Split Rust code into domain modules

src/app/generation.rs

@@ -0,0 +1,252 @@
+use super::*;
+
+#[derive(Clone, Debug)]
+pub(crate) struct ChatMessage {
+    pub(super) id: i32,
+    pub(super) user: bool,
+    pub(super) reasoning: Option<String>,
+    pub(super) reasoning_complete: bool,
+    pub(super) reasoning_open: bool,
+    pub(super) content: String,
+    pub(super) markdown: Vec<markdown::Item>,
+}
+
+impl ChatMessage {
+    pub(super) fn append(&mut self, reasoning: bool, chunk: &str) {
+        if reasoning {
+            self.reasoning.get_or_insert_default().push_str(chunk);
+        } else {
+            self.reasoning_complete = true;
+            self.content.push_str(chunk);
+        }
+    }
+
+    pub(super) fn refresh_markdown(&mut self) {
+        if !self.user {
+            let content = if self.reasoning.is_some() {
+                self.content.trim_start()
+            } else {
+                &self.content
+            };
+            self.markdown = markdown::parse(content).collect();
+        }
+    }
+}
+
+impl From<StoredMessage> for ChatMessage {
+    fn from(message: StoredMessage) -> Self {
+        let mut message = Self {
+            id: message.id,
+            user: message.user,
+            reasoning: message.reasoning,
+            reasoning_complete: message.reasoning_complete,
+            reasoning_open: false,
+            content: message.content,
+            markdown: Vec::new(),
+        };
+        message.refresh_markdown();
+        message
+    }
+}
+
+impl App {
+    pub(super) fn start_generation(&mut self) {
+        if self.generating || self.selected_session.is_none() {
+            return;
+        }
+        let prompt = self.composer.trim().to_owned();
+        if prompt.is_empty() {
+            return;
+        }
+        let model = match ModelChoice::from_id(&self.preferences.selected_model) {
+            Some(model) => model,
+            None => {
+                self.error = Some("The selected model is not supported.".into());
+                return;
+            }
+        };
+        let effective = self.preferences.generation().and_then(|generation| {
+            self.preferences.runtime().and_then(|runtime| {
+                crate::settings::effective_settings(model, &generation, &runtime, &models_path())
+            })
+        });
+        let effective = match effective {
+            Ok(settings) => settings,
+            Err(error) => {
+                self.error = Some(error);
+                return;
+            }
+        };
+        let assistant_reasoning = effective.turn.reasoning_mode != ReasoningMode::Direct;
+        let session_id = self
+            .selected_session
+            .expect("a selected session was checked");
+        #[cfg(target_os = "macos")]
+        let mut messages = self
+            .conversation
+            .iter()
+            .map(|message| ChatTurn {
+                user: message.user,
+                skip_previous_eos: false,
+                reasoning: message.reasoning.clone(),
+                reasoning_complete: message.reasoning_complete,
+                content: message.content.clone(),
+            })
+            .collect::<Vec<_>>();
+        #[cfg(target_os = "macos")]
+        messages.push(ChatTurn {
+            user: true,
+            skip_previous_eos: false,
+            reasoning: None,
+            reasoning_complete: true,
+            content: prompt.clone(),
+        });
+
+        #[cfg(target_os = "macos")]
+        {
+            let Some(service) = &self.generation_service else {
+                self.error = Some("The model runtime is unavailable.".into());
+                return;
+            };
+            let Some(database) = &mut self.database else {
+                return;
+            };
+            let saved = match database.start_chat_turn(session_id, &prompt, assistant_reasoning) {
+                Ok(turn) => turn,
+                Err(error) => {
+                    self.error = Some(format!("Could not save the chat turn: {error}"));
+                    return;
+                }
+            };
+            let idle_timeout =
+                Duration::from_secs(self.preferences.idle_timeout_minutes.max(1) as u64 * 60);
+            self.active_generation = match service.generate(
+                effective.engine,
+                effective.turn,
+                messages,
+                CheckpointTarget::Local(session_checkpoint_path(session_id)),
+                idle_timeout,
+            ) {
+                Ok(active) => Some(active),
+                Err(error) => {
+                    self.generation_service = None;
+                    self.error = Some(error);
+                    return;
+                }
+            };
+            let user = ChatMessage::from(saved.0);
+            let mut assistant = ChatMessage::from(saved.1);
+            assistant.reasoning_open = assistant_reasoning;
+            self.composer.clear();
+            self.conversation.push(user);
+            self.conversation.push(assistant);
+            self.generating = true;
+            self.tokens_per_second = None;
+            self.error = None;
+        }
+        #[cfg(not(target_os = "macos"))]
+        {
+            let _ = effective;
+            self.error = Some("Local Metal generation requires macOS.".into());
+            return;
+        }
+    }
+
+    pub(super) fn poll_generation(&mut self) -> bool {
+        #[cfg(target_os = "macos")]
+        let Some(active) = &mut self.active_generation else {
+            self.generating = false;
+            return false;
+        };
+        #[cfg(target_os = "macos")]
+        let mut transcript_changed = false;
+        #[cfg(target_os = "macos")]
+        let mut context_changed = false;
+        #[cfg(target_os = "macos")]
+        loop {
+            match active.events.try_recv() {
+                Ok(GenerationEvent::Loading) => {}
+                Ok(GenerationEvent::Chunk { reasoning, content }) => {
+                    if let Some(message) = self.conversation.last_mut()
+                        && !message.user
+                    {
+                        message.append(reasoning, &content);
+                        transcript_changed = true;
+                    }
+                }
+                Ok(GenerationEvent::Context {
+                    used,
+                    limit,
+                    tokens_per_second,
+                }) => {
+                    self.context_used = used;
+                    self.context_limit = limit;
+                    self.tokens_per_second = tokens_per_second;
+                    context_changed = true;
+                }
+                Ok(GenerationEvent::Finished(result)) => {
+                    self.generating = false;
+                    if let Err(error) = result {
+                        self.error = Some(error);
+                    }
+                    self.active_generation = None;
+                    break;
+                }
+                Err(TryRecvError::Empty) => break,
+                Err(TryRecvError::Disconnected) => {
+                    self.generating = false;
+                    self.active_generation = None;
+                    self.error = Some("The model runtime stopped unexpectedly.".into());
+                    break;
+                }
+            }
+        }
+        #[cfg(target_os = "macos")]
+        if transcript_changed && let Some(message) = self.conversation.last_mut() {
+            message.refresh_markdown();
+        }
+        #[cfg(target_os = "macos")]
+        if transcript_changed
+            && let Some(message) = self.conversation.last()
+            && let Some(database) = &mut self.database
+            && let Err(error) = database.update_message(
+                message.id,
+                message.reasoning.as_deref(),
+                message.reasoning_complete,
+                &message.content,
+            )
+        {
+            if let Some(active) = &self.active_generation {
+                active.cancel.store(true, Ordering::Relaxed);
+            }
+            self.error = Some(format!("Could not save generated chat text: {error}"));
+        }
+        #[cfg(target_os = "macos")]
+        if context_changed
+            && let Some(session_id) = self.selected_session
+            && let Some(database) = &mut self.database
+        {
+            if let Err(error) = database.update_session_context(
+                session_id,
+                self.context_used,
+                self.context_limit,
+                self.tokens_per_second,
+            ) {
+                self.error = Some(format!("Could not save context usage: {error}"));
+            } else if let Some(session) = self
+                .projects
+                .iter_mut()
+                .flat_map(|project| &mut project.sessions)
+                .find(|session| session.id == session_id)
+            {
+                session.context_used = self.context_used as i32;
+                session.context_limit = self.context_limit as i32;
+                session.last_tokens_per_second = self.tokens_per_second;
+            }
+        }
+        #[cfg(target_os = "macos")]
+        return transcript_changed;
+        #[cfg(not(target_os = "macos"))]
+        false
+    }
+}

src/app/model_manager.rs

@@ -0,0 +1,173 @@
+use super::*;
+
+#[derive(Debug)]
+pub(super) enum ModelDownload {
+    Idle,
+    Active(ActiveDownload),
+    Complete(ManagedArtifactId, ModelOperation, DownloadProgress),
+    Failed(ManagedArtifactId, String, DownloadProgress),
+}
+
+#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+pub(super) enum ModelOperation {
+    Download,
+    Validate,
+}
+
+#[derive(Debug)]
+pub(super) struct ActiveDownload {
+    pub(super) artifact: ManagedArtifactId,
+    pub(super) operation: ModelOperation,
+    pub(super) progress: DownloadProgress,
+    pub(super) sampled_at: Instant,
+    pub(super) sampled_bytes: u64,
+    pub(super) bytes_per_second: f64,
+    pub(super) verified_bytes: Arc<AtomicU64>,
+    pub(super) cancel: Arc<AtomicBool>,
+    pub(super) result: mpsc::Receiver<Result<DownloadOutcome, String>>,
+    pub(super) stopping: bool,
+}
+
+impl App {
+    pub(super) fn open_model_manager(&mut self) -> Task<Message> {
+        if let Some(id) = self.model_manager_window {
+            return window::gain_focus(id);
+        }
+        let (id, open) = window::open(window::Settings {
+            size: Size::new(760.0, 560.0),
+            min_size: Some(Size::new(620.0, 420.0)),
+            icon: Some(app_icon()),
+            ..Default::default()
+        });
+        self.model_manager_window = Some(id);
+        open.map(Message::ModelManagerOpened)
+    }
+
+    pub(super) fn start_model_operation(
+        &mut self,
+        artifact: ManagedArtifactId,
+        operation: ModelOperation,
+    ) {
+        if matches!(self.model_download, ModelDownload::Active(_)) {
+            return;
+        }
+        let models_path = models_path();
+        let progress = match operation {
+            ModelOperation::Download => model::artifact_download_progress(artifact, &models_path),
+            ModelOperation::Validate => model::artifact_verification_progress(artifact, 0),
+        };
+        let cancel = Arc::new(AtomicBool::new(false));
+        let worker_cancel = Arc::clone(&cancel);
+        let verified_bytes = Arc::new(AtomicU64::new(0));
+        let worker_verified_bytes = Arc::clone(&verified_bytes);
+        let (result_sender, result_receiver) = mpsc::channel();
+        let thread_name = match operation {
+            ModelOperation::Download => "model-download",
+            ModelOperation::Validate => "model-validation",
+        };
+        if let Err(error) = thread::Builder::new()
+            .name(thread_name.to_owned())
+            .spawn(move || {
+                let result = match operation {
+                    ModelOperation::Download => model::download_managed_artifact(
+                        artifact,
+                        &models_path,
+                        &worker_cancel,
+                        &worker_verified_bytes,
+                    ),
+                    ModelOperation::Validate => model::validate_managed_artifact(
+                        artifact,
+                        &models_path,
+                        &worker_cancel,
+                        &worker_verified_bytes,
+                    ),
+                };
+                let _ = result_sender.send(result);
+            })
+        {
+            self.error = Some(format!("Could not start {thread_name}: {error}"));
+            return;
+        }
+        self.pending_model_delete = None;
+        self.model_download = ModelDownload::Active(ActiveDownload {
+            artifact,
+            operation,
+            sampled_at: Instant::now(),
+            sampled_bytes: progress.completed(),
+            progress,
+            bytes_per_second: 0.0,
+            verified_bytes,
+            cancel,
+            result: result_receiver,
+            stopping: false,
+        });
+    }
+}
+
+impl App {
+    pub(super) fn update_download_progress(&mut self) {
+        let ModelDownload::Active(download) = &mut self.model_download else {
+            return;
+        };
+        let now = Instant::now();
+        let verified = download.verified_bytes.load(Ordering::Relaxed);
+        let mut progress = match download.operation {
+            ModelOperation::Download => {
+                model::artifact_download_progress(download.artifact, &models_path())
+            }
+            ModelOperation::Validate => {
+                model::artifact_verification_progress(download.artifact, verified)
+            }
+        };
+        if download.operation == ModelOperation::Download
+            && let Some(verification) = &mut progress.verification
+        {
+            verification.verified = verified.min(verification.total);
+        }
+        let elapsed = now.duration_since(download.sampled_at).as_secs_f64();
+        let completed = progress.completed();
+        let phase_changed = progress.phase != download.progress.phase;
+        let transferred = completed.saturating_sub(download.sampled_bytes);
+        if phase_changed {
+            download.bytes_per_second = 0.0;
+        } else if transferred > 0 && elapsed > 0.0 {
+            let current = transferred as f64 / elapsed;
+            download.bytes_per_second = if download.bytes_per_second == 0.0 {
+                current
+            } else {
+                download.bytes_per_second * 0.75 + current * 0.25
+            };
+        }
+        download.progress = progress;
+        download.sampled_at = now;
+        download.sampled_bytes = completed;
+
+        let result = match download.result.try_recv() {
+            Ok(result) => Some(result),
+            Err(TryRecvError::Empty) => None,
+            Err(TryRecvError::Disconnected) => {
+                Some(Err("Download worker stopped unexpectedly.".into()))
+            }
+        };
+        let artifact = download.artifact;
+        let operation = download.operation;
+        let progress = download.progress.clone();
+        if let Some(result) = result {
+            match result {
+                Ok(DownloadOutcome::Complete) => {
+                    let progress = model::artifact_download_progress(artifact, &models_path());
+                    self.model_download = ModelDownload::Complete(artifact, operation, progress);
+                    self.error = None;
+                }
+                Ok(DownloadOutcome::Stopped) => {
+                    self.model_download = ModelDownload::Idle;
+                    self.error = None;
+                }
+                Err(error) => {
+                    self.error = Some(error.clone());
+                    self.model_download = ModelDownload::Failed(artifact, error, progress);
+                }
+            }
+        }
+    }
+}

src/app/preferences.rs

@@ -0,0 +1,476 @@
+use super::*;
+
+#[derive(Clone)]
+pub(super) struct PreferenceDraft {
+    pub(super) model: ModelChoice,
+    pub(super) dspark_enabled: bool,
+    pub(super) idle_timeout_minutes: String,
+    pub(super) endpoint_port: String,
+    pub(super) context_tokens: String,
+    pub(super) max_generated_tokens: String,
+    pub(super) system_prompt: String,
+    pub(super) temperature: String,
+    pub(super) top_p: String,
+    pub(super) min_p: String,
+    pub(super) seed: String,
+    pub(super) reasoning_mode: ReasoningMode,
+    pub(super) cpu_threads: String,
+    pub(super) power_percent: String,
+    pub(super) prefill_chunk: String,
+    pub(super) quality: bool,
+    pub(super) warm_weights: bool,
+    pub(super) mtp_draft_tokens: String,
+    pub(super) mtp_margin: String,
+    pub(super) glm_mtp: bool,
+    pub(super) glm_mtp_timing: bool,
+    pub(super) dspark_confidence_threshold: String,
+    pub(super) dspark_strict: bool,
+    pub(super) ssd_streaming: bool,
+    pub(super) ssd_streaming_cold: bool,
+    pub(super) ssd_cache: String,
+    pub(super) ssd_full_layers: String,
+    pub(super) ssd_preload_experts: String,
+    pub(super) directional_steering_file: String,
+    pub(super) directional_steering_ffn: String,
+    pub(super) directional_steering_attn: String,
+    pub(super) simulated_used_memory_gib: String,
+    pub(super) expert_profile_path: String,
+}
+
+impl PreferenceDraft {
+    pub(super) fn from_saved(preferences: &AppPreferences) -> Result<Self, String> {
+        let model = ModelChoice::from_id(&preferences.selected_model)
+            .ok_or_else(|| format!("Unsupported model: {}", preferences.selected_model))?;
+        let generation = preferences.generation()?;
+        let runtime = preferences.runtime()?;
+        runtime.validate(model)?;
+        let execution = &runtime.execution;
+        let speculative = &runtime.speculative;
+        Ok(Self {
+            model,
+            dspark_enabled: speculative.dspark_enabled,
+            idle_timeout_minutes: preferences.idle_timeout_minutes.to_string(),
+            endpoint_port: preferences.endpoint_port.to_string(),
+            context_tokens: generation.context_tokens.to_string(),
+            max_generated_tokens: generation.max_generated_tokens.to_string(),
+            system_prompt: generation.system_prompt,
+            temperature: generation
+                .temperature
+                .map_or_else(String::new, |value| value.to_string()),
+            top_p: generation
+                .top_p
+                .map_or_else(String::new, |value| value.to_string()),
+            min_p: generation
+                .min_p
+                .map_or_else(String::new, |value| value.to_string()),
+            seed: generation
+                .seed
+                .map_or_else(String::new, |value| value.to_string()),
+            reasoning_mode: generation.reasoning_mode,
+            cpu_threads: execution
+                .cpu_threads
+                .map_or_else(String::new, |value| value.to_string()),
+            power_percent: execution
+                .power_percent
+                .map_or_else(String::new, |value| value.to_string()),
+            prefill_chunk: execution
+                .prefill_chunk
+                .map_or_else(String::new, |value| value.to_string()),
+            quality: execution.quality,
+            warm_weights: execution.warm_weights,
+            mtp_draft_tokens: speculative.mtp_draft_tokens.to_string(),
+            mtp_margin: speculative.mtp_margin.to_string(),
+            glm_mtp: speculative.glm_mtp,
+            glm_mtp_timing: speculative.glm_mtp_timing,
+            dspark_confidence_threshold: speculative
+                .dspark_confidence_threshold
+                .map_or_else(String::new, |value| value.to_string()),
+            dspark_strict: speculative.dspark_strict,
+            ssd_streaming: runtime.ssd.enabled,
+            ssd_streaming_cold: runtime.ssd.cold,
+            ssd_cache: runtime
+                .ssd
+                .cache
+                .map_or_else(String::new, |cache| match cache {
+                    StreamingCacheBudget::Experts(experts) => experts.to_string(),
+                    StreamingCacheBudget::Gib(gib) => format!("{gib}GB"),
+                }),
+            ssd_full_layers: runtime
+                .ssd
+                .full_layers
+                .map_or_else(String::new, |value| value.to_string()),
+            ssd_preload_experts: runtime
+                .ssd
+                .preload_experts
+                .map_or_else(String::new, |value| value.to_string()),
+            directional_steering_file: runtime.steering.file.unwrap_or_default(),
+            directional_steering_ffn: runtime
+                .steering
+                .ffn_scale
+                .map_or_else(String::new, |value| value.to_string()),
+            directional_steering_attn: runtime
+                .steering
+                .attention_scale
+                .map_or_else(String::new, |value| value.to_string()),
+            simulated_used_memory_gib: runtime
+                .diagnostics
+                .simulated_used_memory_gib
+                .map_or_else(String::new, |value| value.to_string()),
+            expert_profile_path: runtime.diagnostics.expert_profile_path.unwrap_or_default(),
+        })
+    }
+
+    pub(super) fn generation(&self) -> Result<GenerationPreferences, String> {
+        let preferences = GenerationPreferences {
+            context_tokens: parse_positive_i32("Context tokens", &self.context_tokens)?,
+            max_generated_tokens: parse_positive_i32(
+                "Maximum generated tokens",
+                &self.max_generated_tokens,
+            )?,
+            system_prompt: self.system_prompt.clone(),
+            temperature: parse_optional_f32("Temperature", &self.temperature)?,
+            top_p: parse_optional_f32("Top-p", &self.top_p)?,
+            min_p: parse_optional_f32("Min-p", &self.min_p)?,
+            seed: parse_optional_u64("Seed", &self.seed)?,
+            reasoning_mode: self.reasoning_mode,
+        };
+        preferences.validate()?;
+        Ok(preferences)
+    }
+
+    pub(super) fn reset(&mut self) {
+        let defaults = GenerationPreferences::default();
+        let execution = ExecutionPreferences::default();
+        let speculative = SpeculativePreferences::default();
+        let ssd = SsdPreferences::default();
+        self.model = ModelChoice::default();
+        self.dspark_enabled = false;
+        self.idle_timeout_minutes = "10".into();
+        self.endpoint_port = "4000".into();
+        self.context_tokens = defaults.context_tokens.to_string();
+        self.max_generated_tokens = defaults.max_generated_tokens.to_string();
+        self.system_prompt = defaults.system_prompt;
+        self.temperature.clear();
+        self.top_p.clear();
+        self.min_p.clear();
+        self.seed.clear();
+        self.reasoning_mode = defaults.reasoning_mode;
+        self.cpu_threads.clear();
+        self.power_percent.clear();
+        self.prefill_chunk.clear();
+        self.quality = execution.quality;
+        self.warm_weights = execution.warm_weights;
+        self.mtp_draft_tokens = speculative.mtp_draft_tokens.to_string();
+        self.mtp_margin = speculative.mtp_margin.to_string();
+        self.glm_mtp = speculative.glm_mtp;
+        self.glm_mtp_timing = speculative.glm_mtp_timing;
+        self.dspark_confidence_threshold.clear();
+        self.dspark_strict = speculative.dspark_strict;
+        self.ssd_streaming = ssd.enabled;
+        self.ssd_streaming_cold = ssd.cold;
+        self.ssd_cache.clear();
+        self.ssd_full_layers.clear();
+        self.ssd_preload_experts.clear();
+        self.directional_steering_file.clear();
+        self.directional_steering_ffn.clear();
+        self.directional_steering_attn.clear();
+        self.simulated_used_memory_gib.clear();
+        self.expert_profile_path.clear();
+    }
+
+    pub(super) fn execution(&self) -> Result<ExecutionPreferences, String> {
+        Ok(ExecutionPreferences {
+            cpu_threads: parse_optional_u32("CPU helper threads", &self.cpu_threads)?,
+            power_percent: parse_optional_u8("GPU power", &self.power_percent)?,
+            prefill_chunk: parse_optional_u32("Prefill chunk", &self.prefill_chunk)?,
+            quality: self.quality,
+            warm_weights: self.warm_weights,
+        })
+    }
+
+    pub(super) fn speculative(&self) -> Result<SpeculativePreferences, String> {
+        Ok(SpeculativePreferences {
+            mtp_draft_tokens: parse_positive_i32("MTP draft tokens", &self.mtp_draft_tokens)?,
+            mtp_margin: parse_f32("MTP margin", &self.mtp_margin)?,
+            glm_mtp: self.glm_mtp,
+            glm_mtp_timing: self.glm_mtp_timing,
+            dspark_enabled: self.dspark_enabled,
+            dspark_confidence_threshold: parse_optional_f32(
+                "DSpark confidence",
+                &self.dspark_confidence_threshold,
+            )?,
+            dspark_strict: self.dspark_strict,
+        })
+    }
+
+    pub(super) fn runtime(&self) -> Result<RuntimePreferences, String> {
+        Ok(RuntimePreferences {
+            execution: self.execution()?,
+            speculative: self.speculative()?,
+            ssd: SsdPreferences {
+                enabled: self.ssd_streaming,
+                cold: self.ssd_streaming_cold,
+                cache: parse_streaming_cache(&self.ssd_cache)?,
+                full_layers: parse_optional_u32("SSD full-layer count", &self.ssd_full_layers)?,
+                preload_experts: parse_optional_u32(
+                    "SSD preload experts",
+                    &self.ssd_preload_experts,
+                )?,
+            },
+            steering: SteeringPreferences {
+                file: optional_text(&self.directional_steering_file),
+                ffn_scale: parse_optional_f32(
+                    "Directional FFN scale",
+                    &self.directional_steering_ffn,
+                )?,
+                attention_scale: parse_optional_f32(
+                    "Directional attention scale",
+                    &self.directional_steering_attn,
+                )?,
+            },
+            diagnostics: DiagnosticPreferences {
+                simulated_used_memory_gib: parse_optional_gib(
+                    "Simulated used memory",
+                    &self.simulated_used_memory_gib,
+                )?,
+                expert_profile_path: optional_text(&self.expert_profile_path),
+            },
+        })
+    }
+}
+
+fn parse_positive_i32(name: &str, value: &str) -> Result<i32, String> {
+    value
+        .trim()
+        .parse::<i32>()
+        .ok()
+        .filter(|value| *value > 0)
+        .ok_or_else(|| format!("{name} must be a positive whole number."))
+}
+
+fn parse_optional_f32(name: &str, value: &str) -> Result<Option<f32>, String> {
+    let value = value.trim();
+    if value.is_empty() {
+        Ok(None)
+    } else {
+        value
+            .parse()
+            .map(Some)
+            .map_err(|_| format!("{name} must be a number or left blank for the DS4 default."))
+    }
+}
+
+fn parse_f32(name: &str, value: &str) -> Result<f32, String> {
+    value
+        .trim()
+        .parse()
+        .map_err(|_| format!("{name} must be a number."))
+}
+
+fn parse_optional_u64(name: &str, value: &str) -> Result<Option<u64>, String> {
+    let value = value.trim();
+    if value.is_empty() {
+        Ok(None)
+    } else {
+        value
+            .parse()
+            .map(Some)
+            .map_err(|_| format!("{name} must be a positive whole number or left blank."))
+    }
+}
+
+fn parse_optional_u32(name: &str, value: &str) -> Result<Option<u32>, String> {
+    parse_optional_number(name, value)
+}
+
+fn parse_optional_u8(name: &str, value: &str) -> Result<Option<u8>, String> {
+    parse_optional_number(name, value)
+}
+
+fn parse_optional_number<T: std::str::FromStr>(
+    name: &str,
+    value: &str,
+) -> Result<Option<T>, String> {
+    let value = value.trim();
+    if value.is_empty() {
+        Ok(None)
+    } else {
+        value
+            .parse()
+            .map(Some)
+            .map_err(|_| format!("{name} must be a positive whole number or left blank."))
+    }
+}
+
+pub(super) fn parse_streaming_cache(value: &str) -> Result<Option<StreamingCacheBudget>, String> {
+    let value = value.trim();
+    if value.is_empty() {
+        return Ok(None);
+    }
+    if value.len() > 2
+        && value
+            .get(value.len() - 2..)
+            .is_some_and(|suffix| suffix.eq_ignore_ascii_case("gb"))
+    {
+        return parse_gib("SSD cache budget", value)
+            .map(|gib| Some(StreamingCacheBudget::Gib(gib)));
+    }
+    if !value.chars().all(|character| character.is_ascii_digit()) {
+        return Err(
+            "SSD cache budget must be a positive expert count or whole GiB value such as 64GB."
+                .into(),
+        );
+    }
+    value
+        .parse::<u32>()
+        .ok()
+        .filter(|value| *value > 0)
+        .map(StreamingCacheBudget::Experts)
+        .map(Some)
+        .ok_or_else(|| {
+            "SSD cache budget must be a positive expert count or whole GiB value such as 64GB."
+                .into()
+        })
+}
+
+pub(super) fn parse_optional_gib(name: &str, value: &str) -> Result<Option<u64>, String> {
+    let value = value.trim();
+    if value.is_empty() {
+        Ok(None)
+    } else {
+        parse_gib(name, value).map(Some)
+    }
+}
+
+fn parse_gib(name: &str, value: &str) -> Result<u64, String> {
+    let value = value
+        .get(value.len().saturating_sub(2)..)
+        .filter(|suffix| suffix.eq_ignore_ascii_case("gb"))
+        .map_or(value, |_| &value[..value.len() - 2]);
+    if !value.chars().all(|character| character.is_ascii_digit()) {
+        return Err(format!("{name} must be a positive whole GiB value."));
+    }
+    value
+        .parse::<u64>()
+        .ok()
+        .filter(|value| *value > 0 && *value <= u64::MAX / GIB)
+        .ok_or_else(|| format!("{name} must be a positive whole GiB value."))
+}
+
+fn optional_text(value: &str) -> Option<String> {
+    let value = value.trim();
+    (!value.is_empty()).then(|| value.to_owned())
+}
+
+impl App {
+    pub(super) fn open_preferences(&mut self) {
+        if self.database.is_none() || self.pending_project_path.is_some() || self.choosing_folder {
+            return;
+        }
+        match PreferenceDraft::from_saved(&self.preferences) {
+            Ok(draft) => {
+                self.preference_draft = draft;
+                self.preference_error = None;
+                self.preferences_open = true;
+            }
+            Err(error) => self.error = Some(error),
+        }
+    }
+
+    pub(super) fn save_preferences(&mut self) {
+        let Ok(idle_timeout_minutes) = self
+            .preference_draft
+            .idle_timeout_minutes
+            .trim()
+            .parse::<i32>()
+        else {
+            self.preference_error = Some("Idle timeout must be a whole number.".into());
+            return;
+        };
+        if !(1..=1440).contains(&idle_timeout_minutes) {
+            self.preference_error = Some("Idle timeout must be between 1 and 1440 minutes.".into());
+            return;
+        }
+        let Ok(endpoint_port) = self.preference_draft.endpoint_port.trim().parse::<u16>() else {
+            self.preference_error = Some("Endpoint port must be a whole number.".into());
+            return;
+        };
+        if endpoint_port == 0 {
+            self.preference_error = Some("Endpoint port must be between 1 and 65535.".into());
+            return;
+        }
+        let generation = match self.preference_draft.generation() {
+            Ok(generation) => generation,
+            Err(error) => {
+                self.preference_error = Some(error);
+                return;
+            }
+        };
+
+        let model = self.preference_draft.model;
+        let runtime = match self.preference_draft.runtime() {
+            Ok(runtime) => runtime,
+            Err(error) => {
+                self.preference_error = Some(error);
+                return;
+            }
+        };
+        if let Err(error) = runtime.validate(model) {
+            self.preference_error = Some(error);
+            return;
+        }
+        #[cfg(target_os = "macos")]
+        let pending_endpoint = if self.preferences.endpoint_port != i32::from(endpoint_port)
+            || self._endpoint.is_none()
+        {
+            let Some(generation) = &self.generation_service else {
+                self.preference_error = Some("The model runtime is unavailable.".into());
+                return;
+            };
+            match crate::server::ServerHandle::spawn(
+                generation.clone(),
+                Arc::clone(&self.runtime_preferences),
+                models_path(),
+                application_support_path().join("kv-cache").join("http"),
+                endpoint_port,
+                Arc::clone(&self.metrics),
+            ) {
+                Ok(endpoint) => Some(endpoint),
+                Err(error) => {
+                    self.preference_error = Some(error);
+                    return;
+                }
+            }
+        } else {
+            None
+        };
+        let Some(database) = &mut self.database else {
+            return;
+        };
+        match database.update_preferences(
+            model.id(),
+            idle_timeout_minutes,
+            i32::from(endpoint_port),
+            &generation,
+            &runtime,
+        ) {
+            Ok(preferences) => {
+                self.preferences = preferences;
+                #[cfg(target_os = "macos")]
+                if let Ok(mut runtime_preferences) = self.runtime_preferences.write() {
+                    *runtime_preferences = self.preferences.clone();
+                }
+                #[cfg(target_os = "macos")]
+                if let Some(endpoint) = pending_endpoint {
+                    self._endpoint = Some(endpoint);
+                }
+                self.preference_draft = PreferenceDraft::from_saved(&self.preferences)
+                    .expect("the saved model was selected from the supported catalog");
+                self.preferences_open = false;
+                self.preference_error = None;
+                self.error = None;
+            }
+            Err(error) => self.preference_error = Some(error),
+        }
+    }
+}

src/app/projects.rs

@@ -0,0 +1,102 @@
+use super::*;
+
+impl App {
+    pub(super) fn prepare_project(&mut self, path: PathBuf) {
+        let Ok(path) = fs::canonicalize(path) else {
+            self.error = Some("The selected folder is no longer available.".into());
+            return;
+        };
+        let Some(path_text) = path.to_str() else {
+            self.error = Some("The selected folder path is not valid UTF-8.".into());
+            return;
+        };
+        if self
+            .projects
+            .iter()
+            .any(|item| item.project.path == path_text)
+        {
+            self.error = Some("That project is already in the sidebar.".into());
+            return;
+        }
+
+        self.project_name_input = path
+            .file_name()
+            .and_then(|name| name.to_str())
+            .unwrap_or("Project")
+            .to_owned();
+        self.pending_project_path = Some(path);
+        self.error = None;
+    }
+
+    pub(super) fn create_project(&mut self) {
+        let name = self.project_name_input.trim();
+        if name.is_empty() {
+            self.error = Some("Project name cannot be empty.".into());
+            return;
+        }
+        let Some(path) = &self.pending_project_path else {
+            return;
+        };
+        let Some(path) = path.to_str() else {
+            self.error = Some("The selected folder path is not valid UTF-8.".into());
+            return;
+        };
+        let Some(database) = &mut self.database else {
+            return;
+        };
+
+        match database.create_project(name, path) {
+            Ok(project) => {
+                self.selected_project = Some(project.id);
+                self.selected_session = None;
+                self.pending_project_path = None;
+                self.project_name_input.clear();
+                self.error = None;
+                self.reload_projects();
+            }
+            Err(error) => self.error = Some(error),
+        }
+    }
+
+    pub(super) fn create_session(&mut self) {
+        if self.generating {
+            self.error = Some("Stop the active generation before creating a session.".into());
+            return;
+        }
+        let Some(project_id) = self.selected_project else {
+            self.error = Some("Select a project first.".into());
+            return;
+        };
+        let default_number = self
+            .selected_project()
+            .map(|project| project.sessions.len() + 1)
+            .unwrap_or(1);
+        let title = format!("Session {default_number}");
+        let Some(database) = &mut self.database else {
+            return;
+        };
+
+        match database.create_session(project_id, &title) {
+            Ok(session) => {
+                self.selected_session = Some(session.id);
+                self.conversation.clear();
+                self.composer.clear();
+                self.context_used = 0;
+                self.context_limit = self.preferences.context_tokens.max(0) as u32;
+                self.tokens_per_second = None;
+                self.error = None;
+                self.reload_projects();
+            }
+            Err(error) => self.error = Some(error),
+        }
+    }
+
+    pub(super) fn reload_projects(&mut self) {
+        if let Some(database) = &mut self.database {
+            match database.load_projects() {
+                Ok(projects) => self.projects = projects,
+                Err(error) => self.error = Some(error),
+            }
+        }
+    }
+}

src/app/view/chat.rs

@@ -0,0 +1,210 @@
+use super::*;
+use iced::widget::column;
+
+impl App {
+    pub(super) fn chat_detail(&self) -> Element<'_, Message> {
+        let Some(item) = self.selected_project() else {
+            let open_project_content = row![icon(ICON_FOLDER_PLUS, 17), text("Open project…"),]
+                .spacing(8)
+                .align_y(Alignment::Center);
+            let open_project = if self.database.is_some() && !self.choosing_folder {
+                action_button(open_project_content).on_press(Message::ChooseProjectFolder)
+            } else {
+                action_button(open_project_content)
+            };
+            return container(
+                column![
+                    icon(ICON_SPARK, 36),
+                    text("Start a local coding session").size(28),
+                    text("Choose a project folder to create your first session.").size(14),
+                    Space::with_height(10),
+                    open_project,
+                ]
+                .spacing(10)
+                .align_x(Alignment::Center),
+            )
+            .center_x(Length::Fill)
+            .center_y(Length::Fill)
+            .into();
+        };
+
+        let project = &item.project;
+        let selected_title = self
+            .selected_session(item)
+            .map(|session| session.title.as_str())
+            .unwrap_or(project.name.as_str());
+        let header = row![
+            icon(ICON_FOLDER, 19),
+            text(selected_title).size(18),
+            icon(ICON_MORE, 18),
+            Space::with_width(Length::Fill),
+        ]
+        .spacing(10)
+        .align_y(Alignment::Center);
+
+        let body: Element<'_, Message> = if let Some(session) = self.selected_session(item) {
+            let mut messages = column![].spacing(12);
+            if self.conversation.is_empty() {
+                messages = messages.push(
+                    column![
+                        text(&session.title).size(26),
+                        text("Run DeepSeek locally with the Rust Metal engine.").size(14),
+                    ]
+                    .spacing(8),
+                );
+            } else {
+                let markdown_style = markdown::Style::from_palette(app_theme().palette());
+                for (index, message) in self.conversation.iter().enumerate() {
+                    let label = if message.user { "You" } else { "DS4" };
+                    let active = self.generating && index + 1 == self.conversation.len();
+                    let mut body = column![text(label).size(11)].spacing(5);
+                    if let Some(reasoning) = &message.reasoning {
+                        let reasoning_label =
+                            match (message.reasoning_open, message.reasoning_complete, active) {
+                                (true, false, true) => "▾ Thinking",
+                                (false, false, true) => "› Thinking",
+                                (true, false, false) => "▾ Reasoning (stopped)",
+                                (false, false, false) => "› Reasoning (stopped)",
+                                (true, true, _) => "▾ Reasoning",
+                                (false, true, _) => "› Reasoning",
+                            };
+                        body = body.push(
+                            button(text(reasoning_label).size(12))
+                                .padding(0)
+                                .style(button::text)
+                                .on_press(Message::ToggleReasoning(index)),
+                        );
+                        if message.reasoning_open {
+                            body = body.push(
+                                text(if reasoning.is_empty() && active {
+                                    "Thinking…"
+                                } else {
+                                    reasoning
+                                })
+                                .size(13)
+                                .color(muted_text()),
+                            );
+                        }
+                    }
+                    if !message.content.is_empty() {
+                        if message.user || message.markdown.is_empty() {
+                            let content = if message.reasoning.is_some() {
+                                message.content.trim_start()
+                            } else {
+                                &message.content
+                            };
+                            body = body.push(text(content).size(14));
+                        } else {
+                            body = body.push(
+                                markdown::view(
+                                    &message.markdown,
+                                    markdown::Settings::with_text_size(14),
+                                    markdown_style,
+                                )
+                                .map(Message::OpenLink),
+                            );
+                        }
+                    } else if active && message.reasoning.is_none() {
+                        body = body.push(text("Loading model…").size(14));
+                    }
+                    let user = message.user;
+                    messages = messages.push(
+                        container(body)
+                            .padding(14)
+                            .width(Length::Fill)
+                            .style(move |theme| chat_message_style(theme, user)),
+                    );
+                }
+            }
+            let composer = text_input("Ask DS4Server anything…", &self.composer)
+                .on_input(Message::ComposerChanged)
+                .on_submit(Message::SubmitPrompt)
+                .padding(12)
+                .size(14);
+            let action = if self.generating {
+                action_button(text("Stop").size(12)).on_press(Message::StopGeneration)
+            } else if self.composer.trim().is_empty() {
+                action_button(icon(ICON_SEND, 18)).padding(8)
+            } else {
+                action_button(icon(ICON_SEND, 18))
+                    .padding(8)
+                    .on_press(Message::SubmitPrompt)
+            };
+            let context_fraction = if self.context_limit == 0 {
+                0.0
+            } else {
+                self.context_used.min(self.context_limit) as f32 / self.context_limit as f32
+            };
+            let conversation = column![
+                scrollable(messages)
+                    .id(chat_scroll_id())
+                    .height(Length::Fill),
+                container(
+                    column![
+                        composer,
+                        progress_bar(0.0..=1.0, context_fraction).height(3),
+                        row![
+                            icon(ICON_PAPERCLIP, 19),
+                            text(format!(
+                                "{} / {} tokens ({:.0}%) • {}",
+                                self.context_used,
+                                self.context_limit,
+                                context_fraction * 100.0,
+                                self.tokens_per_second.map_or_else(
+                                    || "— tok/s".to_owned(),
+                                    |speed| format!("{speed:.1} tok/s")
+                                )
+                            ))
+                            .size(11)
+                            .color(muted_text()),
+                            Space::with_width(Length::Fill),
+                            icon(ICON_MODEL, 16),
+                            text(
+                                ModelChoice::from_id(&self.preferences.selected_model)
+                                    .unwrap_or_default()
+                                    .to_string(),
+                            )
+                            .size(12),
+                            action,
+                        ]
+                        .align_y(Alignment::Center),
+                    ]
+                    .spacing(8),
+                )
+                .padding(16)
+                .width(Length::Fill)
+                .style(overview_style),
+            ]
+            .height(Length::Fill)
+            .spacing(8);
+            container(conversation)
+                .max_width(860)
+                .center_x(Length::Fill)
+                .height(Length::Fill)
+                .into()
+        } else {
+            container(
+                column![
+                    text(if item.sessions.is_empty() {
+                        "No sessions yet"
+                    } else {
+                        "Choose a session"
+                    })
+                    .size(24),
+                    text("Create a session above or select one from the sidebar.").size(14),
+                ]
+                .spacing(8)
+                .align_x(Alignment::Center),
+            )
+            .center_x(Length::Fill)
+            .center_y(Length::Fill)
+            .into()
+        };
+
+        container(column![header, body].spacing(24))
+            .width(Length::Fill)
+            .height(Length::Fill)
+            .padding(24)
+            .into()
+    }
+}

src/app/view/model_manager.rs

@@ -0,0 +1,296 @@
+use super::*;
+use iced::widget::column;
+
+impl App {
+    pub(super) fn model_manager(&self) -> Element<'_, Message> {
+        let busy = matches!(self.model_download, ModelDownload::Active(_));
+        let mut artifacts = column![];
+        for (index, artifact) in model::managed_artifacts(&models_path()).iter().enumerate() {
+            if index > 0 {
+                artifacts = artifacts.push(horizontal_rule(1));
+            }
+            artifacts = artifacts.push(model_artifact_row(artifact, busy));
+        }
+
+        let mut content = column![
+            text("Model Manager").size(26),
+            text("Download, verify, or remove locally stored model files.")
+                .size(14)
+                .color(muted_text()),
+        ]
+        .spacing(8);
+        if !matches!(self.model_download, ModelDownload::Idle) {
+            content = content.push(
+                container(model_download_status(&self.model_download))
+                    .padding(16)
+                    .style(overview_style),
+            );
+        }
+        if let Some(error) = &self.error {
+            content = content.push(text(error).style(iced::widget::text::danger));
+        }
+        content = content.push(Space::with_height(8)).push(
+            container(scrollable(artifacts).height(Length::Fill))
+                .height(Length::Fill)
+                .style(overview_style),
+        );
+
+        let base: Element<'_, Message> = container(content)
+            .width(Length::Fill)
+            .height(Length::Fill)
+            .padding(28)
+            .into();
+        let Some(artifact) = self.pending_model_delete else {
+            return base;
+        };
+        let confirmation = container(
+            column![
+                text("Delete model file?").size(22),
+                text(format!(
+                    "Delete {artifact}, including any resumable partial download?"
+                ))
+                .size(13),
+                row![
+                    Space::with_width(Length::Fill),
+                    action_button("Cancel").on_press(Message::CancelDeleteArtifact),
+                    danger_button("Delete").on_press(Message::ConfirmDeleteArtifact),
+                ]
+                .spacing(8),
+            ]
+            .spacing(14),
+        )
+        .padding(22)
+        .width(460)
+        .style(overview_style);
+        stack![
+            base,
+            opaque(
+                container(confirmation)
+                    .center_x(Length::Fill)
+                    .center_y(Length::Fill)
+                    .style(|_| container::Style::default()
+                        .background(Color::from_rgba8(0, 0, 0, 0.68)))
+            )
+        ]
+        .into()
+    }
+}
+
+fn model_artifact_row(artifact: &ManagedArtifact, busy: bool) -> Element<'static, Message> {
+    let status = match artifact.state {
+        ManagedArtifactState::Missing => "Not downloaded",
+        ManagedArtifactState::Partial => "Partial download",
+        ManagedArtifactState::NeedsVerification => "Downloaded; verification required",
+        ManagedArtifactState::Ready => "Ready and verified",
+    };
+    let download_label = match artifact.state {
+        ManagedArtifactState::Ready => "Downloaded",
+        _ if artifact.stored > 0 => "Resume",
+        _ => "Download",
+    };
+    let download = if artifact.state == ManagedArtifactState::Ready || busy {
+        action_button(download_label)
+    } else {
+        action_button(download_label).on_press(Message::DownloadArtifact(artifact.id))
+    };
+    let validate = if artifact.can_validate() && !busy {
+        action_button("Validate").on_press(Message::ValidateArtifact(artifact.id))
+    } else {
+        action_button("Validate")
+    };
+    let delete = if artifact.stored > 0 && !busy {
+        danger_button("Delete").on_press(Message::DeleteArtifact(artifact.id))
+    } else {
+        danger_button("Delete")
+    };
+
+    container(
+        row![
+            icon(ICON_MODEL, 28),
+            column![
+                text(artifact.id.to_string()).size(16),
+                text(status).size(13).color(muted_text()),
+                text(format!(
+                    "{} on disk • {} expected",
+                    format_bytes(artifact.stored),
+                    format_bytes(artifact.expected),
+                ))
+                .size(12)
+                .color(muted_text()),
+            ]
+            .spacing(5),
+            Space::with_width(Length::Fill),
+            row![download, validate, delete]
+                .spacing(8)
+                .align_y(Alignment::Center),
+        ]
+        .spacing(16)
+        .align_y(Alignment::Center),
+    )
+    .width(Length::Fill)
+    .padding([18, 20])
+    .into()
+}
+
+pub(super) fn download_status_bar(download: &ActiveDownload) -> Element<'_, Message> {
+    let progress = &download.progress;
+    let percent = progress.fraction() * 100.0;
+    let measurement = if progress.verification.is_some() {
+        format!(
+            "{} verified of {} ({percent:.1}%)",
+            format_bytes(progress.completed()),
+            format_bytes(progress.active_total()),
+        )
+    } else {
+        format!(
+            "{} of {} ({percent:.1}%)",
+            format_bytes(progress.completed()),
+            format_bytes(progress.active_total()),
+        )
+    };
+    let transfer = if download.bytes_per_second > 0.0 && progress.remaining() > 0 {
+        format!(
+            "{}/s • about {} remaining",
+            format_bytes(download.bytes_per_second as u64),
+            format_duration(progress.remaining() as f64 / download.bytes_per_second),
+        )
+    } else {
+        "Calculating time remaining…".to_owned()
+    };
+    let stop = if download.stopping {
+        danger_button("Stopping…")
+    } else {
+        danger_button("Stop").on_press(Message::StopModelDownload)
+    };
+
+    container(
+        row![
+            text(phase_text(progress.phase)).size(12),
+            progress_bar(0.0..=1.0, progress.fraction())
+                .width(180)
+                .height(7),
+            text(measurement).size(12),
+            text(transfer).size(12),
+            Space::with_width(Length::Fill),
+            stop,
+        ]
+        .spacing(12)
+        .align_y(Alignment::Center),
+    )
+    .width(Length::Fill)
+    .padding([7, 14])
+    .style(sidebar_style)
+    .into()
+}
+
+fn model_download_status(download: &ModelDownload) -> Element<'_, Message> {
+    let (progress, heading, speed, failed) = match download {
+        ModelDownload::Idle => unreachable!("idle operations are not displayed"),
+        ModelDownload::Active(active) => (
+            &active.progress,
+            if active.stopping {
+                format!("Stopping {}…", active.artifact)
+            } else {
+                phase_text(active.progress.phase)
+            },
+            active.bytes_per_second,
+            false,
+        ),
+        ModelDownload::Complete(artifact, operation, progress) => (
+            progress,
+            match operation {
+                ModelOperation::Download => format!("{artifact} is downloaded and verified."),
+                ModelOperation::Validate => format!("{artifact} passed validation."),
+            },
+            0.0,
+            false,
+        ),
+        ModelDownload::Failed(artifact, error, progress) => (
+            progress,
+            format!("{artifact} operation failed: {error}"),
+            0.0,
+            true,
+        ),
+    };
+
+    let percent = progress.fraction() * 100.0;
+    let measurement = if progress.verification.is_some() {
+        format!(
+            "{} verified of {} ({percent:.1}%) • {} remaining",
+            format_bytes(progress.completed()),
+            format_bytes(progress.active_total()),
+            format_bytes(progress.remaining()),
+        )
+    } else {
+        format!(
+            "{} of {} ({percent:.1}%) • {} remaining",
+            format_bytes(progress.completed()),
+            format_bytes(progress.active_total()),
+            format_bytes(progress.remaining()),
+        )
+    };
+    let heading = if failed {
+        text(heading).size(12).style(iced::widget::text::danger)
+    } else {
+        text(heading).size(12)
+    };
+    let mut heading_row = row![heading, Space::with_width(Length::Fill)]
+        .align_y(Alignment::Center)
+        .spacing(8);
+    if let ModelDownload::Active(active) = download {
+        heading_row = heading_row.push(if active.stopping {
+            danger_button("Stopping…")
+        } else {
+            danger_button("Stop").on_press(Message::StopModelDownload)
+        });
+    }
+    let mut status = column![
+        heading_row,
+        progress_bar(0.0..=1.0, progress.fraction()).height(8),
+        text(measurement).size(12),
+    ]
+    .spacing(6);
+    if speed > 0.0 && progress.remaining() > 0 {
+        status = status.push(
+            text(format!(
+                "{}/s • about {} remaining",
+                format_bytes(speed as u64),
+                format_duration(progress.remaining() as f64 / speed),
+            ))
+            .size(12),
+        );
+    }
+    status.into()
+}
+
+fn phase_text(phase: DownloadPhase) -> String {
+    match phase {
+        DownloadPhase::Pending(artifact) => format!("Ready to download {artifact}."),
+        DownloadPhase::Downloading(artifact) => format!("Downloading {artifact}…"),
+        DownloadPhase::Verifying(artifact) => format!("Verifying {artifact}…"),
+        DownloadPhase::Complete => "Download complete.".to_owned(),
+    }
+}
+
+pub(super) fn format_bytes(bytes: u64) -> String {
+    const GB: f64 = 1_000_000_000.0;
+    const MB: f64 = 1_000_000.0;
+    if bytes >= 1_000_000_000 {
+        format!("{:.1} GB", bytes as f64 / GB)
+    } else {
+        format!("{:.1} MB", bytes as f64 / MB)
+    }
+}
+
+pub(super) fn format_duration(seconds: f64) -> String {
+    let seconds = seconds.max(0.0).round() as u64;
+    let hours = seconds / 3600;
+    let minutes = seconds % 3600 / 60;
+    if hours > 0 {
+        format!("{hours}h {minutes}m")
+    } else if minutes > 0 {
+        format!("{minutes}m {}s", seconds % 60)
+    } else {
+        format!("{seconds}s")
+    }
+}

src/app/view/preferences.rs

@@ -0,0 +1,440 @@
+use super::*;
+use iced::widget::column;
+
+impl App {
+    pub(super) fn preferences_panel(&self) -> Element<'_, Message> {
+        let dspark_toggle: Option<fn(bool) -> Message> = self
+            .preference_draft
+            .model
+            .supports_dspark()
+            .then_some(Message::PreferenceDsparkChanged);
+        let dspark = checkbox(
+            "Enable DSpark for this model",
+            self.preference_draft.dspark_enabled,
+        )
+        .on_toggle_maybe(dspark_toggle);
+        let glm_mtp_toggle: Option<fn(bool) -> Message> = (self.preference_draft.model
+            == ModelChoice::Glm52)
+            .then_some(Message::PreferenceGlmMtpChanged);
+        let glm_mtp_timing_toggle: Option<fn(bool) -> Message> = (self.preference_draft.model
+            == ModelChoice::Glm52)
+            .then_some(Message::PreferenceGlmMtpTimingChanged);
+        let dspark_strict_toggle: Option<fn(bool) -> Message> = self
+            .preference_draft
+            .model
+            .supports_dspark()
+            .then_some(Message::PreferenceDsparkStrictChanged);
+        let effective = self
+            .preference_draft
+            .generation()
+            .and_then(|generation| {
+                self.preference_draft.runtime().and_then(|runtime| {
+                    crate::settings::effective_settings(
+                        self.preference_draft.model,
+                        &generation,
+                        &runtime,
+                        &models_path(),
+                    )
+                })
+            })
+            .ok();
+        let engine = effective.as_ref().map(|settings| &settings.engine);
+        let turn = effective.as_ref().map(|settings| &settings.turn);
+        let mut power = text_input("100", &self.preference_draft.power_percent);
+        let mut prefill = text_input("Automatic", &self.preference_draft.prefill_chunk);
+        let mut ssd_full_layers = text_input("Automatic", &self.preference_draft.ssd_full_layers);
+        let mut steering_file = text_input(
+            "Direction-vector file path",
+            &self.preference_draft.directional_steering_file,
+        );
+        let mut steering_ffn =
+            text_input("Automatic", &self.preference_draft.directional_steering_ffn);
+        let mut steering_attn = text_input("0", &self.preference_draft.directional_steering_attn);
+        let mut dspark_confidence = text_input(
+            "0.9 (DS4 default)",
+            &self.preference_draft.dspark_confidence_threshold,
+        );
+        if self.preference_draft.model != ModelChoice::Glm52 {
+            power = power.on_input(Message::PreferencePowerChanged);
+            prefill = prefill.on_input(Message::PreferencePrefillChunkChanged);
+            steering_file = steering_file.on_input(Message::PreferenceSteeringFileChanged);
+            steering_ffn = steering_ffn.on_input(Message::PreferenceSteeringFfnChanged);
+            steering_attn = steering_attn.on_input(Message::PreferenceSteeringAttnChanged);
+        } else {
+            ssd_full_layers = ssd_full_layers.on_input(Message::PreferenceSsdFullLayersChanged);
+        }
+        if self.preference_draft.model.supports_dspark() {
+            dspark_confidence =
+                dspark_confidence.on_input(Message::PreferenceDsparkConfidenceChanged);
+        }
+
+        let model_group = preference_group(
+            "MODEL & LIFECYCLE",
+            column![
+                pick_list(
+                    &MODEL_CHOICES[..],
+                    Some(self.preference_draft.model),
+                    Message::PreferenceModelChanged,
+                )
+                .width(Length::Fill),
+                text(format!(
+                    "Main: {}{}",
+                    engine.map_or_else(
+                        || "Invalid settings".to_owned(),
+                        |engine| engine.artifacts.model.display().to_string(),
+                    ),
+                    engine
+                        .and_then(|engine| engine.artifacts.mtp.as_ref())
+                        .map_or_else(String::new, |path| format!(
+                            " • support: {}",
+                            path.display()
+                        )),
+                ))
+                .size(12),
+                row![
+                    text_input("10", &self.preference_draft.idle_timeout_minutes)
+                        .on_input(Message::PreferenceTimeoutChanged)
+                        .width(90)
+                        .padding(9),
+                    text("minutes before unloading the model").size(13),
+                ]
+                .spacing(10)
+                .align_y(Alignment::Center),
+                text("Enter a whole number from 1 to 1440.").size(12),
+            ]
+            .spacing(10),
+        );
+        let endpoint_group = preference_group(
+            "LOCAL ENDPOINT",
+            column![
+                preference_input_row(
+                    "Port",
+                    text_input("4000", &self.preference_draft.endpoint_port)
+                        .on_input(Message::PreferenceEndpointPortChanged),
+                ),
+                text("Listens on 127.0.0.1. Saving a changed port restarts the local endpoint.")
+                    .size(12),
+            ]
+            .spacing(10),
+        );
+        let generation_group = preference_group(
+            "GENERATION",
+            column![
+                preference_input_row(
+                    "Context tokens",
+                    text_input("32768", &self.preference_draft.context_tokens)
+                        .on_input(Message::PreferenceContextChanged),
+                ),
+                preference_input_row(
+                    "Maximum generated tokens",
+                    text_input("50000", &self.preference_draft.max_generated_tokens)
+                        .on_input(Message::PreferenceMaxTokensChanged),
+                ),
+                text("System prompt").size(13),
+                text_input(
+                    "You are a helpful assistant",
+                    &self.preference_draft.system_prompt,
+                )
+                .on_input(Message::PreferenceSystemPromptChanged)
+                .padding(9),
+                Space::with_height(4),
+                text("SAMPLING & REASONING").size(11).color(muted_text()),
+                preference_input_row(
+                    "Temperature",
+                    text_input("DS4 default", &self.preference_draft.temperature)
+                        .on_input(Message::PreferenceTemperatureChanged),
+                ),
+                preference_input_row(
+                    "Top-p",
+                    text_input("DS4 default", &self.preference_draft.top_p)
+                        .on_input(Message::PreferenceTopPChanged),
+                ),
+                preference_input_row(
+                    "Min-p",
+                    text_input("DS4 default", &self.preference_draft.min_p)
+                        .on_input(Message::PreferenceMinPChanged),
+                ),
+                preference_input_row(
+                    "Seed",
+                    text_input("Random", &self.preference_draft.seed)
+                        .on_input(Message::PreferenceSeedChanged),
+                ),
+                row![
+                    text("Reasoning").size(13).width(Length::Fill),
+                    pick_list(
+                        &REASONING_MODES[..],
+                        Some(self.preference_draft.reasoning_mode),
+                        Message::PreferenceReasoningChanged,
+                    )
+                    .width(240),
+                ]
+                .spacing(12)
+                .align_y(Alignment::Center),
+                text("Blank sampling values retain DS4's model-family defaults. Think Max needs at least 393216 context tokens.")
+                    .size(12),
+                text(turn.map_or_else(
+                    || "Effective settings will appear after valid values are entered.".to_owned(),
+                    |settings| format!(
+                        "Effective: {} context • {} max • temp {} • top-p {} • min-p {} • seed {} • {} • system prompt {}",
+                        settings.context_tokens,
+                        settings.max_generated_tokens,
+                        settings.temperature,
+                        settings.top_p,
+                        settings.min_p,
+                        settings.seed.map_or_else(|| "random".to_owned(), |seed| seed.to_string()),
+                        settings.reasoning_mode,
+                        if settings.system_prompt.is_empty() { "off" } else { "on" },
+                    ),
+                ))
+                .size(12),
+            ]
+            .spacing(10),
+        );
+        let execution_group = preference_group(
+            "EXECUTION",
+            column![
+                preference_input_row(
+                    "CPU helper threads",
+                    text_input("Automatic", &self.preference_draft.cpu_threads)
+                        .on_input(Message::PreferenceCpuThreadsChanged),
+                ),
+                preference_input_row("GPU power percent", power),
+                preference_input_row("Prefill chunk", prefill),
+                checkbox("Prefer exact quality kernels", self.preference_draft.quality)
+                    .on_toggle(Message::PreferenceQualityChanged),
+                checkbox("Warm mapped weights at load time", self.preference_draft.warm_weights)
+                    .on_toggle(Message::PreferenceWarmWeightsChanged),
+                text(if self.preference_draft.model == ModelChoice::Glm52 {
+                    "GLM 5.2 uses full GPU power and selects prefill chunks automatically."
+                } else {
+                    "Blank numeric values preserve DS4's automatic engine behavior."
+                })
+                .size(12),
+                text(engine.as_ref().map_or_else(
+                    || "Effective execution settings will appear after valid values are entered."
+                        .to_owned(),
+                    |engine| {
+                        let settings = engine.execution;
+                        format!(
+                            "Metal engine: threads {} • power {}% • prefill {} • quality {} • warm weights {}",
+                            if settings.cpu_threads == 0 { "auto".to_owned() } else { settings.cpu_threads.to_string() },
+                            if settings.power_percent == 0 { 100 } else { settings.power_percent },
+                            if settings.prefill_chunk == 0 { "auto".to_owned() } else { settings.prefill_chunk.to_string() },
+                            if settings.quality { "on" } else { "off" },
+                            if settings.warm_weights { "on" } else { "off" },
+                        )
+                    },
+                ))
+                .size(12),
+            ]
+            .spacing(10),
+        );
+        let acceleration_group = preference_group(
+            "ACCELERATION & MEMORY",
+            column![
+                text("SPECULATIVE DECODING").size(11).color(muted_text()),
+                preference_input_row(
+                    "MTP draft tokens",
+                    text_input("1", &self.preference_draft.mtp_draft_tokens)
+                        .on_input(Message::PreferenceMtpDraftChanged),
+                ),
+                preference_input_row(
+                    "MTP verifier margin",
+                    text_input("3", &self.preference_draft.mtp_margin)
+                        .on_input(Message::PreferenceMtpMarginChanged),
+                ),
+                checkbox("Enable integrated GLM MTP", self.preference_draft.glm_mtp)
+                    .on_toggle_maybe(glm_mtp_toggle),
+                checkbox(
+                    "Log GLM MTP timing counters",
+                    self.preference_draft.glm_mtp_timing,
+                )
+                .on_toggle_maybe(glm_mtp_timing_toggle),
+                dspark,
+                preference_input_row("DSpark confidence threshold", dspark_confidence),
+                checkbox(
+                    "DSpark target-only decode",
+                    self.preference_draft.dspark_strict,
+                )
+                .on_toggle_maybe(dspark_strict_toggle),
+                text(if self.preference_draft.model.supports_dspark() {
+                    "DSpark uses the managed support artifact; entering a threshold or enabling strict mode also enables DSpark."
+                } else if self.preference_draft.model == ModelChoice::Glm52 {
+                    "GLM MTP is integrated; DSpark is unavailable for this model."
+                } else {
+                    "No managed MTP support artifact is available for this model."
+                })
+                .size(12),
+                text(engine.as_ref().map_or_else(
+                    || "Effective speculative settings will appear after valid values are entered."
+                        .to_owned(),
+                    |engine| {
+                        let settings = engine.speculative;
+                        format!(
+                            "Engine: MTP draft {} • margin {} • GLM MTP {} • timing {} • DSpark {} • confidence {}{} • target-only {}",
+                            settings.mtp_draft_tokens,
+                            settings.mtp_margin,
+                            if settings.glm_mtp { "on" } else { "off" },
+                            if settings.glm_mtp_timing { "on" } else { "off" },
+                            if settings.dspark { "on" } else { "off" },
+                            settings.dspark_confidence_threshold,
+                            if settings.dspark_confidence_threshold_set { " explicit" } else { " default" },
+                            if settings.dspark_strict { "on" } else { "off" },
+                        )
+                    },
+                ))
+                .size(12),
+                Space::with_height(6),
+                text("SSD STREAMING").size(11).color(muted_text()),
+                checkbox("Enable SSD-backed model streaming", self.preference_draft.ssd_streaming)
+                    .on_toggle(Message::PreferenceSsdChanged),
+                checkbox("Skip automatic expert preload", self.preference_draft.ssd_streaming_cold)
+                    .on_toggle(Message::PreferenceSsdColdChanged),
+                preference_input_row(
+                    "Expert cache count or GiB",
+                    text_input("Automatic, 128, or 64GB", &self.preference_draft.ssd_cache)
+                        .on_input(Message::PreferenceSsdCacheChanged),
+                ),
+                preference_input_row("Fully resident GLM layers", ssd_full_layers),
+                preference_input_row(
+                    "Explicit expert preload count",
+                    text_input("Automatic", &self.preference_draft.ssd_preload_experts)
+                        .on_input(Message::PreferenceSsdPreloadChanged),
+                ),
+                text("A blank full-layer value is automatic; an explicit 0 disables fully resident GLM layers. SSD streaming and DSpark are mutually exclusive.")
+                    .size(12),
+                text(engine.as_ref().map_or_else(
+                    || "Effective SSD settings will appear after valid values are entered."
+                        .to_owned(),
+                    |engine| {
+                        let settings = engine.ssd;
+                        let cache = if settings.cache_bytes > 0 {
+                            format!("{} GiB", settings.cache_bytes / GIB)
+                        } else if settings.cache_experts > 0 {
+                            format!("{} experts", settings.cache_experts)
+                        } else {
+                            "auto".to_owned()
+                        };
+                        format!(
+                            "Engine: streaming {} • cold {} • cache {} • full layers {}{} • preload {}",
+                            if settings.enabled { "on" } else { "off" },
+                            if settings.cold { "on" } else { "off" },
+                            cache,
+                            settings.full_layers,
+                            if settings.full_layers_set { " explicit" } else { " auto" },
+                            if settings.preload_experts == 0 { "auto".to_owned() } else { settings.preload_experts.to_string() },
+                        )
+                    },
+                ))
+                .size(12),
+            ]
+            .spacing(10),
+        );
+        let steering_group = preference_group(
+            "STEERING & DIAGNOSTICS",
+            column![
+                text("DIRECTIONAL STEERING").size(11).color(muted_text()),
+                text("Direction-vector file").size(13),
+                steering_file.padding(9),
+                preference_input_row("FFN scale", steering_ffn),
+                preference_input_row("Attention scale", steering_attn),
+                text(if self.preference_draft.model == ModelChoice::Glm52 {
+                    "Directional steering is not supported for GLM 5.2."
+                } else {
+                    "With a file and no explicit scale, DS4 defaults the FFN scale to 1. Scales accept -100 through 100."
+                })
+                .size(12),
+                text(engine.as_ref().map_or_else(
+                    || "Effective steering settings will appear after valid values are entered."
+                        .to_owned(),
+                    |engine| format!(
+                        "Engine: file {} • FFN scale {} • attention scale {}",
+                        if engine.steering.file.is_some() { "set" } else { "off" },
+                        engine.steering.ffn_scale,
+                        engine.steering.attention_scale,
+                    ),
+                ))
+                .size(12),
+                Space::with_height(6),
+                text("ADVANCED DIAGNOSTICS").size(11).color(muted_text()),
+                preference_input_row(
+                    "Simulated used memory (GiB)",
+                    text_input("Disabled", &self.preference_draft.simulated_used_memory_gib)
+                        .on_input(Message::PreferenceSimulatedMemoryChanged),
+                ),
+                text("Routed expert profile output").size(13),
+                text_input("Output file path", &self.preference_draft.expert_profile_path)
+                    .on_input(Message::PreferenceExpertProfileChanged)
+                    .padding(9),
+                text(engine.as_ref().map_or_else(
+                    || "Effective diagnostic settings will appear after valid values are entered."
+                        .to_owned(),
+                    |engine| format!(
+                        "{} load: simulated memory {} • expert profile {}",
+                        engine.model,
+                        if engine.diagnostics.simulated_used_memory_bytes == 0 {
+                            "off".to_owned()
+                        } else {
+                            format!("{} GiB", engine.diagnostics.simulated_used_memory_bytes / GIB)
+                        },
+                        if engine.diagnostics.expert_profile_path.is_some() { "set" } else { "off" },
+                    ),
+                ))
+                .size(12),
+            ]
+            .spacing(10),
+        );
+        let mut fields = column![
+            model_group,
+            endpoint_group,
+            generation_group,
+            execution_group,
+            acceleration_group,
+            steering_group,
+        ]
+        .spacing(12);
+
+        if let Some(error) = &self.preference_error {
+            fields = fields.push(text(error).style(iced::widget::text::danger));
+        }
+        let header = row![
+            icon(ICON_SETTINGS, 22),
+            text("Preferences").size(24),
+            Space::with_width(Length::Fill),
+            text("⌘,").size(12),
+        ]
+        .spacing(10)
+        .align_y(Alignment::Center);
+        let footer = row![
+            action_button("Reset DS4 defaults").on_press(Message::ResetPreferences),
+            Space::with_width(Length::Fill),
+            action_button("Cancel").on_press(Message::DismissPanel),
+            action_button("Save").on_press(Message::SavePreferences),
+        ]
+        .spacing(8);
+
+        let panel = container(
+            column![
+                header,
+                scrollable(container(fields).padding(iced::Padding::ZERO.right(18)))
+                    .height(Length::Fill),
+                footer
+            ]
+            .spacing(16),
+        )
+        .padding(24)
+        .width(700)
+        .height(Length::Fill)
+        .max_height(660)
+        .style(overview_style);
+        opaque(
+            container(panel)
+                .padding(24)
+                .center_x(Length::Fill)
+                .center_y(Length::Fill)
+                .style(|_| {
+                    container::Style::default().background(Color::from_rgba8(0, 0, 0, 0.68))
+                }),
+        )
+    }
+}

src/app/view/stats.rs

@@ -0,0 +1,371 @@
+use super::*;
+use iced::widget::column;
+
+impl App {
+    pub(super) fn stats_dashboard(&self) -> Element<'_, Message> {
+        let stats = &self.metrics_snapshot;
+        let context_fraction = if stats.context_limit == 0 {
+            0.0
+        } else {
+            stats.context_used.min(stats.context_limit) as f32 / stats.context_limit as f32
+        };
+        let cache_fraction = if stats.last_prompt_tokens == 0 {
+            0.0
+        } else {
+            stats.last_cached_tokens as f32 / stats.last_prompt_tokens as f32
+        };
+        let cache_hit_fraction = if stats.kv_lookups == 0 {
+            0.0
+        } else {
+            stats.kv_hits as f32 / stats.kv_lookups as f32
+        };
+        let endpoint = if stats.server_listening {
+            format!("Listening · 127.0.0.1:{}", stats.server_port)
+        } else {
+            "Stopped".to_owned()
+        };
+        let phase_color = match stats.phase {
+            crate::metrics::RuntimePhase::Generating => Color::from_rgb8(84, 170, 255),
+            crate::metrics::RuntimePhase::Prefilling | crate::metrics::RuntimePhase::Loading => {
+                Color::from_rgb8(240, 180, 70)
+            }
+            crate::metrics::RuntimePhase::Ready => Color::from_rgb8(72, 176, 112),
+            crate::metrics::RuntimePhase::Failed => Color::from_rgb8(220, 80, 86),
+            crate::metrics::RuntimePhase::Unloaded => muted_text(),
+        };
+        let heading = container(
+            row![
+                column![
+                    text("Runtime observability").size(24),
+                    text(format!(
+                        "{} · {} · uptime {}",
+                        stats.model,
+                        stats.source.label(),
+                        format_duration(stats.uptime_seconds as f64)
+                    ))
+                    .size(12)
+                    .color(muted_text()),
+                ]
+                .spacing(5),
+                Space::with_width(Length::Fill),
+                container(text(stats.phase.label()).size(12).color(phase_color))
+                    .padding([7, 11])
+                    .style(move |_| status_badge_style(phase_color)),
+            ]
+            .align_y(Alignment::Center),
+        )
+        .padding(16)
+        .style(overview_style);
+
+        let headline = column![
+            row![
+                metric_card(
+                    "DECODE",
+                    format!("{:.1} tok/s", stats.decode_tokens_per_second),
+                    format!(
+                        "{} completion tokens total",
+                        format_count(stats.completion_tokens)
+                    ),
+                ),
+                metric_card(
+                    "PREFILL",
+                    format!("{:.1} tok/s", stats.prefill_tokens_per_second),
+                    format!("{} prompt tokens total", format_count(stats.prompt_tokens)),
+                ),
+            ]
+            .spacing(10),
+            row![
+                metric_card(
+                    "CONTEXT",
+                    format!(
+                        "{} / {}",
+                        format_count(u64::from(stats.context_used)),
+                        format_count(u64::from(stats.context_limit))
+                    ),
+                    format!("{:.0}% occupied", context_fraction * 100.0),
+                ),
+                metric_card(
+                    "WORK",
+                    format!(
+                        "{} active · {} queued",
+                        stats.http_active, stats.queue_depth
+                    ),
+                    format!("{} runtime requests", format_count(stats.runtime_requests)),
+                ),
+            ]
+            .spacing(10),
+        ]
+        .spacing(10);
+
+        let throughput = stats_panel(
+            "MODEL ACTIVITY · LAST 24 SECONDS",
+            column![
+                mini_chart(
+                    &self.metrics_history,
+                    |point| point.decode_tokens_per_second,
+                    Color::from_rgb8(84, 170, 255),
+                ),
+                row![
+                    text("Decode")
+                        .size(12)
+                        .color(Color::from_rgb8(84, 170, 255)),
+                    Space::with_width(Length::Fill),
+                    text(format!("{:.1} tok/s", stats.decode_tokens_per_second))
+                        .size(12)
+                        .color(muted_text()),
+                ],
+                mini_chart(
+                    &self.metrics_history,
+                    |point| point.prefill_tokens_per_second,
+                    Color::from_rgb8(157, 119, 255),
+                ),
+                row![
+                    text("Prefill")
+                        .size(12)
+                        .color(Color::from_rgb8(157, 119, 255)),
+                    Space::with_width(Length::Fill),
+                    text(format!("{:.1} tok/s", stats.prefill_tokens_per_second))
+                        .size(12)
+                        .color(muted_text()),
+                ],
+            ]
+            .spacing(7)
+            .into(),
+        );
+        let requests = stats_panel(
+            "SERVER REQUEST RATE · LAST 24 SECONDS",
+            column![
+                mini_chart(
+                    &self.metrics_history,
+                    |point| point.http_requests_per_second,
+                    Color::from_rgb8(72, 176, 112),
+                ),
+                row![
+                    text(format!("{} requests", format_count(stats.http_requests))).size(12),
+                    Space::with_width(Length::Fill),
+                    text(format!(
+                        "{} errors · {} streaming",
+                        stats.http_errors, stats.http_streaming_requests
+                    ))
+                    .size(12)
+                    .color(muted_text()),
+                ],
+            ]
+            .spacing(7)
+            .into(),
+        );
+        let latest = self.metrics_history.back().copied().unwrap_or_default();
+        let kv_io = stats_panel(
+            "KV CHECKPOINT I/O · LAST 24 SECONDS",
+            column![
+                mini_chart(
+                    &self.metrics_history,
+                    |point| point.kv_read_bytes_per_second,
+                    Color::from_rgb8(67, 194, 203),
+                ),
+                row![
+                    text(if stats.kv_read_active {
+                        "Disk read · active"
+                    } else {
+                        "Disk read"
+                    })
+                    .size(12)
+                    .color(Color::from_rgb8(67, 194, 203)),
+                    Space::with_width(Length::Fill),
+                    text(format_rate(latest.kv_read_bytes_per_second))
+                        .size(12)
+                        .color(muted_text()),
+                ],
+                mini_chart(
+                    &self.metrics_history,
+                    |point| point.kv_write_bytes_per_second,
+                    Color::from_rgb8(240, 180, 70),
+                ),
+                row![
+                    text(if stats.kv_write_active {
+                        "Disk write · active"
+                    } else {
+                        "Disk write"
+                    })
+                    .size(12)
+                    .color(Color::from_rgb8(240, 180, 70)),
+                    Space::with_width(Length::Fill),
+                    text(format_rate(latest.kv_write_bytes_per_second))
+                        .size(12)
+                        .color(muted_text()),
+                ],
+            ]
+            .spacing(7)
+            .into(),
+        );
+
+        let model = stats_panel(
+            "MODEL CORE",
+            column![
+                metric_row("State", stats.phase.label()),
+                metric_row("Loaded model", stats.model),
+                metric_row("Mapped weights", format_bytes(stats.model_bytes)),
+                metric_row("Tensors", format_count(stats.tensor_count)),
+                metric_row("Vocabulary", format_count(stats.vocabulary_size)),
+                metric_row("Last load", format_milliseconds(stats.model_load_ms)),
+                metric_row(
+                    "Lifecycle",
+                    format!(
+                        "{} loads · {} unloads",
+                        stats.model_loads, stats.model_unloads
+                    ),
+                ),
+            ]
+            .spacing(9)
+            .into(),
+        );
+        let runtime = stats_panel(
+            "GENERATION",
+            column![
+                metric_row("Last runtime", format_milliseconds(stats.last_runtime_ms)),
+                metric_row(
+                    "Average runtime",
+                    format_milliseconds(stats.average_runtime_ms)
+                ),
+                metric_row("Last prompt", format_count(stats.last_prompt_tokens)),
+                metric_row("Last reused", format_count(stats.last_cached_tokens)),
+                metric_row(
+                    "Last completion",
+                    format_count(stats.last_completion_tokens)
+                ),
+                metric_row("Cached tokens total", format_count(stats.cached_tokens)),
+                metric_row(
+                    "Cache reuse",
+                    format!("{:.0}% of last prompt", cache_fraction * 100.0),
+                ),
+                metric_row(
+                    "Results",
+                    format!(
+                        "{} completed · {} failed",
+                        stats.completed_requests, stats.failed_requests
+                    ),
+                ),
+            ]
+            .spacing(9)
+            .into(),
+        );
+        let cache = stats_panel(
+            "KV CACHE",
+            column![
+                metric_row(
+                    "Total",
+                    format!(
+                        "{} · {} files",
+                        format_bytes(stats.kv_bytes),
+                        stats.kv_files
+                    )
+                ),
+                metric_row(
+                    "Local sessions",
+                    format!(
+                        "{} · {} files",
+                        format_bytes(stats.local_kv_bytes),
+                        stats.local_kv_files
+                    ),
+                ),
+                metric_row(
+                    "HTTP transient",
+                    format!(
+                        "{} · {} files",
+                        format_bytes(stats.http_kv_bytes),
+                        stats.http_kv_files
+                    ),
+                ),
+                metric_row("Checkpoint writes", format_count(stats.checkpoint_writes)),
+                metric_row(
+                    "Exact hits",
+                    format!(
+                        "{} · {} memory / {} disk",
+                        stats.kv_hits, stats.kv_memory_hits, stats.kv_disk_hits
+                    ),
+                ),
+                metric_row(
+                    "Misses",
+                    format!("{} · {} invalid", stats.kv_misses, stats.kv_invalid),
+                ),
+                metric_row("Lookups", format_count(stats.kv_lookups)),
+                metric_row(
+                    "Exact hit rate",
+                    format!("{:.1}%", cache_hit_fraction * 100.0)
+                ),
+                metric_row("Prefix hits", format_count(stats.kv_prefix_hits)),
+                metric_row(
+                    "Reads",
+                    format!(
+                        "{} · {} · {} errors · last {}",
+                        stats.kv_read_operations,
+                        format_bytes(stats.kv_read_bytes),
+                        stats.kv_read_errors,
+                        format_milliseconds(stats.last_kv_read_ms),
+                    ),
+                ),
+                metric_row(
+                    "Writes",
+                    format!(
+                        "{} · {} · {} errors · last {}",
+                        stats.kv_write_operations,
+                        format_bytes(stats.kv_write_bytes),
+                        stats.kv_write_errors,
+                        format_milliseconds(stats.last_kv_write_ms),
+                    ),
+                ),
+                progress_bar(0.0..=1.0, cache_fraction.min(1.0)).height(4),
+            ]
+            .spacing(9)
+            .into(),
+        );
+        let server = stats_panel(
+            "LOCAL SERVER",
+            column![
+                metric_row("Endpoint", endpoint),
+                metric_row("Active", format_count(u64::from(stats.http_active))),
+                metric_row("Completed", format_count(stats.http_completed)),
+                metric_row("Chat completions", format_count(stats.http_chat_requests)),
+                metric_row("Model queries", format_count(stats.http_model_requests)),
+                metric_row(
+                    "Runtime sources",
+                    format!(
+                        "{} local · {} HTTP",
+                        stats.local_requests, stats.endpoint_generations
+                    ),
+                ),
+                metric_row("Received", format_bytes(stats.http_bytes_received)),
+                metric_row("Last latency", format_milliseconds(stats.last_http_ms)),
+                metric_row(
+                    "Average latency",
+                    format_milliseconds(stats.average_http_ms)
+                ),
+            ]
+            .spacing(9)
+            .into(),
+        );
+
+        scrollable(
+            container(
+                column![
+                    heading,
+                    headline,
+                    throughput,
+                    kv_io,
+                    requests,
+                    row![model, runtime].spacing(10),
+                    row![cache, server].spacing(10),
+                    text("Counters are published by the runtime with relaxed atomics and sampled by the UI every 200 ms.")
+                        .size(11)
+                        .color(muted_text()),
+                ]
+                .spacing(12),
+            )
+            .padding(24)
+            .max_width(960)
+            .center_x(Length::Fill),
+        )
+        .height(Length::Fill)
+        .into()
+    }
+}

src/engine/metal/checkpoint.rs

@@ -0,0 +1,367 @@
+use super::*;
+
+impl Executor {
+    pub(in crate::engine) fn save_checkpoint(
+        &mut self,
+        path: &Path,
+        tag: [u8; 32],
+        progress: &mut impl FnMut(u64),
+    ) -> Result<(), String> {
+        if let Some(parent) = path.parent() {
+            fs::create_dir_all(parent).map_err(|error| error.to_string())?;
+        }
+        let temporary = path.with_extension("tmp");
+        let mut file = File::create(&temporary).map_err(|error| error.to_string())?;
+        let mut reported = 0_u64;
+        let mut pending = 0_u64;
+        self.write_checkpoint(&mut file, tag, &mut |bytes| {
+            reported += bytes;
+            pending += bytes;
+            if pending >= CHECKPOINT_IO_CHUNK as u64 {
+                progress(pending);
+                pending = 0;
+            }
+        })?;
+        progress(pending);
+        let total = file.metadata().map_err(|error| error.to_string())?.len();
+        progress(total.saturating_sub(reported));
+        file.sync_all().map_err(|error| error.to_string())?;
+        fs::rename(&temporary, path).map_err(|error| error.to_string())?;
+        self.checkpoint_tag = tag;
+        Ok(())
+    }
+
+    pub(in crate::engine) fn load_checkpoint(
+        &mut self,
+        path: &Path,
+        progress: &mut impl FnMut(u64),
+    ) -> Result<bool, String> {
+        let mut file = match File::open(path) {
+            Ok(file) => file,
+            Err(error) if error.kind() == std::io::ErrorKind::NotFound => return Ok(false),
+            Err(error) => return Err(error.to_string()),
+        };
+        let mut reported = 0_u64;
+        let mut pending = 0_u64;
+        self.read_checkpoint(&mut file, &mut |bytes| {
+            reported += bytes;
+            pending += bytes;
+            if pending >= CHECKPOINT_IO_CHUNK as u64 {
+                progress(pending);
+                pending = 0;
+            }
+        })?;
+        progress(pending);
+        let total = file.metadata().map_err(|error| error.to_string())?.len();
+        progress(total.saturating_sub(reported));
+        Ok(true)
+    }
+
+    fn write_checkpoint(
+        &self,
+        file: &mut File,
+        tag: [u8; 32],
+        progress: &mut impl FnMut(u64),
+    ) -> Result<(), String> {
+        let shape = self.model.shape;
+        file.write_all(CHECKPOINT_MAGIC)
+            .map_err(|error| error.to_string())?;
+        for value in [
+            CHECKPOINT_VERSION,
+            self.session.context,
+            self.session.raw_cap,
+            shape.layers,
+            shape.head_dim as u32,
+            shape.indexer_head_dim as u32,
+            shape.vocab as u32,
+            u32::from(self.quality),
+        ] {
+            write_u32(file, value)?;
+        }
+        write_u64(file, self.model.main.len())?;
+        write_u64(file, self.model_modified.0)?;
+        write_u32(file, self.model_modified.1)?;
+        for weight in [self.weights.token_embedding, self.weights.output] {
+            write_u64(file, weight.offset)?;
+            write_u64(file, weight.bytes)?;
+            write_u32(file, weight.kind)?;
+        }
+        file.write_all(&tag).map_err(|error| error.to_string())?;
+        let token_count = u32::try_from(self.tokens.len())
+            .map_err(|_| "KV checkpoint has too many tokens".to_owned())?;
+        let raw_live = token_count.min(self.session.raw_cap);
+        write_u32(file, token_count)?;
+        write_u32(file, raw_live)?;
+        for &token in &self.tokens {
+            write_u32(file, token as u32)?;
+        }
+        for &logit in &self.logits {
+            write_u32(file, logit.to_bits())?;
+        }
+        for layer in &self.session.layers {
+            write_u32(
+                file,
+                layer.compression.as_ref().map_or(0, |state| state.rows),
+            )?;
+        }
+        for layer in &self.session.layers {
+            write_u32(file, layer.indexer.as_ref().map_or(0, |state| state.rows))?;
+        }
+
+        let mut chunk = vec![0; CHECKPOINT_IO_CHUNK];
+        let raw_first = token_count - raw_live;
+        for layer in &self.session.layers {
+            for position in raw_first..token_count {
+                let physical = position % self.session.raw_cap;
+                write_buffer(
+                    file,
+                    &layer.raw_cache,
+                    u64::from(physical) * shape.head_dim * 4,
+                    shape.head_dim * 4,
+                    &mut chunk,
+                    progress,
+                )?;
+            }
+            if let Some(state) = &layer.compression {
+                write_buffer(
+                    file,
+                    &state.cache,
+                    0,
+                    u64::from(state.rows) * shape.head_dim * 2,
+                    &mut chunk,
+                    progress,
+                )?;
+                let bytes = compressor_state_bytes(state.ratio, shape.head_dim);
+                write_buffer(file, &state.state_kv, 0, bytes, &mut chunk, progress)?;
+                write_buffer(file, &state.state_score, 0, bytes, &mut chunk, progress)?;
+            }
+            if let Some(state) = &layer.indexer {
+                write_buffer(
+                    file,
+                    &state.cache,
+                    0,
+                    u64::from(state.rows) * shape.indexer_head_dim * 4,
+                    &mut chunk,
+                    progress,
+                )?;
+                let bytes = compressor_state_bytes(state.ratio, shape.indexer_head_dim);
+                write_buffer(file, &state.state_kv, 0, bytes, &mut chunk, progress)?;
+                write_buffer(file, &state.state_score, 0, bytes, &mut chunk, progress)?;
+            }
+        }
+        Ok(())
+    }
+
+    fn read_checkpoint(
+        &mut self,
+        file: &mut File,
+        progress: &mut impl FnMut(u64),
+    ) -> Result<(), String> {
+        let mut magic = [0; 8];
+        file.read_exact(&mut magic)
+            .map_err(|error| error.to_string())?;
+        if &magic != CHECKPOINT_MAGIC {
+            return Err("KV checkpoint has an invalid signature".into());
+        }
+        let shape = self.model.shape;
+        let header = [
+            CHECKPOINT_VERSION,
+            self.session.context,
+            self.session.raw_cap,
+            shape.layers,
+            shape.head_dim as u32,
+            shape.indexer_head_dim as u32,
+            shape.vocab as u32,
+            u32::from(self.quality),
+        ];
+        for expected in header {
+            if read_u32(file)? != expected {
+                return Err("KV checkpoint does not match the current executor".into());
+            }
+        }
+        if read_u64(file)? != self.model.main.len() {
+            return Err("KV checkpoint was written for a different model".into());
+        }
+        if read_u64(file)? != self.model_modified.0 || read_u32(file)? != self.model_modified.1 {
+            return Err("KV checkpoint model file has changed".into());
+        }
+        for weight in [self.weights.token_embedding, self.weights.output] {
+            if read_u64(file)? != weight.offset
+                || read_u64(file)? != weight.bytes
+                || read_u32(file)? != weight.kind
+            {
+                return Err("KV checkpoint was written for a different model layout".into());
+            }
+        }
+        let mut checkpoint_tag = [0; 32];
+        file.read_exact(&mut checkpoint_tag)
+            .map_err(|error| error.to_string())?;
+        let token_count = read_u32(file)?;
+        let raw_live = read_u32(file)?;
+        if token_count > self.session.context || raw_live != token_count.min(self.session.raw_cap) {
+            return Err("KV checkpoint token count is invalid".into());
+        }
+        let mut tokens = Vec::with_capacity(token_count as usize);
+        for _ in 0..token_count {
+            let token = read_u32(file)?;
+            if u64::from(token) >= shape.vocab {
+                return Err("KV checkpoint contains an invalid token".into());
+            }
+            tokens.push(token as i32);
+        }
+        let mut logits = Vec::with_capacity(shape.vocab as usize);
+        for _ in 0..shape.vocab {
+            logits.push(f32::from_bits(read_u32(file)?));
+        }
+        let mut compressed_rows = Vec::with_capacity(shape.layers as usize);
+        let mut indexer_rows = Vec::with_capacity(shape.layers as usize);
+        for layer in 0..shape.layers {
+            let rows = read_u32(file)?;
+            let ratio = compression_ratio(layer);
+            if rows != token_count.checked_div(ratio).unwrap_or(0) {
+                return Err("KV checkpoint compressed row count is invalid".into());
+            }
+            compressed_rows.push(rows);
+        }
+        for layer in 0..shape.layers {
+            let rows = read_u32(file)?;
+            let expected = if compression_ratio(layer) == 4 {
+                token_count / 4
+            } else {
+                0
+            };
+            if rows != expected {
+                return Err("KV checkpoint indexer row count is invalid".into());
+            }
+            indexer_rows.push(rows);
+        }
+
+        self.reset()?;
+        let mut chunk = vec![0; CHECKPOINT_IO_CHUNK];
+        let raw_first = token_count - raw_live;
+        for (index, layer) in self.session.layers.iter_mut().enumerate() {
+            for position in raw_first..token_count {
+                let physical = position % self.session.raw_cap;
+                read_buffer(
+                    file,
+                    &layer.raw_cache,
+                    u64::from(physical) * shape.head_dim * 4,
+                    shape.head_dim * 4,
+                    &mut chunk,
+                    progress,
+                )?;
+            }
+            if let Some(state) = &mut layer.compression {
+                state.rows = compressed_rows[index];
+                read_buffer(
+                    file,
+                    &state.cache,
+                    0,
+                    u64::from(state.rows) * shape.head_dim * 2,
+                    &mut chunk,
+                    progress,
+                )?;
+                let bytes = compressor_state_bytes(state.ratio, shape.head_dim);
+                read_buffer(file, &state.state_kv, 0, bytes, &mut chunk, progress)?;
+                read_buffer(file, &state.state_score, 0, bytes, &mut chunk, progress)?;
+            }
+            if let Some(state) = &mut layer.indexer {
+                state.rows = indexer_rows[index];
+                read_buffer(
+                    file,
+                    &state.cache,
+                    0,
+                    u64::from(state.rows) * shape.indexer_head_dim * 4,
+                    &mut chunk,
+                    progress,
+                )?;
+                let bytes = compressor_state_bytes(state.ratio, shape.indexer_head_dim);
+                read_buffer(file, &state.state_kv, 0, bytes, &mut chunk, progress)?;
+                read_buffer(file, &state.state_score, 0, bytes, &mut chunk, progress)?;
+            }
+        }
+        let mut trailing = [0];
+        if file
+            .read(&mut trailing)
+            .map_err(|error| error.to_string())?
+            != 0
+        {
+            self.reset()?;
+            return Err("KV checkpoint has trailing data".into());
+        }
+        self.session.position = token_count;
+        self.tokens = tokens;
+        self.logits = logits;
+        self.checkpoint_tag = checkpoint_tag;
+        Ok(())
+    }
+}
+
+fn compressor_state_bytes(ratio: u32, head_dim: u64) -> u64 {
+    let coefficient = if ratio == 4 { 2 } else { 1 };
+    coefficient * head_dim * coefficient * u64::from(ratio) * 4
+}
+
+fn write_buffer(
+    file: &mut File,
+    buffer: &Buffer,
+    mut offset: u64,
+    mut bytes: u64,
+    chunk: &mut [u8],
+    progress: &mut impl FnMut(u64),
+) -> Result<(), String> {
+    while bytes != 0 {
+        let length = bytes.min(chunk.len() as u64) as usize;
+        buffer.read(offset, &mut chunk[..length])?;
+        file.write_all(&chunk[..length])
+            .map_err(|error| error.to_string())?;
+        progress(length as u64);
+        offset += length as u64;
+        bytes -= length as u64;
+    }
+    Ok(())
+}
+
+fn read_buffer(
+    file: &mut File,
+    buffer: &Buffer,
+    mut offset: u64,
+    mut bytes: u64,
+    chunk: &mut [u8],
+    progress: &mut impl FnMut(u64),
+) -> Result<(), String> {
+    while bytes != 0 {
+        let length = bytes.min(chunk.len() as u64) as usize;
+        file.read_exact(&mut chunk[..length])
+            .map_err(|error| error.to_string())?;
+        buffer.write(offset, &chunk[..length])?;
+        progress(length as u64);
+        offset += length as u64;
+        bytes -= length as u64;
+    }
+    Ok(())
+}
+
+fn write_u32(file: &mut File, value: u32) -> Result<(), String> {
+    file.write_all(&value.to_le_bytes())
+        .map_err(|error| error.to_string())
+}
+
+fn write_u64(file: &mut File, value: u64) -> Result<(), String> {
+    file.write_all(&value.to_le_bytes())
+        .map_err(|error| error.to_string())
+}
+
+fn read_u32(file: &mut File) -> Result<u32, String> {
+    let mut bytes = [0; 4];
+    file.read_exact(&mut bytes)
+        .map_err(|error| error.to_string())?;
+    Ok(u32::from_le_bytes(bytes))
+}
+
+fn read_u64(file: &mut File) -> Result<u64, String> {
+    let mut bytes = [0; 8];
+    file.read_exact(&mut bytes)
+        .map_err(|error| error.to_string())?;
+    Ok(u64::from_le_bytes(bytes))
+}

src/engine/metal/gpu.rs

@@ -0,0 +1,926 @@
+use super::*;
+
+#[repr(C)]
+pub(super) struct GpuTensor {
+    _private: [u8; 0],
+}
+
+unsafe extern "C" {
+    pub(super) fn ds4_gpu_init() -> i32;
+    pub(super) fn ds4_gpu_cleanup();
+    pub(super) fn ds4_gpu_set_model_map_range(
+        model_map: *const c_void,
+        model_size: u64,
+        map_offset: u64,
+        map_size: u64,
+        max_tensor_bytes: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_set_quality(quality: bool);
+    pub(super) fn ds4_gpu_tensor_alloc(bytes: u64) -> *mut GpuTensor;
+    pub(super) fn ds4_gpu_tensor_view(
+        base: *const GpuTensor,
+        offset: u64,
+        bytes: u64,
+    ) -> *mut GpuTensor;
+    pub(super) fn ds4_gpu_tensor_free(tensor: *mut GpuTensor);
+    pub(super) fn ds4_gpu_tensor_fill_f32(tensor: *mut GpuTensor, value: f32, count: u64) -> i32;
+    pub(super) fn ds4_gpu_tensor_read(
+        tensor: *const GpuTensor,
+        offset: u64,
+        data: *mut c_void,
+        bytes: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_tensor_write(
+        tensor: *mut GpuTensor,
+        offset: u64,
+        data: *const c_void,
+        bytes: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_tensor_copy(
+        dst: *mut GpuTensor,
+        dst_offset: u64,
+        src: *const GpuTensor,
+        src_offset: u64,
+        bytes: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_tensor_copy_f32_to_f16(
+        dst: *mut GpuTensor,
+        dst_offset: u64,
+        src: *const GpuTensor,
+        src_offset: u64,
+        count: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_begin_commands() -> i32;
+    pub(super) fn ds4_gpu_end_commands() -> i32;
+
+    pub(super) fn ds4_gpu_embed_tokens_hc_tensor(
+        out: *mut GpuTensor,
+        tokens: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        vocab: u32,
+        rows: u32,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+
+    pub(super) fn ds4_gpu_embed_token_hc_tensor(
+        out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        vocab: u32,
+        token: u32,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_rms_norm_plain_tensor(
+        out: *mut GpuTensor,
+        x: *const GpuTensor,
+        n: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_rms_norm_weight_tensor(
+        out: *mut GpuTensor,
+        x: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        n: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_rms_scale_project_f16_tensor(
+        out: *mut GpuTensor,
+        scale: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u32,
+        output: u32,
+        x: *const GpuTensor,
+        rows: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_f16_tensor(
+        out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        rows: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_q8_0_tensor(
+        out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        rows: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_q8_0_pair_tensor(
+        out_a: *mut GpuTensor,
+        out_b: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight_a: u64,
+        weight_b: u64,
+        input: u64,
+        output_a: u64,
+        output_b: u64,
+        x: *const GpuTensor,
+        rows: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_f16_pair_tensor(
+        out_a: *mut GpuTensor,
+        out_b: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight_a: u64,
+        weight_b: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        rows: u64,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_f16_pair_compressor_store_tensor(
+        out_kv: *mut GpuTensor,
+        out_score: *mut GpuTensor,
+        state_kv: *mut GpuTensor,
+        state_score: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight_kv: u64,
+        weight_score: u64,
+        ape: u64,
+        ape_type: u32,
+        input: u64,
+        width: u32,
+        x: *const GpuTensor,
+        ratio: u32,
+        pos: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_split_weighted_sum_norm_tensor(
+        out: *mut GpuTensor,
+        norm: *mut GpuTensor,
+        split: *mut GpuTensor,
+        mix: *const GpuTensor,
+        residual: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        scale: u64,
+        base: u64,
+        norm_weight: u64,
+        embd: u32,
+        hc: u32,
+        iterations: u32,
+        eps: f32,
+        norm_eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_dsv4_qkv_rms_norm_rows_tensor(
+        q_out: *mut GpuTensor,
+        q: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        q_weight: u64,
+        q_n: u32,
+        kv_out: *mut GpuTensor,
+        kv: *const GpuTensor,
+        kv_weight: u64,
+        kv_n: u32,
+        rows: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attn_q_b_f16_head_rms_rope_tail_tensor(
+        out: *mut GpuTensor,
+        half: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        rows: u32,
+        heads: u32,
+        head_dim: u32,
+        rot: u32,
+        pos: u32,
+        original_context: u32,
+        inverse: bool,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_head_rms_norm_tensor(
+        x: *mut GpuTensor,
+        rows: u32,
+        heads: u32,
+        head_dim: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_head_rms_norm_rope_tail_tensor(
+        x: *mut GpuTensor,
+        rows: u32,
+        heads: u32,
+        head_dim: u32,
+        rot: u32,
+        pos: u32,
+        original_context: u32,
+        inverse: bool,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_rope_tail_tensor(
+        x: *mut GpuTensor,
+        rows: u32,
+        heads: u32,
+        head_dim: u32,
+        rot: u32,
+        pos: u32,
+        original_context: u32,
+        inverse: bool,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_decode_heads_tensor(
+        heads_out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw_kv: *const GpuTensor,
+        n_raw: u32,
+        raw_cap: u32,
+        raw_start: u32,
+        comp_kv: *const GpuTensor,
+        comp_f16: u32,
+        n_comp: u32,
+        comp_mask: *const GpuTensor,
+        use_mask: u32,
+        heads: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_decode_raw_batch_heads_tensor(
+        heads: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw_kv: *const GpuTensor,
+        tokens: u32,
+        pos: u32,
+        n_raw: u32,
+        raw_cap: u32,
+        raw_start: u32,
+        window: u32,
+        heads_count: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_decode_mixed_batch_heads_tensor(
+        heads: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw_kv: *const GpuTensor,
+        compressed: *const GpuTensor,
+        compressed_f16: u32,
+        compressed_mask: *const GpuTensor,
+        use_mask: u32,
+        tokens: u32,
+        pos: u32,
+        n_raw: u32,
+        raw_cap: u32,
+        raw_start: u32,
+        n_comp: u32,
+        window: u32,
+        ratio: u32,
+        heads_count: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_indexed_mixed_batch_heads_tensor(
+        heads: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw_kv: *const GpuTensor,
+        comp_kv: *const GpuTensor,
+        comp_f16: u32,
+        topk: *const GpuTensor,
+        tokens: u32,
+        pos: u32,
+        n_raw: u32,
+        raw_cap: u32,
+        raw_start: u32,
+        n_comp: u32,
+        top_k: u32,
+        window: u32,
+        ratio: u32,
+        heads: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_indexer_score_one_tensor(
+        scores: *mut GpuTensor,
+        q: *const GpuTensor,
+        weights: *const GpuTensor,
+        index_comp: *const GpuTensor,
+        n_comp: u32,
+        heads: u32,
+        head_dim: u32,
+        scale: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_indexer_scores_decode_batch_tensor(
+        scores: *mut GpuTensor,
+        q: *const GpuTensor,
+        weights: *const GpuTensor,
+        index_comp: *const GpuTensor,
+        n_comp: u32,
+        tokens: u32,
+        pos: u32,
+        heads: u32,
+        head_dim: u32,
+        ratio: u32,
+        scale: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_indexer_topk_tensor(
+        selected: *mut GpuTensor,
+        scores: *const GpuTensor,
+        n_comp: u32,
+        tokens: u32,
+        top_k: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_dsv4_indexer_qat_tensor(
+        x: *mut GpuTensor,
+        rows: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_compressor_update_tensor(
+        kv: *const GpuTensor,
+        score: *const GpuTensor,
+        state_kv: *mut GpuTensor,
+        state_score: *mut GpuTensor,
+        cache: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        ape: u64,
+        ape_type: u32,
+        norm: u64,
+        norm_type: u32,
+        head_dim: u32,
+        ratio: u32,
+        pos: u32,
+        row: u32,
+        rot: u32,
+        original_context: u32,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+        rms_eps: f32,
+        state_already_stored: bool,
+    ) -> i32;
+    pub(super) fn ds4_gpu_compressor_prefill_state_ratio4_tensor(
+        state_kv: *mut GpuTensor,
+        state_score: *mut GpuTensor,
+        kv_tail: *const GpuTensor,
+        score_tail: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        ape: u64,
+        ape_type: u32,
+        head_dim: u32,
+        pos: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_dsv4_fp8_kv_quantize_tensor(
+        x: *mut GpuTensor,
+        rows: u32,
+        head_dim: u32,
+        rot: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_kv_fp8_store_raw_tensor(
+        kv: *mut GpuTensor,
+        raw_cache: *mut GpuTensor,
+        raw_cap: u32,
+        row: u32,
+        head_dim: u32,
+        rot: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_store_raw_kv_batch_tensor(
+        raw_cache: *mut GpuTensor,
+        kv: *const GpuTensor,
+        raw_cap: u32,
+        pos: u32,
+        rows: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_compressor_prefill_tensor(
+        cache: *mut GpuTensor,
+        state_kv: *mut GpuTensor,
+        state_score: *mut GpuTensor,
+        kv: *const GpuTensor,
+        score: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        ape: u64,
+        ape_type: u32,
+        norm: u64,
+        norm_type: u32,
+        head_dim: u32,
+        ratio: u32,
+        pos: u32,
+        rows: u32,
+        rot: u32,
+        original_context: u32,
+        quantize_fp8: bool,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+        rms_eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_compressor_prefill_ratio4_replay_tensor(
+        cache: *mut GpuTensor,
+        state_kv: *mut GpuTensor,
+        state_score: *mut GpuTensor,
+        kv: *const GpuTensor,
+        score: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        ape: u64,
+        ape_type: u32,
+        norm: u64,
+        norm_type: u32,
+        head_dim: u32,
+        pos: u32,
+        rows: u32,
+        rot: u32,
+        original_context: u32,
+        quantize_fp8: bool,
+        freq_base: f32,
+        freq_scale: f32,
+        ext_factor: f32,
+        attn_factor: f32,
+        beta_fast: f32,
+        beta_slow: f32,
+        rms_eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_prefill_raw_heads_tensor(
+        heads: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw: *const GpuTensor,
+        rows: u32,
+        window: u32,
+        heads_count: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_prefill_static_mixed_heads_tensor(
+        heads: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        sinks: u64,
+        q: *const GpuTensor,
+        raw: *const GpuTensor,
+        compressed: *const GpuTensor,
+        compressed_f16: u32,
+        rows: u32,
+        compressed_rows: u32,
+        window: u32,
+        ratio: u32,
+        heads_count: u32,
+        head_dim: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_indexer_scores_prefill_tensor(
+        scores: *mut GpuTensor,
+        q: *const GpuTensor,
+        weights: *const GpuTensor,
+        compressed: *const GpuTensor,
+        compressed_rows: u32,
+        rows: u32,
+        heads: u32,
+        head_dim: u32,
+        ratio: u32,
+        scale: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_output_q8_batch_f16_tensor(
+        out_half: *mut GpuTensor,
+        low: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight_a: u64,
+        weight_b: u64,
+        group_dim: u64,
+        rank: u64,
+        groups: u32,
+        output: u64,
+        heads: *const GpuTensor,
+        rows: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_output_q8_batch_tensor(
+        out: *mut GpuTensor,
+        low: *mut GpuTensor,
+        group_scratch: *mut GpuTensor,
+        low_scratch: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight_a: u64,
+        weight_b: u64,
+        group_dim: u64,
+        rank: u64,
+        groups: u32,
+        output: u64,
+        heads: *const GpuTensor,
+        rows: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_router_select_batch_tensor(
+        selected: *mut GpuTensor,
+        weights: *mut GpuTensor,
+        probs: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        bias: u64,
+        hash: u64,
+        hash_rows: u32,
+        expert_groups: u32,
+        groups_used: u32,
+        has_bias: bool,
+        hash_mode: bool,
+        logits: *const GpuTensor,
+        tokens: *const GpuTensor,
+        experts: u32,
+        experts_used: u32,
+        scale: f32,
+        rows: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_routed_moe_batch_tensor(
+        out: *mut GpuTensor,
+        gate: *mut GpuTensor,
+        up: *mut GpuTensor,
+        mid: *mut GpuTensor,
+        experts_out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        gate_weight: u64,
+        up_weight: u64,
+        down_weight: u64,
+        gate_type: u32,
+        down_type: u32,
+        gate_expert_bytes: u64,
+        gate_row_bytes: u64,
+        down_expert_bytes: u64,
+        down_row_bytes: u64,
+        input: u32,
+        middle: u32,
+        output: u32,
+        selected: *const GpuTensor,
+        weights: *const GpuTensor,
+        total_experts: u32,
+        used_experts: u32,
+        clamp: f32,
+        x: *const GpuTensor,
+        layer: u32,
+        rows: u32,
+        mid_f16: *mut bool,
+        force_resident: bool,
+    ) -> i32;
+    pub(super) fn ds4_gpu_swiglu_tensor(
+        out: *mut GpuTensor,
+        gate: *const GpuTensor,
+        up: *const GpuTensor,
+        count: u32,
+        clamp: f32,
+        scale: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_expand_split_half_tensor(
+        out: *mut GpuTensor,
+        block_half: *const GpuTensor,
+        residual: *const GpuTensor,
+        split: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_expand_split_tensor(
+        out: *mut GpuTensor,
+        block: *const GpuTensor,
+        residual: *const GpuTensor,
+        split: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_expand_add_split_tensor(
+        out: *mut GpuTensor,
+        block: *const GpuTensor,
+        add: *const GpuTensor,
+        residual: *const GpuTensor,
+        split: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_attention_output_low_q8_tensor(
+        low: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        group_dim: u64,
+        rank: u64,
+        groups: u32,
+        heads: *const GpuTensor,
+    ) -> i32;
+    pub(super) fn ds4_gpu_matmul_q8_0_hc_expand_tensor(
+        out_hc: *mut GpuTensor,
+        block_out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        residual: *const GpuTensor,
+        split: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_router_select_tensor(
+        selected: *mut GpuTensor,
+        weights: *mut GpuTensor,
+        probs: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        bias: u64,
+        hash: u64,
+        hash_rows: u32,
+        token: u32,
+        experts: u32,
+        used: u32,
+        scale: f32,
+        expert_groups: u32,
+        groups_used: u32,
+        has_bias: bool,
+        hash_mode: bool,
+        logits: *const GpuTensor,
+    ) -> i32;
+    pub(super) fn ds4_gpu_routed_moe_one_tensor(
+        out: *mut GpuTensor,
+        gate: *mut GpuTensor,
+        up: *mut GpuTensor,
+        mid: *mut GpuTensor,
+        experts_out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        gate_weight: u64,
+        up_weight: u64,
+        down_weight: u64,
+        gate_type: u32,
+        down_type: u32,
+        gate_expert_bytes: u64,
+        gate_row_bytes: u64,
+        down_expert_bytes: u64,
+        down_row_bytes: u64,
+        input: u32,
+        middle: u32,
+        output: u32,
+        selected: *const GpuTensor,
+        weights: *const GpuTensor,
+        total_experts: u32,
+        used_experts: u32,
+        clamp: f32,
+        x: *const GpuTensor,
+        add: *const GpuTensor,
+        layer: u32,
+        force_resident: bool,
+    ) -> i32;
+    pub(super) fn ds4_gpu_shared_gate_up_swiglu_q8_0_tensor(
+        gate: *mut GpuTensor,
+        up: *mut GpuTensor,
+        mid: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        gate_weight: u64,
+        up_weight: u64,
+        input: u64,
+        output: u64,
+        x: *const GpuTensor,
+        clamp: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_shared_down_hc_expand_q8_0_tensor(
+        out_hc: *mut GpuTensor,
+        shared_out: *mut GpuTensor,
+        map: *const c_void,
+        size: u64,
+        weight: u64,
+        input: u64,
+        output: u64,
+        middle: *const GpuTensor,
+        routed: *const GpuTensor,
+        residual: *const GpuTensor,
+        split: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_output_hc_weights_tensor(
+        out: *mut GpuTensor,
+        pre: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        scale: u64,
+        base: u64,
+        hc: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_weighted_sum_norm_tensor(
+        out: *mut GpuTensor,
+        norm: *mut GpuTensor,
+        residual: *const GpuTensor,
+        weights: *const GpuTensor,
+        map: *const c_void,
+        size: u64,
+        norm_weight: u64,
+        embd: u32,
+        hc: u32,
+        eps: f32,
+    ) -> i32;
+    pub(super) fn ds4_gpu_hc_weighted_sum_tensor(
+        out: *mut GpuTensor,
+        residual: *const GpuTensor,
+        weights: *const GpuTensor,
+        embd: u32,
+        hc: u32,
+    ) -> i32;
+}
+
+pub(super) struct Context;
+
+impl Context {
+    pub(super) fn open(model: &Model, quality: bool) -> Result<Self, String> {
+        check(unsafe { ds4_gpu_init() }, "Metal initialization")?;
+        let data_offset = model.main.data_offset();
+        if let Err(error) = check(
+            unsafe {
+                ds4_gpu_set_model_map_range(
+                    model.main.map_ptr().cast(),
+                    model.main.len(),
+                    data_offset,
+                    model.main.len() - data_offset,
+                    model.main.max_tensor_bytes(),
+                )
+            },
+            "model mapping",
+        ) {
+            unsafe { ds4_gpu_cleanup() };
+            return Err(error);
+        }
+        unsafe { ds4_gpu_set_quality(quality) };
+        Ok(Self)
+    }
+}
+
+impl Drop for Context {
+    fn drop(&mut self) {
+        unsafe { ds4_gpu_cleanup() };
+    }
+}
+
+pub(super) struct Commands(bool);
+
+impl Commands {
+    pub(super) fn begin() -> Result<Self, String> {
+        check(
+            unsafe { ds4_gpu_begin_commands() },
+            "beginning Metal commands",
+        )?;
+        Ok(Self(true))
+    }
+
+    pub(super) fn finish(mut self) -> Result<(), String> {
+        self.0 = false;
+        check(
+            unsafe { ds4_gpu_end_commands() },
+            "executing Metal commands",
+        )
+    }
+}
+
+impl Drop for Commands {
+    fn drop(&mut self) {
+        if self.0 {
+            unsafe { ds4_gpu_end_commands() };
+        }
+    }
+}
+
+pub(super) struct Buffer(NonNull<GpuTensor>);
+
+impl Buffer {
+    pub(super) fn floats(count: u64) -> Result<Self, String> {
+        Self::bytes(count * 4)
+    }
+
+    pub(super) fn bytes(bytes: u64) -> Result<Self, String> {
+        NonNull::new(unsafe { ds4_gpu_tensor_alloc(bytes) })
+            .map(Self)
+            .ok_or_else(|| format!("Metal could not allocate {bytes} bytes"))
+    }
+
+    pub(super) fn view(&self, offset: u64, bytes: u64) -> Result<Self, String> {
+        NonNull::new(unsafe { ds4_gpu_tensor_view(self.raw(), offset, bytes) })
+            .map(Self)
+            .ok_or_else(|| "Metal could not create a tensor view".to_owned())
+    }
+
+    pub(super) fn read_f32(&self, values: &mut [f32]) -> Result<(), String> {
+        check(
+            unsafe {
+                ds4_gpu_tensor_read(
+                    self.raw(),
+                    0,
+                    values.as_mut_ptr().cast(),
+                    std::mem::size_of_val(values) as u64,
+                )
+            },
+            "reading Metal output",
+        )
+    }
+
+    pub(super) fn read(&self, offset: u64, values: &mut [u8]) -> Result<(), String> {
+        check(
+            unsafe {
+                ds4_gpu_tensor_read(
+                    self.raw(),
+                    offset,
+                    values.as_mut_ptr().cast(),
+                    values.len() as u64,
+                )
+            },
+            "reading a Metal buffer",
+        )
+    }
+
+    pub(super) fn write(&self, offset: u64, values: &[u8]) -> Result<(), String> {
+        check(
+            unsafe {
+                ds4_gpu_tensor_write(
+                    self.raw(),
+                    offset,
+                    values.as_ptr().cast(),
+                    values.len() as u64,
+                )
+            },
+            "restoring a Metal buffer",
+        )
+    }
+
+    pub(super) fn write_i32(&self, values: &[i32]) -> Result<(), String> {
+        check(
+            unsafe {
+                ds4_gpu_tensor_write(
+                    self.raw(),
+                    0,
+                    values.as_ptr().cast(),
+                    std::mem::size_of_val(values) as u64,
+                )
+            },
+            "uploading tokens",
+        )
+    }
+
+    pub(super) fn fill(&self, value: f32, count: u64) -> Result<(), String> {
+        call(
+            unsafe { ds4_gpu_tensor_fill_f32(self.raw(), value, count) },
+            "initializing a Metal buffer",
+        )
+    }
+
+    pub(super) fn raw(&self) -> *mut GpuTensor {
+        self.0.as_ptr()
+    }
+}
+
+impl Drop for Buffer {
+    fn drop(&mut self) {
+        unsafe { ds4_gpu_tensor_free(self.raw()) };
+    }
+}

src/engine/validation.rs

@@ -0,0 +1,994 @@
+use super::*;
+
+pub(crate) fn validate_model_artifact(
+    path: &Path,
+    expected: ModelChoice,
+    support: bool,
+) -> Result<(), String> {
+    if support {
+        let model = Gguf::open(path)?;
+        validate_dspark(&model, &FLASH)
+    } else {
+        let model = Model::open_main(path, expected)?;
+        let summary = model.summary();
+        if summary.tensor_count == 0
+            || model
+                .render_prompt("", "", ReasoningMode::Direct)
+                .is_empty()
+        {
+            return Err("model intake produced an empty tensor directory or prompt".into());
+        }
+        let _ = model.tokenize("");
+        let eos = model.eos_token();
+        let _ = model.token_bytes(eos);
+        if !model.is_stop_token(eos) {
+            return Err("tokenizer EOS marker is not a stop token".into());
+        }
+        model.tensor_data("token_embd.weight")?;
+        Ok(())
+    }
+}
+
+pub(super) fn validate_main(model: &Gguf, expected: ModelChoice) -> Result<Shape, String> {
+    let family = if model.bytes("general.architecture").ok() == Some(b"glm-dsa") {
+        ModelFamily::Glm
+    } else {
+        ModelFamily::DeepSeek
+    };
+    let shape = match family {
+        ModelFamily::Glm => GLM,
+        ModelFamily::DeepSeek => match model.u32("deepseek4.block_count")? {
+            43 => FLASH,
+            61 => PRO,
+            layers => return Err(format!("unsupported DeepSeek layer count: {layers}")),
+        },
+    };
+    if shape.model != expected {
+        return Err(format!(
+            "{} contains {}, but the selected model is {expected}",
+            model.path().display(),
+            shape.model
+        ));
+    }
+    validate_metadata(model, &shape)?;
+    validate_tensors(model, &shape)?;
+    Ok(shape)
+}
+
+fn validate_metadata(model: &Gguf, shape: &Shape) -> Result<(), String> {
+    let prefix = if shape.family == ModelFamily::Glm {
+        "glm-dsa"
+    } else {
+        "deepseek4"
+    };
+    for (key, expected) in [
+        ("block_count", u64::from(shape.layers)),
+        ("embedding_length", shape.embd),
+        ("vocab_size", shape.vocab),
+        ("attention.head_count", shape.heads),
+        ("attention.head_count_kv", shape.head_kv),
+        ("attention.key_length", shape.head_dim),
+        ("attention.value_length", shape.value_dim),
+        ("rope.dimension_count", shape.rot),
+        ("attention.q_lora_rank", shape.lora_q),
+        ("expert_count", shape.experts),
+        ("expert_used_count", shape.experts_used),
+        ("expert_feed_forward_length", shape.ff_expert),
+        ("expert_shared_count", shape.expert_shared),
+        ("attention.indexer.head_count", shape.indexer_heads),
+        ("attention.indexer.key_length", shape.indexer_head_dim),
+        ("attention.indexer.top_k", shape.indexer_top_k),
+    ] {
+        expect_u64(model, &format!("{prefix}.{key}"), expected)?;
+    }
+    expect_float(
+        model,
+        &format!("{prefix}.attention.layer_norm_rms_epsilon"),
+        shape.rms_epsilon,
+    )?;
+    expect_float(
+        model,
+        &format!("{prefix}.expert_weights_scale"),
+        shape.expert_weight_scale,
+    )?;
+    if !model.boolean(&format!("{prefix}.expert_weights_norm"))? {
+        return Err(format!("{prefix}.expert_weights_norm must be true"));
+    }
+    expect_float(model, &format!("{prefix}.rope.freq_base"), shape.rope_base)?;
+
+    if shape.family == ModelFamily::Glm {
+        for (key, expected) in [
+            ("context_length", shape.original_context),
+            ("feed_forward_length", shape.ff_dense),
+            ("attention.kv_lora_rank", shape.kv_lora),
+            ("attention.key_length_mla", shape.key_mla),
+            ("attention.value_length_mla", shape.value_mla),
+            ("expert_group_count", 1),
+            ("expert_group_used_count", 1),
+            ("expert_gating_func", 2),
+            ("leading_dense_block_count", u64::from(shape.leading_dense)),
+            ("nextn_predict_layers", u64::from(shape.nextn)),
+        ] {
+            expect_u64(model, &format!("{prefix}.{key}"), expected)?;
+        }
+        return Ok(());
+    }
+
+    for (key, expected) in [
+        ("attention.output_group_count", shape.out_groups),
+        ("attention.output_lora_rank", shape.lora_o),
+        ("hash_layer_count", u64::from(shape.hash_layers)),
+        ("attention.sliding_window", shape.sliding_window),
+        ("hyper_connection.count", shape.hc),
+        ("hyper_connection.sinkhorn_iterations", shape.hc_sinkhorn),
+    ] {
+        expect_u64(model, &format!("{prefix}.{key}"), expected)?;
+    }
+    for key in ["expert_group_count", "expert_group_used_count"] {
+        if let Some(Value::U32(value)) = model.metadata.get(&format!("{prefix}.{key}"))
+            && *value != 0
+        {
+            return Err(format!("{prefix}.{key} must be zero"));
+        }
+    }
+    for (key, expected) in [
+        ("hyper_connection.epsilon", shape.hc_epsilon),
+        (
+            "attention.compress_rope_freq_base",
+            shape.compress_rope_base,
+        ),
+    ] {
+        expect_float(model, &format!("{prefix}.{key}"), expected)?;
+    }
+    for (key, expected) in [
+        ("rope.scaling.factor", shape.rope_scale),
+        ("rope.scaling.yarn_beta_fast", shape.rope_beta_fast),
+        ("rope.scaling.yarn_beta_slow", shape.rope_beta_slow),
+    ] {
+        if model.metadata.contains_key(&format!("{prefix}.{key}")) {
+            expect_float(model, &format!("{prefix}.{key}"), expected)?;
+        }
+    }
+    if model
+        .metadata
+        .contains_key("deepseek4.rope.scaling.original_context_length")
+    {
+        expect_u64(
+            model,
+            "deepseek4.rope.scaling.original_context_length",
+            shape.original_context,
+        )?;
+    }
+    let ratios = model.u32s("deepseek4.attention.compress_ratios")?;
+    let clamps = model.f32s("deepseek4.swiglu_clamp_exp")?;
+    if ratios.len() < shape.layers as usize || clamps.len() < shape.layers as usize {
+        return Err("DeepSeek per-layer metadata is shorter than the layer count".into());
+    }
+    for layer in 0..shape.layers as usize {
+        let expected = compression_ratio(shape, layer as u32);
+        if ratios[layer] != expected {
+            return Err(format!(
+                "layer {layer} compression ratio is {}, expected {expected}",
+                ratios[layer]
+            ));
+        }
+        if !float_eq(clamps[layer], shape.swiglu_clamp) {
+            return Err(format!("layer {layer} has an invalid SwiGLU clamp"));
+        }
+    }
+    Ok(())
+}
+
+fn validate_tensors(model: &Gguf, shape: &Shape) -> Result<(), String> {
+    match shape.family {
+        ModelFamily::DeepSeek => validate_deepseek_tensors(model, shape),
+        ModelFamily::Glm => validate_glm_tensors(model, shape),
+    }
+}
+
+fn validate_deepseek_tensors(model: &Gguf, shape: &Shape) -> Result<(), String> {
+    let hc_dim = shape.embd * shape.hc;
+    let hc_mix = 2 * shape.hc + shape.hc * shape.hc;
+    let q_dim = shape.heads * shape.head_dim;
+    let output_low = shape.out_groups * shape.lora_o;
+    expect(
+        model,
+        "token_embd.weight",
+        &[F16],
+        &[shape.embd, shape.vocab],
+    )?;
+    expect(model, "output_hc_base.weight", &[F32], &[shape.hc])?;
+    expect(model, "output_hc_fn.weight", &[F16], &[hc_dim, shape.hc])?;
+    expect(model, "output_hc_scale.weight", &[F32], &[1])?;
+    expect(model, "output_norm.weight", &[F32], &[shape.embd])?;
+    expect(model, "output.weight", DENSE, &[shape.embd, shape.vocab])?;
+
+    for layer in 0..shape.layers {
+        let name = |suffix: &str| format!("blk.{layer}.{suffix}");
+        expect(model, &name("hc_attn_fn.weight"), &[F16], &[hc_dim, hc_mix])?;
+        expect(model, &name("hc_attn_scale.weight"), &[F32], &[3])?;
+        expect(model, &name("hc_attn_base.weight"), &[F32], &[hc_mix])?;
+        expect(model, &name("attn_norm.weight"), &[F32], &[shape.embd])?;
+        expect(
+            model,
+            &name("attn_q_a.weight"),
+            DENSE,
+            &[shape.embd, shape.lora_q],
+        )?;
+        expect(
+            model,
+            &name("attn_q_a_norm.weight"),
+            &[F32],
+            &[shape.lora_q],
+        )?;
+        expect(
+            model,
+            &name("attn_q_b.weight"),
+            DENSE,
+            &[shape.lora_q, q_dim],
+        )?;
+        expect(
+            model,
+            &name("attn_kv.weight"),
+            DENSE,
+            &[shape.embd, shape.head_dim],
+        )?;
+        expect(
+            model,
+            &name("attn_kv_a_norm.weight"),
+            &[F32],
+            &[shape.head_dim],
+        )?;
+        expect(model, &name("attn_sinks.weight"), &[F32], &[shape.heads])?;
+        expect(
+            model,
+            &name("attn_output_a.weight"),
+            DENSE,
+            &[
+                shape.head_dim * (shape.heads / shape.out_groups),
+                output_low,
+            ],
+        )?;
+        expect(
+            model,
+            &name("attn_output_b.weight"),
+            DENSE,
+            &[output_low, shape.embd],
+        )?;
+
+        let ratio = compression_ratio(shape, layer);
+        if ratio != 0 {
+            let compression_width = if ratio == 4 { 2 } else { 1 } * shape.head_dim;
+            expect(
+                model,
+                &name("attn_compressor_ape.weight"),
+                &[F16],
+                &[compression_width, u64::from(ratio)],
+            )?;
+            expect(
+                model,
+                &name("attn_compressor_kv.weight"),
+                &[F16],
+                &[shape.embd, compression_width],
+            )?;
+            expect(
+                model,
+                &name("attn_compressor_gate.weight"),
+                &[F16],
+                &[shape.embd, compression_width],
+            )?;
+            expect(
+                model,
+                &name("attn_compressor_norm.weight"),
+                &[F32],
+                &[shape.head_dim],
+            )?;
+        }
+        if ratio == 4 {
+            let index_q = shape.indexer_heads * shape.indexer_head_dim;
+            let index_width = 2 * shape.indexer_head_dim;
+            expect(
+                model,
+                &name("indexer.attn_q_b.weight"),
+                &[F16, Q8_0],
+                &[shape.lora_q, index_q],
+            )?;
+            expect(
+                model,
+                &name("indexer.proj.weight"),
+                &[F16],
+                &[shape.embd, shape.indexer_heads],
+            )?;
+            expect(
+                model,
+                &name("indexer_compressor_ape.weight"),
+                &[F16],
+                &[index_width, 4],
+            )?;
+            expect(
+                model,
+                &name("indexer_compressor_kv.weight"),
+                &[F16],
+                &[shape.embd, index_width],
+            )?;
+            expect(
+                model,
+                &name("indexer_compressor_gate.weight"),
+                &[F16],
+                &[shape.embd, index_width],
+            )?;
+            expect(
+                model,
+                &name("indexer_compressor_norm.weight"),
+                &[F32],
+                &[shape.indexer_head_dim],
+            )?;
+        }
+        expect(model, &name("hc_ffn_fn.weight"), &[F16], &[hc_dim, hc_mix])?;
+        expect(model, &name("hc_ffn_scale.weight"), &[F32], &[3])?;
+        expect(model, &name("hc_ffn_base.weight"), &[F32], &[hc_mix])?;
+        expect(model, &name("ffn_norm.weight"), &[F32], &[shape.embd])?;
+        expect(
+            model,
+            &name("ffn_gate_inp.weight"),
+            &[F16],
+            &[shape.embd, shape.experts],
+        )?;
+        expect_optional(model, &name("exp_probs_b.bias"), &[F32], &[shape.experts])?;
+        expect(
+            model,
+            &name("ffn_gate_exps.weight"),
+            ROUTED,
+            &[shape.embd, shape.ff_expert, shape.experts],
+        )?;
+        expect(
+            model,
+            &name("ffn_up_exps.weight"),
+            ROUTED,
+            &[shape.embd, shape.ff_expert, shape.experts],
+        )?;
+        expect(
+            model,
+            &name("ffn_down_exps.weight"),
+            ROUTED,
+            &[shape.ff_expert, shape.embd, shape.experts],
+        )?;
+        same_type(
+            model,
+            &name("ffn_gate_exps.weight"),
+            &name("ffn_up_exps.weight"),
+        )?;
+        expect(
+            model,
+            &name("ffn_gate_shexp.weight"),
+            DENSE,
+            &[shape.embd, shape.ff_expert],
+        )?;
+        expect(
+            model,
+            &name("ffn_up_shexp.weight"),
+            DENSE,
+            &[shape.embd, shape.ff_expert],
+        )?;
+        expect(
+            model,
+            &name("ffn_down_shexp.weight"),
+            DENSE,
+            &[shape.ff_expert, shape.embd],
+        )?;
+        if layer < shape.hash_layers {
+            expect(
+                model,
+                &name("ffn_gate_tid2eid.weight"),
+                &[I32],
+                &[shape.experts_used, shape.vocab],
+            )?;
+        }
+    }
+    Ok(())
+}
+
+fn validate_glm_tensors(model: &Gguf, shape: &Shape) -> Result<(), String> {
+    let q_dim = shape.heads * shape.key_mla;
+    let q_nope = shape.key_mla - shape.rot;
+    let index_q = shape.indexer_heads * shape.indexer_head_dim;
+    expect(
+        model,
+        "token_embd.weight",
+        DENSE,
+        &[shape.embd, shape.vocab],
+    )?;
+    expect(model, "output_norm.weight", &[F32], &[shape.embd])?;
+    expect(model, "output.weight", DENSE, &[shape.embd, shape.vocab])?;
+    for layer in 0..shape.layers {
+        let name = |suffix: &str| format!("blk.{layer}.{suffix}");
+        expect(model, &name("attn_norm.weight"), &[F32], &[shape.embd])?;
+        expect(
+            model,
+            &name("attn_q_a.weight"),
+            DENSE,
+            &[shape.embd, shape.lora_q],
+        )?;
+        expect(
+            model,
+            &name("attn_q_a_norm.weight"),
+            &[F32],
+            &[shape.lora_q],
+        )?;
+        expect(
+            model,
+            &name("attn_q_b.weight"),
+            DENSE,
+            &[shape.lora_q, q_dim],
+        )?;
+        expect(
+            model,
+            &name("attn_kv_a_mqa.weight"),
+            DENSE,
+            &[shape.embd, shape.head_dim],
+        )?;
+        expect(
+            model,
+            &name("attn_kv_a_norm.weight"),
+            &[F32],
+            &[shape.kv_lora],
+        )?;
+        expect(
+            model,
+            &name("attn_k_b.weight"),
+            DENSE,
+            &[q_nope, shape.kv_lora, shape.heads],
+        )?;
+        expect(
+            model,
+            &name("attn_v_b.weight"),
+            DENSE,
+            &[shape.kv_lora, shape.value_mla, shape.heads],
+        )?;
+        expect(
+            model,
+            &name("attn_output.weight"),
+            DENSE,
+            &[shape.heads * shape.value_mla, shape.embd],
+        )?;
+        expect(
+            model,
+            &name("indexer.attn_k.weight"),
+            DENSE,
+            &[shape.embd, shape.indexer_head_dim],
+        )?;
+        expect(
+            model,
+            &name("indexer.attn_q_b.weight"),
+            DENSE,
+            &[shape.lora_q, index_q],
+        )?;
+        expect(
+            model,
+            &name("indexer.k_norm.weight"),
+            &[F32],
+            &[shape.indexer_head_dim],
+        )?;
+        expect(
+            model,
+            &name("indexer.k_norm.bias"),
+            &[F32],
+            &[shape.indexer_head_dim],
+        )?;
+        expect(
+            model,
+            &name("indexer.proj.weight"),
+            &[F32],
+            &[shape.embd, shape.indexer_heads],
+        )?;
+        expect(model, &name("ffn_norm.weight"), &[F32], &[shape.embd])?;
+        if layer < shape.leading_dense {
+            expect(
+                model,
+                &name("ffn_gate.weight"),
+                DENSE,
+                &[shape.embd, shape.ff_dense],
+            )?;
+            expect(
+                model,
+                &name("ffn_up.weight"),
+                DENSE,
+                &[shape.embd, shape.ff_dense],
+            )?;
+            expect(
+                model,
+                &name("ffn_down.weight"),
+                DENSE,
+                &[shape.ff_dense, shape.embd],
+            )?;
+        } else {
+            expect(
+                model,
+                &name("ffn_gate_inp.weight"),
+                &[F32],
+                &[shape.embd, shape.experts],
+            )?;
+            expect(model, &name("exp_probs_b.bias"), &[F32], &[shape.experts])?;
+            expect(
+                model,
+                &name("ffn_gate_exps.weight"),
+                ROUTED,
+                &[shape.embd, shape.ff_expert, shape.experts],
+            )?;
+            expect(
+                model,
+                &name("ffn_up_exps.weight"),
+                ROUTED,
+                &[shape.embd, shape.ff_expert, shape.experts],
+            )?;
+            expect(
+                model,
+                &name("ffn_down_exps.weight"),
+                ROUTED,
+                &[shape.ff_expert, shape.embd, shape.experts],
+            )?;
+            same_type(
+                model,
+                &name("ffn_gate_exps.weight"),
+                &name("ffn_up_exps.weight"),
+            )?;
+            expect(
+                model,
+                &name("ffn_gate_shexp.weight"),
+                DENSE,
+                &[shape.embd, shape.ff_expert],
+            )?;
+            expect(
+                model,
+                &name("ffn_up_shexp.weight"),
+                DENSE,
+                &[shape.embd, shape.ff_expert],
+            )?;
+            expect(
+                model,
+                &name("ffn_down_shexp.weight"),
+                DENSE,
+                &[shape.ff_expert, shape.embd],
+            )?;
+        }
+        if layer + shape.nextn >= shape.layers {
+            expect(
+                model,
+                &name("nextn.eh_proj.weight"),
+                DENSE,
+                &[2 * shape.embd, shape.embd],
+            )?;
+            expect(model, &name("nextn.enorm.weight"), &[F32], &[shape.embd])?;
+            expect(model, &name("nextn.hnorm.weight"), &[F32], &[shape.embd])?;
+            expect(
+                model,
+                &name("nextn.shared_head_norm.weight"),
+                &[F32],
+                &[shape.embd],
+            )?;
+        }
+    }
+    Ok(())
+}
+
+pub(super) fn validate_dspark(model: &Gguf, shape: &Shape) -> Result<(), String> {
+    if shape.model != ModelChoice::DeepSeekV4Flash {
+        return Err("DSpark support is available only for DeepSeek V4 Flash".into());
+    }
+    let block_size = first_u32(
+        model,
+        &[
+            "deepseek4.dspark.block_size",
+            "deepseek4.dspark_block_size",
+            "dspark.block_size",
+        ],
+    )?;
+    let markov_rank = first_u32(
+        model,
+        &[
+            "deepseek4.dspark.markov_rank",
+            "deepseek4.dspark_markov_rank",
+            "dspark.markov_rank",
+        ],
+    )?;
+    let noise_token = first_u32(
+        model,
+        &[
+            "deepseek4.dspark.noise_token_id",
+            "deepseek4.dspark_noise_token_id",
+            "dspark.noise_token_id",
+        ],
+    )?;
+    let targets = first_u32s(
+        model,
+        &[
+            "deepseek4.dspark.target_layer_ids",
+            "deepseek4.dspark_target_layer_ids",
+            "dspark.target_layer_ids",
+        ],
+    )?;
+    if !(1..=16).contains(&block_size) || markov_rank == 0 || noise_token >= shape.vocab as u32 {
+        return Err("invalid DSpark block, Markov, or noise-token metadata".into());
+    }
+    if targets.is_empty()
+        || targets.len() > 8
+        || targets.windows(2).any(|pair| pair[0] >= pair[1])
+        || targets.iter().any(|layer| *layer >= shape.layers)
+    {
+        return Err("invalid DSpark target-layer metadata".into());
+    }
+    let stages = model
+        .tensors
+        .keys()
+        .filter_map(|name| {
+            name.strip_prefix("mtp.")?
+                .split('.')
+                .next()?
+                .parse::<u32>()
+                .ok()
+        })
+        .max()
+        .map_or(0, |stage| stage + 1);
+    if !(1..=8).contains(&stages) {
+        return Err(format!("invalid DSpark stage count: {stages}"));
+    }
+    for stage in 0..stages {
+        validate_dspark_block(model, shape, stage)?;
+        if stage == 0 {
+            expect(
+                model,
+                &format!("mtp.{stage}.main_proj.weight"),
+                DSPARK_DENSE,
+                &[targets.len() as u64 * shape.embd, shape.embd],
+            )?;
+            expect(
+                model,
+                &format!("mtp.{stage}.main_norm.weight"),
+                &[F32],
+                &[shape.embd],
+            )?;
+        }
+    }
+    let prefix = format!("mtp.{}", stages - 1);
+    expect(
+        model,
+        &format!("{prefix}.norm.weight"),
+        &[F32],
+        &[shape.embd],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.hc_head_base.weight"),
+        &[F32],
+        &[shape.hc],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.hc_head_fn.weight"),
+        PLAIN,
+        &[shape.embd * shape.hc, shape.hc],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.hc_head_scale.weight"),
+        &[F32],
+        &[1],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.markov_head.markov_w1.weight"),
+        DSPARK_DENSE,
+        &[u64::from(markov_rank), shape.vocab],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.markov_head.markov_w2.weight"),
+        DSPARK_DENSE,
+        &[u64::from(markov_rank), shape.vocab],
+    )?;
+    expect(
+        model,
+        &format!("{prefix}.confidence_head.proj.weight"),
+        DSPARK_DENSE,
+        &[shape.embd + u64::from(markov_rank), 1],
+    )?;
+    Ok(())
+}
+
+fn validate_dspark_block(model: &Gguf, shape: &Shape, stage: u32) -> Result<(), String> {
+    let hc_dim = shape.embd * shape.hc;
+    let hc_mix = 2 * shape.hc + shape.hc * shape.hc;
+    let q_dim = shape.heads * shape.head_dim;
+    let output_low = shape.out_groups * shape.lora_o;
+    let name = |suffix: &str| format!("mtp.{stage}.{suffix}");
+    for (suffix, types, dims) in [
+        ("hc_attn_fn.weight", PLAIN, vec![hc_dim, hc_mix]),
+        ("hc_attn_scale.weight", &[F32][..], vec![3]),
+        ("hc_attn_base.weight", &[F32][..], vec![hc_mix]),
+        ("attn_norm.weight", &[F32][..], vec![shape.embd]),
+        (
+            "attn_q_a.weight",
+            DSPARK_DENSE,
+            vec![shape.embd, shape.lora_q],
+        ),
+        ("attn_q_a_norm.weight", &[F32][..], vec![shape.lora_q]),
+        ("attn_q_b.weight", DSPARK_DENSE, vec![shape.lora_q, q_dim]),
+        (
+            "attn_kv.weight",
+            DSPARK_DENSE,
+            vec![shape.embd, shape.head_dim],
+        ),
+        ("attn_kv_a_norm.weight", &[F32][..], vec![shape.head_dim]),
+        ("attn_sinks.weight", &[F32][..], vec![shape.heads]),
+        (
+            "attn_output_a.weight",
+            DSPARK_DENSE,
+            vec![
+                shape.head_dim * (shape.heads / shape.out_groups),
+                output_low,
+            ],
+        ),
+        (
+            "attn_output_b.weight",
+            DSPARK_DENSE,
+            vec![output_low, shape.embd],
+        ),
+        ("hc_ffn_fn.weight", PLAIN, vec![hc_dim, hc_mix]),
+        ("hc_ffn_scale.weight", &[F32][..], vec![3]),
+        ("hc_ffn_base.weight", &[F32][..], vec![hc_mix]),
+        ("ffn_norm.weight", &[F32][..], vec![shape.embd]),
+        (
+            "ffn_gate_inp.weight",
+            DSPARK_DENSE,
+            vec![shape.embd, shape.experts],
+        ),
+        ("exp_probs_b.bias", &[F32][..], vec![shape.experts]),
+        (
+            "ffn_gate_exps.weight",
+            ROUTED,
+            vec![shape.embd, shape.ff_expert, shape.experts],
+        ),
+        (
+            "ffn_up_exps.weight",
+            ROUTED,
+            vec![shape.embd, shape.ff_expert, shape.experts],
+        ),
+        (
+            "ffn_down_exps.weight",
+            ROUTED,
+            vec![shape.ff_expert, shape.embd, shape.experts],
+        ),
+        (
+            "ffn_gate_shexp.weight",
+            DSPARK_DENSE,
+            vec![shape.embd, shape.ff_expert],
+        ),
+        (
+            "ffn_up_shexp.weight",
+            DSPARK_DENSE,
+            vec![shape.embd, shape.ff_expert],
+        ),
+        (
+            "ffn_down_shexp.weight",
+            DSPARK_DENSE,
+            vec![shape.ff_expert, shape.embd],
+        ),
+    ] {
+        expect(model, &name(suffix), types, &dims)?;
+    }
+    same_type(
+        model,
+        &name("ffn_gate_exps.weight"),
+        &name("ffn_up_exps.weight"),
+    )
+}
+
+fn expect(model: &Gguf, name: &str, types: &[u32], dims: &[u64]) -> Result<(), String> {
+    validate_tensor(name, model.tensor(name)?, types, dims)
+}
+
+fn expect_optional(model: &Gguf, name: &str, types: &[u32], dims: &[u64]) -> Result<(), String> {
+    match model.tensors.get(name) {
+        Some(tensor) => validate_tensor(name, tensor, types, dims),
+        None => Ok(()),
+    }
+}
+
+fn validate_tensor(name: &str, tensor: &Tensor, types: &[u32], dims: &[u64]) -> Result<(), String> {
+    if !types.contains(&tensor.kind) {
+        return Err(format!(
+            "tensor {name} has unsupported type {}",
+            tensor.kind
+        ));
+    }
+    if tensor.dims != dims {
+        return Err(format!(
+            "tensor {name} has dimensions {:?}, expected {dims:?}",
+            tensor.dims
+        ));
+    }
+    Ok(())
+}
+
+fn same_type(model: &Gguf, first: &str, second: &str) -> Result<(), String> {
+    if model.tensor(first)?.kind != model.tensor(second)?.kind {
+        Err(format!(
+            "tensors {first} and {second} use different quantizations"
+        ))
+    } else {
+        Ok(())
+    }
+}
+
+fn expect_u64(model: &Gguf, key: &str, expected: u64) -> Result<(), String> {
+    let actual = model.u64(key)?;
+    if actual == expected {
+        Ok(())
+    } else {
+        Err(format!("{key} is {actual}, expected {expected}"))
+    }
+}
+
+fn expect_float(model: &Gguf, key: &str, expected: f32) -> Result<(), String> {
+    let actual = model.f32(key)?;
+    if float_eq(actual, expected) {
+        Ok(())
+    } else {
+        Err(format!("{key} is {actual}, expected {expected}"))
+    }
+}
+
+fn float_eq(actual: f32, expected: f32) -> bool {
+    actual.is_finite() && (actual - expected).abs() <= expected.abs().max(1.0) * 1.0e-6
+}
+
+fn compression_ratio(shape: &Shape, layer: u32) -> u32 {
+    match shape.model {
+        ModelChoice::DeepSeekV4Flash if layer < 2 => 0,
+        ModelChoice::DeepSeekV4Pro if layer < 2 => 128,
+        ModelChoice::DeepSeekV4Flash | ModelChoice::DeepSeekV4Pro if layer.is_multiple_of(2) => 4,
+        ModelChoice::DeepSeekV4Flash | ModelChoice::DeepSeekV4Pro => 128,
+        ModelChoice::Glm52 => 0,
+    }
+}
+
+fn first_u32(model: &Gguf, keys: &[&str]) -> Result<u32, String> {
+    keys.iter()
+        .find_map(|key| model.u32(key).ok())
+        .ok_or_else(|| format!("required DSpark metadata is missing: {}", keys[0]))
+}
+
+fn first_u32s<'a>(model: &'a Gguf, keys: &[&str]) -> Result<&'a [u32], String> {
+    keys.iter()
+        .find_map(|key| model.u32s(key).ok())
+        .ok_or_else(|| format!("required DSpark metadata is missing: {}", keys[0]))
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn installed_ds4_fixture_opens_and_renders_a_prompt() {
+        let path = Path::new(
+            "../ds4/gguf/DeepSeek-V4-Flash-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-chat-v2-imatrix.gguf",
+        );
+        if !path.exists() {
+            return;
+        }
+        let model = Model::open_main(path, ModelChoice::DeepSeekV4Flash).unwrap();
+        let summary = model.summary();
+        assert_eq!(summary.model, ModelChoice::DeepSeekV4Flash);
+        assert_eq!(summary.vocabulary_size, 129_280);
+        assert_eq!(
+            model.tokenize("Hello, world! 1234\nint café = 7;\n中文テスト"),
+            [
+                19_923, 14, 2_058, 3, 223, 6_895, 22, 201, 650, 57_664, 438, 223, 25, 510, 21_134,
+                109_288,
+            ]
+        );
+        assert_eq!(
+            model.render_prompt("", "Hello", ReasoningMode::Direct),
+            [0, 128_803, 19_923, 128_804, 128_822]
+        );
+        assert_eq!(model.tokenizer.tokenize_rendered("｜DSML｜").len(), 1);
+        assert!(model.is_stop_token_for_reasoning(128_822, ReasoningMode::Direct));
+        assert!(!model.is_stop_token_for_reasoning(128_822, ReasoningMode::High));
+        let think_start = *model
+            .render_prompt("", "Hello", ReasoningMode::High)
+            .last()
+            .unwrap();
+        assert_eq!(
+            model.render_conversation(
+                "",
+                &[
+                    ChatTurn {
+                        user: true,
+                        skip_previous_eos: false,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                    ChatTurn {
+                        user: false,
+                        skip_previous_eos: false,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                    ChatTurn {
+                        user: true,
+                        skip_previous_eos: false,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                ],
+                ReasoningMode::Direct,
+            ),
+            [
+                0, 128_803, 19_923, 128_804, 128_822, 19_923, 1, 128_803, 19_923, 128_804, 128_822,
+            ]
+        );
+        assert_eq!(
+            model.render_continuation("Hello", ReasoningMode::Direct, false),
+            [1, 128_803, 19_923, 128_804, 128_822]
+        );
+        assert_eq!(
+            model.render_continuation("Hello", ReasoningMode::Direct, true),
+            [128_803, 19_923, 128_804, 128_822]
+        );
+        assert_eq!(
+            model.render_conversation(
+                "",
+                &[
+                    ChatTurn {
+                        user: true,
+                        skip_previous_eos: false,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                    ChatTurn {
+                        user: false,
+                        skip_previous_eos: false,
+                        reasoning: Some("Hello".into()),
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                    ChatTurn {
+                        user: true,
+                        skip_previous_eos: false,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: "Hello".into(),
+                    },
+                ],
+                ReasoningMode::Direct,
+            ),
+            [
+                0,
+                128_803,
+                19_923,
+                128_804,
+                think_start,
+                19_923,
+                128_822,
+                19_923,
+                1,
+                128_803,
+                19_923,
+                128_804,
+                128_822,
+            ]
+        );
+    }
+
+    #[test]
+    fn installed_dspark_fixture_passes_the_target_layout() {
+        let path = Path::new("../ds4/gguf/DeepSeek-V4-Flash-DSpark-support.gguf");
+        if path.exists() {
+            validate_model_artifact(path, ModelChoice::DeepSeekV4Flash, true).unwrap();
+        }
+    }
+}

src/model.rs

@@ -1,10 +1,12 @@
-use std::fmt;
+mod transfer;
-use std::fs::{self, File, OpenOptions};
-use std::io::{Read, Write};
-use std::path::{Path, PathBuf};
-use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
 
-use sha2::{Digest, Sha256};
+pub(crate) use transfer::{
+    delete_managed_artifact, download_managed_artifact, validate_managed_artifact,
+};
+
+use std::fmt;
+use std::fs;
+use std::path::{Path, PathBuf};
 
 pub(crate) const MODEL_CHOICES: [ModelChoice; 3] = [
     ModelChoice::DeepSeekV4Flash,
@@ -379,505 +381,3 @@ pub(crate) fn artifact_verification_progress(
         }),
     }
 }
-
-pub(crate) fn download_managed_artifact(
-    id: ManagedArtifactId,
-    models_path: &Path,
-    cancel: &AtomicBool,
-    verified_bytes: &AtomicU64,
-) -> Result<DownloadOutcome, String> {
-    download_artifact_with_cancel(
-        id.model(),
-        id.artifact(),
-        models_path,
-        cancel,
-        verified_bytes,
-    )
-}
-
-pub(crate) fn validate_managed_artifact(
-    id: ManagedArtifactId,
-    models_path: &Path,
-    cancel: &AtomicBool,
-    verified_bytes: &AtomicU64,
-) -> Result<DownloadOutcome, String> {
-    let model = id.model();
-    let artifact = id.artifact();
-    let destination = artifact.path(model, models_path);
-    let partial = artifact.partial_path(model, models_path);
-    let (path, promote) = if destination.exists() {
-        (destination.clone(), false)
-    } else if partial.exists() {
-        (partial.clone(), true)
-    } else {
-        return Err(format!("{} is not downloaded", artifact.label));
-    };
-
-    match verify(&path, artifact, model, cancel, verified_bytes) {
-        Ok(DownloadOutcome::Stopped) => return Ok(DownloadOutcome::Stopped),
-        Ok(DownloadOutcome::Complete) => {}
-        Err(error) => {
-            let marker = artifact.verification_path(model, models_path);
-            if let Err(remove_error) = fs::remove_file(marker)
-                && remove_error.kind() != std::io::ErrorKind::NotFound
-            {
-                return Err(format!(
-                    "{error}; could not remove checksum marker: {remove_error}"
-                ));
-            }
-            return Err(error);
-        }
-    }
-    if promote {
-        fs::rename(partial, destination).map_err(|error| error.to_string())?;
-    }
-    mark_verified(model, artifact, models_path)?;
-    Ok(DownloadOutcome::Complete)
-}
-
-pub(crate) fn delete_managed_artifact(
-    id: ManagedArtifactId,
-    models_path: &Path,
-) -> Result<(), String> {
-    let model = id.model();
-    let artifact = id.artifact();
-    for path in [
-        artifact.path(model, models_path),
-        artifact.partial_path(model, models_path),
-        artifact.verification_path(model, models_path),
-    ] {
-        match fs::remove_file(path) {
-            Ok(()) => {}
-            Err(error) if error.kind() == std::io::ErrorKind::NotFound => {}
-            Err(error) => return Err(error.to_string()),
-        }
-    }
-    Ok(())
-}
-
-#[cfg(test)]
-fn download_artifact(
-    model: ModelChoice,
-    artifact: &Artifact,
-    models_path: &Path,
-) -> Result<DownloadOutcome, String> {
-    download_artifact_with_cancel(
-        model,
-        artifact,
-        models_path,
-        &AtomicBool::new(false),
-        &AtomicU64::new(0),
-    )
-}
-
-fn download_artifact_with_cancel(
-    model: ModelChoice,
-    artifact: &Artifact,
-    models_path: &Path,
-    cancel: &AtomicBool,
-    verified_bytes: &AtomicU64,
-) -> Result<DownloadOutcome, String> {
-    if cancel.load(Ordering::Relaxed) {
-        return Ok(DownloadOutcome::Stopped);
-    }
-    let destination = artifact.path(model, models_path);
-    if artifact.is_installed(model, models_path) {
-        return Ok(DownloadOutcome::Complete);
-    }
-    if destination.exists() {
-        if verify(&destination, artifact, model, cancel, verified_bytes)?
-            == DownloadOutcome::Stopped
-        {
-            return Ok(DownloadOutcome::Stopped);
-        }
-        mark_verified(model, artifact, models_path)?;
-        return Ok(DownloadOutcome::Complete);
-    }
-
-    let directory = destination
-        .parent()
-        .ok_or_else(|| "model artifact path has no parent directory".to_owned())?;
-    fs::create_dir_all(directory).map_err(|error| error.to_string())?;
-    let partial = artifact.partial_path(model, models_path);
-    let partial_size = partial.metadata().map_or(0, |metadata| metadata.len());
-    if partial_size > artifact.size {
-        File::create(&partial).map_err(|error| error.to_string())?;
-    }
-    if partial.metadata().map_or(0, |metadata| metadata.len()) != artifact.size
-        && download_to_partial(artifact, &partial, cancel)? == DownloadOutcome::Stopped
-    {
-        return Ok(DownloadOutcome::Stopped);
-    }
-
-    if cancel.load(Ordering::Relaxed) {
-        return Ok(DownloadOutcome::Stopped);
-    }
-    match verify(&partial, artifact, model, cancel, verified_bytes) {
-        Ok(DownloadOutcome::Stopped) => return Ok(DownloadOutcome::Stopped),
-        Ok(DownloadOutcome::Complete) => {}
-        Err(error) => {
-            fs::remove_file(&partial).map_err(|remove_error| {
-                format!("{error}; could not remove partial file: {remove_error}")
-            })?;
-            return Err(error);
-        }
-    }
-    fs::rename(partial, destination).map_err(|error| error.to_string())?;
-    mark_verified(model, artifact, models_path)?;
-    Ok(DownloadOutcome::Complete)
-}
-
-fn mark_verified(
-    model: ModelChoice,
-    artifact: &Artifact,
-    models_path: &Path,
-) -> Result<(), String> {
-    fs::write(
-        artifact.verification_path(model, models_path),
-        artifact.sha256,
-    )
-    .map_err(|error| error.to_string())
-}
-
-fn verify(
-    path: &Path,
-    artifact: &Artifact,
-    model: ModelChoice,
-    cancel: &AtomicBool,
-    verified_bytes: &AtomicU64,
-) -> Result<DownloadOutcome, String> {
-    verified_bytes.store(0, Ordering::Relaxed);
-    let size = path.metadata().map_err(|error| error.to_string())?.len();
-    if size != artifact.size {
-        return Err(format!(
-            "{} has size {size}, expected {}",
-            path.display(),
-            artifact.size
-        ));
-    }
-
-    let mut file = File::open(path).map_err(|error| error.to_string())?;
-    let mut hasher = Sha256::new();
-    let mut buffer = vec![0; 1024 * 1024];
-    loop {
-        if cancel.load(Ordering::Relaxed) {
-            return Ok(DownloadOutcome::Stopped);
-        }
-        let count = file.read(&mut buffer).map_err(|error| error.to_string())?;
-        if count == 0 {
-            break;
-        }
-        hasher.update(&buffer[..count]);
-        verified_bytes.fetch_add(count as u64, Ordering::Relaxed);
-    }
-    let actual = hex(&hasher.finalize());
-    if actual != artifact.sha256 {
-        return Err(format!(
-            "Checksum verification failed for {}",
-            path.display()
-        ));
-    }
-    if let Some(support) = artifact.support {
-        crate::engine::validate_model_artifact(path, model, support)?;
-    }
-    Ok(DownloadOutcome::Complete)
-}
-
-fn hex(bytes: &[u8]) -> String {
-    const DIGITS: &[u8; 16] = b"0123456789abcdef";
-    let mut result = String::with_capacity(bytes.len() * 2);
-    for byte in bytes {
-        result.push(DIGITS[(byte >> 4) as usize] as char);
-        result.push(DIGITS[(byte & 0x0f) as usize] as char);
-    }
-    result
-}
-
-fn download_to_partial(
-    artifact: &Artifact,
-    partial: &Path,
-    cancel: &AtomicBool,
-) -> Result<DownloadOutcome, String> {
-    download_url_to_partial(&artifact.url(), partial, cancel)
-}
-
-fn download_url_to_partial(
-    url: &str,
-    partial: &Path,
-    cancel: &AtomicBool,
-) -> Result<DownloadOutcome, String> {
-    let offset = partial.metadata().map_or(0, |metadata| metadata.len());
-    let agent: ureq::Agent = ureq::Agent::config_builder()
-        .https_only(url.starts_with("https://"))
-        .build()
-        .into();
-    let mut request = agent.get(url);
-    if offset > 0 {
-        request = request.header("Range", format!("bytes={offset}-"));
-    }
-    let mut response = request
-        .call()
-        .map_err(|error| format!("Model download failed: {error}"))?;
-    let status = response.status().as_u16();
-    let append = offset > 0 && status == 206;
-    if offset > 0 && status != 200 && status != 206 {
-        return Err(format!(
-            "Model server returned HTTP {status} while resuming at byte {offset}"
-        ));
-    }
-    if append {
-        validate_content_range(&response, offset)?;
-    }
-
-    let mut output = OpenOptions::new()
-        .create(true)
-        .write(true)
-        .append(append)
-        .truncate(!append)
-        .open(partial)
-        .map_err(|error| error.to_string())?;
-    let mut body = response.body_mut().as_reader();
-    let mut buffer = vec![0; 1024 * 1024];
-    loop {
-        if cancel.load(Ordering::Relaxed) {
-            return Ok(DownloadOutcome::Stopped);
-        }
-        let count = body
-            .read(&mut buffer)
-            .map_err(|error| format!("Model download failed: {error}"))?;
-        if count == 0 {
-            break;
-        }
-        output
-            .write_all(&buffer[..count])
-            .map_err(|error| error.to_string())?;
-    }
-    Ok(DownloadOutcome::Complete)
-}
-
-fn validate_content_range(
-    response: &ureq::http::Response<ureq::Body>,
-    offset: u64,
-) -> Result<(), String> {
-    let expected = format!("bytes {offset}-");
-    let content_range = response
-        .headers()
-        .get("content-range")
-        .and_then(|value| value.to_str().ok())
-        .unwrap_or_default();
-    if content_range.starts_with(&expected) {
-        Ok(())
-    } else {
-        Err(format!(
-            "Model server returned an invalid Content-Range while resuming at byte {offset}"
-        ))
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use std::net::TcpListener;
-    use std::thread;
-    use std::time::{SystemTime, UNIX_EPOCH};
-
-    #[test]
-    fn catalog_and_checksum_verification_are_explicit() {
-        assert_eq!(ModelChoice::from_id("glm-5.2"), Some(ModelChoice::Glm52));
-        assert!(ModelChoice::from_id("unknown").is_none());
-        assert_eq!(
-            ModelChoice::DeepSeekV4Flash.main_artifact().size,
-            86_720_111_488
-        );
-        assert_eq!(ModelChoice::Glm52.main_artifact().size, 211_075_856_448);
-        assert_eq!(ModelChoice::DeepSeekV4Flash.artifacts(true).count(), 2);
-        assert_eq!(ModelChoice::Glm52.artifacts(true).count(), 1);
-
-        let id = SystemTime::now()
-            .duration_since(UNIX_EPOCH)
-            .unwrap()
-            .as_nanos();
-        let models_path = std::env::temp_dir().join(format!("ds4-server-models-{id}"));
-        let engine = engine_artifacts(ModelChoice::DeepSeekV4Flash, true, &models_path);
-        assert_eq!(
-            engine.model.file_name(),
-            Some(std::ffi::OsStr::new(FLASH.file_name))
-        );
-        assert_eq!(
-            engine.mtp.as_deref().and_then(Path::file_name),
-            Some(std::ffi::OsStr::new(FLASH_DSPARK.file_name))
-        );
-        let empty = Artifact {
-            label: "empty",
-            file_name: "empty",
-            repository: "",
-            size: 0,
-            sha256: "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
-            support: None,
-        };
-        let partial = empty.partial_path(ModelChoice::DeepSeekV4Flash, &models_path);
-        fs::create_dir_all(partial.parent().unwrap()).unwrap();
-        fs::write(&partial, []).unwrap();
-        download_artifact(ModelChoice::DeepSeekV4Flash, &empty, &models_path).unwrap();
-        assert!(empty.is_installed(ModelChoice::DeepSeekV4Flash, &models_path));
-        assert!(!partial.exists());
-        assert_eq!(
-            fs::read_to_string(empty.verification_path(ModelChoice::DeepSeekV4Flash, &models_path))
-                .unwrap(),
-            empty.sha256
-        );
-        fs::remove_dir_all(models_path).unwrap();
-    }
-
-    #[test]
-    fn verification_reports_bytes_read() {
-        let path = std::env::temp_dir().join(format!(
-            "ds4-server-verify-{}",
-            SystemTime::now()
-                .duration_since(UNIX_EPOCH)
-                .unwrap()
-                .as_nanos()
-        ));
-        fs::write(&path, b"abc").unwrap();
-        let artifact = Artifact {
-            label: "test model",
-            file_name: "unused",
-            repository: "unused",
-            size: 3,
-            sha256: "ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad",
-            support: None,
-        };
-        let verified_bytes = AtomicU64::new(999);
-
-        assert_eq!(
-            verify(
-                &path,
-                &artifact,
-                ModelChoice::DeepSeekV4Flash,
-                &AtomicBool::new(false),
-                &verified_bytes,
-            )
-            .unwrap(),
-            DownloadOutcome::Complete
-        );
-        assert_eq!(verified_bytes.load(Ordering::Relaxed), 3);
-        fs::remove_file(path).unwrap();
-    }
-
-    #[test]
-    fn managed_artifact_inventory_and_delete_include_partial_files() {
-        let models_path = std::env::temp_dir().join(format!(
-            "ds4-server-manager-{}",
-            SystemTime::now()
-                .duration_since(UNIX_EPOCH)
-                .unwrap()
-                .as_nanos()
-        ));
-        let id = ManagedArtifactId::DeepSeekV4Flash;
-        let partial = id.artifact().partial_path(id.model(), &models_path);
-        fs::create_dir_all(partial.parent().unwrap()).unwrap();
-        fs::write(&partial, b"part").unwrap();
-
-        let managed = managed_artifacts(&models_path)
-            .into_iter()
-            .find(|artifact| artifact.id == id)
-            .unwrap();
-        assert_eq!(managed.stored, 4);
-        assert_eq!(managed.state, ManagedArtifactState::Partial);
-
-        delete_managed_artifact(id, &models_path).unwrap();
-        assert!(!partial.exists());
-        fs::remove_dir_all(models_path).unwrap();
-    }
-
-    #[test]
-    fn restart_resumes_at_the_existing_partial_byte() {
-        let content = b"restart-resume works";
-        let offset = 8;
-        let directory = std::env::temp_dir().join(format!(
-            "ds4-server-resume-{}",
-            SystemTime::now()
-                .duration_since(UNIX_EPOCH)
-                .unwrap()
-                .as_nanos()
-        ));
-        fs::create_dir_all(&directory).unwrap();
-        let partial = directory.join("model.gguf.part");
-        fs::write(&partial, &content[..offset]).unwrap();
-
-        let listener = match TcpListener::bind("127.0.0.1:0") {
-            Ok(listener) => listener,
-            Err(error) if error.kind() == std::io::ErrorKind::PermissionDenied => {
-                fs::remove_dir_all(directory).unwrap();
-                return;
-            }
-            Err(error) => panic!("could not start test server: {error}"),
-        };
-        let address = listener.local_addr().unwrap();
-        let server = thread::spawn(move || {
-            let (mut connection, _) = listener.accept().unwrap();
-            let mut request = [0; 2048];
-            let count = connection.read(&mut request).unwrap();
-            let request = String::from_utf8_lossy(&request[..count]).to_ascii_lowercase();
-            assert!(request.contains("range: bytes=8-"));
-            let remaining = &content[offset..];
-            write!(
-                connection,
-                "HTTP/1.1 206 Partial Content\r\nContent-Length: {}\r\nContent-Range: bytes {offset}-{}/{}\r\nConnection: close\r\n\r\n",
-                remaining.len(),
-                content.len() - 1,
-                content.len(),
-            )
-            .unwrap();
-            connection.write_all(remaining).unwrap();
-        });
-
-        let outcome = download_url_to_partial(
-            &format!("http://{address}/model.gguf"),
-            &partial,
-            &AtomicBool::new(false),
-        )
-        .unwrap();
-        server.join().unwrap();
-        assert_eq!(outcome, DownloadOutcome::Complete);
-        assert_eq!(fs::read(&partial).unwrap(), content);
-        fs::remove_dir_all(directory).unwrap();
-    }
-
-    #[test]
-    fn cancellation_keeps_the_partial_file_for_the_next_run() {
-        let directory = std::env::temp_dir().join(format!(
-            "ds4-server-cancel-{}",
-            SystemTime::now()
-                .duration_since(UNIX_EPOCH)
-                .unwrap()
-                .as_nanos()
-        ));
-        let artifact = Artifact {
-            label: "test model",
-            file_name: "model.gguf",
-            repository: "unused",
-            size: 10,
-            sha256: "unused",
-            support: None,
-        };
-        let partial = artifact.partial_path(ModelChoice::DeepSeekV4Flash, &directory);
-        fs::create_dir_all(partial.parent().unwrap()).unwrap();
-        fs::write(&partial, b"part").unwrap();
-        let cancel = AtomicBool::new(true);
-
-        assert_eq!(
-            download_artifact_with_cancel(
-                ModelChoice::DeepSeekV4Flash,
-                &artifact,
-                &directory,
-                &cancel,
-                &AtomicU64::new(0),
-            )
-            .unwrap(),
-            DownloadOutcome::Stopped
-        );
-        assert_eq!(fs::read(&partial).unwrap(), b"part");
-        fs::remove_dir_all(directory).unwrap();
-    }
-}

src/model/transfer.rs

@@ -0,0 +1,507 @@
+use super::*;
+use sha2::{Digest, Sha256};
+use std::fs::{self, File, OpenOptions};
+use std::io::{Read, Write};
+use std::sync::atomic::{AtomicBool, AtomicU64, Ordering};
+
+pub(crate) fn download_managed_artifact(
+    id: ManagedArtifactId,
+    models_path: &Path,
+    cancel: &AtomicBool,
+    verified_bytes: &AtomicU64,
+) -> Result<DownloadOutcome, String> {
+    download_artifact_with_cancel(
+        id.model(),
+        id.artifact(),
+        models_path,
+        cancel,
+        verified_bytes,
+    )
+}
+
+pub(crate) fn validate_managed_artifact(
+    id: ManagedArtifactId,
+    models_path: &Path,
+    cancel: &AtomicBool,
+    verified_bytes: &AtomicU64,
+) -> Result<DownloadOutcome, String> {
+    let model = id.model();
+    let artifact = id.artifact();
+    let destination = artifact.path(model, models_path);
+    let partial = artifact.partial_path(model, models_path);
+    let (path, promote) = if destination.exists() {
+        (destination.clone(), false)
+    } else if partial.exists() {
+        (partial.clone(), true)
+    } else {
+        return Err(format!("{} is not downloaded", artifact.label));
+    };
+
+    match verify(&path, artifact, model, cancel, verified_bytes) {
+        Ok(DownloadOutcome::Stopped) => return Ok(DownloadOutcome::Stopped),
+        Ok(DownloadOutcome::Complete) => {}
+        Err(error) => {
+            let marker = artifact.verification_path(model, models_path);
+            if let Err(remove_error) = fs::remove_file(marker)
+                && remove_error.kind() != std::io::ErrorKind::NotFound
+            {
+                return Err(format!(
+                    "{error}; could not remove checksum marker: {remove_error}"
+                ));
+            }
+            return Err(error);
+        }
+    }
+    if promote {
+        fs::rename(partial, destination).map_err(|error| error.to_string())?;
+    }
+    mark_verified(model, artifact, models_path)?;
+    Ok(DownloadOutcome::Complete)
+}
+
+pub(crate) fn delete_managed_artifact(
+    id: ManagedArtifactId,
+    models_path: &Path,
+) -> Result<(), String> {
+    let model = id.model();
+    let artifact = id.artifact();
+    for path in [
+        artifact.path(model, models_path),
+        artifact.partial_path(model, models_path),
+        artifact.verification_path(model, models_path),
+    ] {
+        match fs::remove_file(path) {
+            Ok(()) => {}
+            Err(error) if error.kind() == std::io::ErrorKind::NotFound => {}
+            Err(error) => return Err(error.to_string()),
+        }
+    }
+    Ok(())
+}
+
+#[cfg(test)]
+fn download_artifact(
+    model: ModelChoice,
+    artifact: &Artifact,
+    models_path: &Path,
+) -> Result<DownloadOutcome, String> {
+    download_artifact_with_cancel(
+        model,
+        artifact,
+        models_path,
+        &AtomicBool::new(false),
+        &AtomicU64::new(0),
+    )
+}
+
+fn download_artifact_with_cancel(
+    model: ModelChoice,
+    artifact: &Artifact,
+    models_path: &Path,
+    cancel: &AtomicBool,
+    verified_bytes: &AtomicU64,
+) -> Result<DownloadOutcome, String> {
+    if cancel.load(Ordering::Relaxed) {
+        return Ok(DownloadOutcome::Stopped);
+    }
+    let destination = artifact.path(model, models_path);
+    if artifact.is_installed(model, models_path) {
+        return Ok(DownloadOutcome::Complete);
+    }
+    if destination.exists() {
+        if verify(&destination, artifact, model, cancel, verified_bytes)?
+            == DownloadOutcome::Stopped
+        {
+            return Ok(DownloadOutcome::Stopped);
+        }
+        mark_verified(model, artifact, models_path)?;
+        return Ok(DownloadOutcome::Complete);
+    }
+
+    let directory = destination
+        .parent()
+        .ok_or_else(|| "model artifact path has no parent directory".to_owned())?;
+    fs::create_dir_all(directory).map_err(|error| error.to_string())?;
+    let partial = artifact.partial_path(model, models_path);
+    let partial_size = partial.metadata().map_or(0, |metadata| metadata.len());
+    if partial_size > artifact.size {
+        File::create(&partial).map_err(|error| error.to_string())?;
+    }
+    if partial.metadata().map_or(0, |metadata| metadata.len()) != artifact.size
+        && download_to_partial(artifact, &partial, cancel)? == DownloadOutcome::Stopped
+    {
+        return Ok(DownloadOutcome::Stopped);
+    }
+
+    if cancel.load(Ordering::Relaxed) {
+        return Ok(DownloadOutcome::Stopped);
+    }
+    match verify(&partial, artifact, model, cancel, verified_bytes) {
+        Ok(DownloadOutcome::Stopped) => return Ok(DownloadOutcome::Stopped),
+        Ok(DownloadOutcome::Complete) => {}
+        Err(error) => {
+            fs::remove_file(&partial).map_err(|remove_error| {
+                format!("{error}; could not remove partial file: {remove_error}")
+            })?;
+            return Err(error);
+        }
+    }
+    fs::rename(partial, destination).map_err(|error| error.to_string())?;
+    mark_verified(model, artifact, models_path)?;
+    Ok(DownloadOutcome::Complete)
+}
+
+fn mark_verified(
+    model: ModelChoice,
+    artifact: &Artifact,
+    models_path: &Path,
+) -> Result<(), String> {
+    fs::write(
+        artifact.verification_path(model, models_path),
+        artifact.sha256,
+    )
+    .map_err(|error| error.to_string())
+}
+
+fn verify(
+    path: &Path,
+    artifact: &Artifact,
+    model: ModelChoice,
+    cancel: &AtomicBool,
+    verified_bytes: &AtomicU64,
+) -> Result<DownloadOutcome, String> {
+    verified_bytes.store(0, Ordering::Relaxed);
+    let size = path.metadata().map_err(|error| error.to_string())?.len();
+    if size != artifact.size {
+        return Err(format!(
+            "{} has size {size}, expected {}",
+            path.display(),
+            artifact.size
+        ));
+    }
+
+    let mut file = File::open(path).map_err(|error| error.to_string())?;
+    let mut hasher = Sha256::new();
+    let mut buffer = vec![0; 1024 * 1024];
+    loop {
+        if cancel.load(Ordering::Relaxed) {
+            return Ok(DownloadOutcome::Stopped);
+        }
+        let count = file.read(&mut buffer).map_err(|error| error.to_string())?;
+        if count == 0 {
+            break;
+        }
+        hasher.update(&buffer[..count]);
+        verified_bytes.fetch_add(count as u64, Ordering::Relaxed);
+    }
+    let actual = hex(&hasher.finalize());
+    if actual != artifact.sha256 {
+        return Err(format!(
+            "Checksum verification failed for {}",
+            path.display()
+        ));
+    }
+    if let Some(support) = artifact.support {
+        crate::engine::validate_model_artifact(path, model, support)?;
+    }
+    Ok(DownloadOutcome::Complete)
+}
+
+fn hex(bytes: &[u8]) -> String {
+    const DIGITS: &[u8; 16] = b"0123456789abcdef";
+    let mut result = String::with_capacity(bytes.len() * 2);
+    for byte in bytes {
+        result.push(DIGITS[(byte >> 4) as usize] as char);
+        result.push(DIGITS[(byte & 0x0f) as usize] as char);
+    }
+    result
+}
+
+fn download_to_partial(
+    artifact: &Artifact,
+    partial: &Path,
+    cancel: &AtomicBool,
+) -> Result<DownloadOutcome, String> {
+    download_url_to_partial(&artifact.url(), partial, cancel)
+}
+
+fn download_url_to_partial(
+    url: &str,
+    partial: &Path,
+    cancel: &AtomicBool,
+) -> Result<DownloadOutcome, String> {
+    let offset = partial.metadata().map_or(0, |metadata| metadata.len());
+    let agent: ureq::Agent = ureq::Agent::config_builder()
+        .https_only(url.starts_with("https://"))
+        .build()
+        .into();
+    let mut request = agent.get(url);
+    if offset > 0 {
+        request = request.header("Range", format!("bytes={offset}-"));
+    }
+    let mut response = request
+        .call()
+        .map_err(|error| format!("Model download failed: {error}"))?;
+    let status = response.status().as_u16();
+    let append = offset > 0 && status == 206;
+    if offset > 0 && status != 200 && status != 206 {
+        return Err(format!(
+            "Model server returned HTTP {status} while resuming at byte {offset}"
+        ));
+    }
+    if append {
+        validate_content_range(&response, offset)?;
+    }
+
+    let mut output = OpenOptions::new()
+        .create(true)
+        .write(true)
+        .append(append)
+        .truncate(!append)
+        .open(partial)
+        .map_err(|error| error.to_string())?;
+    let mut body = response.body_mut().as_reader();
+    let mut buffer = vec![0; 1024 * 1024];
+    loop {
+        if cancel.load(Ordering::Relaxed) {
+            return Ok(DownloadOutcome::Stopped);
+        }
+        let count = body
+            .read(&mut buffer)
+            .map_err(|error| format!("Model download failed: {error}"))?;
+        if count == 0 {
+            break;
+        }
+        output
+            .write_all(&buffer[..count])
+            .map_err(|error| error.to_string())?;
+    }
+    Ok(DownloadOutcome::Complete)
+}
+
+fn validate_content_range(
+    response: &ureq::http::Response<ureq::Body>,
+    offset: u64,
+) -> Result<(), String> {
+    let expected = format!("bytes {offset}-");
+    let content_range = response
+        .headers()
+        .get("content-range")
+        .and_then(|value| value.to_str().ok())
+        .unwrap_or_default();
+    if content_range.starts_with(&expected) {
+        Ok(())
+    } else {
+        Err(format!(
+            "Model server returned an invalid Content-Range while resuming at byte {offset}"
+        ))
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::net::TcpListener;
+    use std::thread;
+    use std::time::{SystemTime, UNIX_EPOCH};
+
+    #[test]
+    fn catalog_and_checksum_verification_are_explicit() {
+        assert_eq!(ModelChoice::from_id("glm-5.2"), Some(ModelChoice::Glm52));
+        assert!(ModelChoice::from_id("unknown").is_none());
+        assert_eq!(
+            ModelChoice::DeepSeekV4Flash.main_artifact().size,
+            86_720_111_488
+        );
+        assert_eq!(ModelChoice::Glm52.main_artifact().size, 211_075_856_448);
+        assert_eq!(ModelChoice::DeepSeekV4Flash.artifacts(true).count(), 2);
+        assert_eq!(ModelChoice::Glm52.artifacts(true).count(), 1);
+
+        let id = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap()
+            .as_nanos();
+        let models_path = std::env::temp_dir().join(format!("ds4-server-models-{id}"));
+        let engine = engine_artifacts(ModelChoice::DeepSeekV4Flash, true, &models_path);
+        assert_eq!(
+            engine.model.file_name(),
+            Some(std::ffi::OsStr::new(FLASH.file_name))
+        );
+        assert_eq!(
+            engine.mtp.as_deref().and_then(Path::file_name),
+            Some(std::ffi::OsStr::new(FLASH_DSPARK.file_name))
+        );
+        let empty = Artifact {
+            label: "empty",
+            file_name: "empty",
+            repository: "",
+            size: 0,
+            sha256: "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
+            support: None,
+        };
+        let partial = empty.partial_path(ModelChoice::DeepSeekV4Flash, &models_path);
+        fs::create_dir_all(partial.parent().unwrap()).unwrap();
+        fs::write(&partial, []).unwrap();
+        download_artifact(ModelChoice::DeepSeekV4Flash, &empty, &models_path).unwrap();
+        assert!(empty.is_installed(ModelChoice::DeepSeekV4Flash, &models_path));
+        assert!(!partial.exists());
+        assert_eq!(
+            fs::read_to_string(empty.verification_path(ModelChoice::DeepSeekV4Flash, &models_path))
+                .unwrap(),
+            empty.sha256
+        );
+        fs::remove_dir_all(models_path).unwrap();
+    }
+
+    #[test]
+    fn verification_reports_bytes_read() {
+        let path = std::env::temp_dir().join(format!(
+            "ds4-server-verify-{}",
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .unwrap()
+                .as_nanos()
+        ));
+        fs::write(&path, b"abc").unwrap();
+        let artifact = Artifact {
+            label: "test model",
+            file_name: "unused",
+            repository: "unused",
+            size: 3,
+            sha256: "ba7816bf8f01cfea414140de5dae2223b00361a396177a9cb410ff61f20015ad",
+            support: None,
+        };
+        let verified_bytes = AtomicU64::new(999);
+
+        assert_eq!(
+            verify(
+                &path,
+                &artifact,
+                ModelChoice::DeepSeekV4Flash,
+                &AtomicBool::new(false),
+                &verified_bytes,
+            )
+            .unwrap(),
+            DownloadOutcome::Complete
+        );
+        assert_eq!(verified_bytes.load(Ordering::Relaxed), 3);
+        fs::remove_file(path).unwrap();
+    }
+
+    #[test]
+    fn managed_artifact_inventory_and_delete_include_partial_files() {
+        let models_path = std::env::temp_dir().join(format!(
+            "ds4-server-manager-{}",
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .unwrap()
+                .as_nanos()
+        ));
+        let id = ManagedArtifactId::DeepSeekV4Flash;
+        let partial = id.artifact().partial_path(id.model(), &models_path);
+        fs::create_dir_all(partial.parent().unwrap()).unwrap();
+        fs::write(&partial, b"part").unwrap();
+
+        let managed = managed_artifacts(&models_path)
+            .into_iter()
+            .find(|artifact| artifact.id == id)
+            .unwrap();
+        assert_eq!(managed.stored, 4);
+        assert_eq!(managed.state, ManagedArtifactState::Partial);
+
+        delete_managed_artifact(id, &models_path).unwrap();
+        assert!(!partial.exists());
+        fs::remove_dir_all(models_path).unwrap();
+    }
+
+    #[test]
+    fn restart_resumes_at_the_existing_partial_byte() {
+        let content = b"restart-resume works";
+        let offset = 8;
+        let directory = std::env::temp_dir().join(format!(
+            "ds4-server-resume-{}",
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .unwrap()
+                .as_nanos()
+        ));
+        fs::create_dir_all(&directory).unwrap();
+        let partial = directory.join("model.gguf.part");
+        fs::write(&partial, &content[..offset]).unwrap();
+
+        let listener = match TcpListener::bind("127.0.0.1:0") {
+            Ok(listener) => listener,
+            Err(error) if error.kind() == std::io::ErrorKind::PermissionDenied => {
+                fs::remove_dir_all(directory).unwrap();
+                return;
+            }
+            Err(error) => panic!("could not start test server: {error}"),
+        };
+        let address = listener.local_addr().unwrap();
+        let server = thread::spawn(move || {
+            let (mut connection, _) = listener.accept().unwrap();
+            let mut request = [0; 2048];
+            let count = connection.read(&mut request).unwrap();
+            let request = String::from_utf8_lossy(&request[..count]).to_ascii_lowercase();
+            assert!(request.contains("range: bytes=8-"));
+            let remaining = &content[offset..];
+            write!(
+                connection,
+                "HTTP/1.1 206 Partial Content\r\nContent-Length: {}\r\nContent-Range: bytes {offset}-{}/{}\r\nConnection: close\r\n\r\n",
+                remaining.len(),
+                content.len() - 1,
+                content.len(),
+            )
+            .unwrap();
+            connection.write_all(remaining).unwrap();
+        });
+
+        let outcome = download_url_to_partial(
+            &format!("http://{address}/model.gguf"),
+            &partial,
+            &AtomicBool::new(false),
+        )
+        .unwrap();
+        server.join().unwrap();
+        assert_eq!(outcome, DownloadOutcome::Complete);
+        assert_eq!(fs::read(&partial).unwrap(), content);
+        fs::remove_dir_all(directory).unwrap();
+    }
+
+    #[test]
+    fn cancellation_keeps_the_partial_file_for_the_next_run() {
+        let directory = std::env::temp_dir().join(format!(
+            "ds4-server-cancel-{}",
+            SystemTime::now()
+                .duration_since(UNIX_EPOCH)
+                .unwrap()
+                .as_nanos()
+        ));
+        let artifact = Artifact {
+            label: "test model",
+            file_name: "model.gguf",
+            repository: "unused",
+            size: 10,
+            sha256: "unused",
+            support: None,
+        };
+        let partial = artifact.partial_path(ModelChoice::DeepSeekV4Flash, &directory);
+        fs::create_dir_all(partial.parent().unwrap()).unwrap();
+        fs::write(&partial, b"part").unwrap();
+        let cancel = AtomicBool::new(true);
+
+        assert_eq!(
+            download_artifact_with_cancel(
+                ModelChoice::DeepSeekV4Flash,
+                &artifact,
+                &directory,
+                &cancel,
+                &AtomicU64::new(0),
+            )
+            .unwrap(),
+            DownloadOutcome::Stopped
+        );
+        assert_eq!(fs::read(&partial).unwrap(), b"part");
+        fs::remove_dir_all(directory).unwrap();
+    }
+}

src/server/http.rs

@@ -0,0 +1,178 @@
+use super::*;
+
+pub(super) fn send_sse_headers(stream: &mut impl Write) -> Result<(), String> {
+    stream
+        .write_all(
+            b"HTTP/1.1 200 OK\r\nContent-Type: text/event-stream\r\nCache-Control: no-cache\r\nConnection: close\r\n\r\n",
+        )
+        .map_err(|error| error.to_string())
+}
+
+pub(super) fn send_sse(stream: &mut impl Write, value: &Value) -> Result<(), String> {
+    let body = serde_json::to_vec(value).map_err(|error| error.to_string())?;
+    stream
+        .write_all(b"data: ")
+        .map_err(|error| error.to_string())?;
+    stream.write_all(&body).map_err(|error| error.to_string())?;
+    stream.write_all(b"\n\n").map_err(|error| error.to_string())
+}
+
+pub(super) fn send_sse_error(stream: &mut impl Write, message: &str) -> Result<(), String> {
+    let body = serde_json::to_vec(&json!({"error": {"message": message, "type": "server_error"}}))
+        .map_err(|error| error.to_string())?;
+    stream
+        .write_all(b"event: error\ndata: ")
+        .map_err(|error| error.to_string())?;
+    stream.write_all(&body).map_err(|error| error.to_string())?;
+    stream.write_all(b"\n\n").map_err(|error| error.to_string())
+}
+
+pub(super) fn send_json(stream: &mut TcpStream, code: u16, value: &Value) -> Result<(), String> {
+    let mut body = serde_json::to_vec(value).map_err(|error| error.to_string())?;
+    body.push(b'\n');
+    send_response(stream, code, Some("application/json"), &body)
+}
+
+pub(super) fn send_error(stream: &mut TcpStream, code: u16, message: &str) -> Result<(), String> {
+    send_json(
+        stream,
+        code,
+        &json!({"error": {"message": message, "type": "invalid_request_error"}}),
+    )
+}
+
+pub(super) fn send_response(
+    stream: &mut TcpStream,
+    code: u16,
+    content_type: Option<&str>,
+    body: &[u8],
+) -> Result<(), String> {
+    let reason = match code {
+        200 => "OK",
+        204 => "No Content",
+        400 => "Bad Request",
+        404 => "Not Found",
+        409 => "Conflict",
+        500 => "Internal Server Error",
+        _ => "Error",
+    };
+    let mut header = format!(
+        "HTTP/1.1 {code} {reason}\r\nContent-Length: {}\r\n",
+        body.len()
+    );
+    if let Some(content_type) = content_type {
+        header.push_str("Content-Type: ");
+        header.push_str(content_type);
+        header.push_str("\r\n");
+    }
+    header.push_str("Connection: close\r\n\r\n");
+    stream
+        .write_all(header.as_bytes())
+        .and_then(|()| stream.write_all(body))
+        .map_err(|error| error.to_string())
+}
+
+pub(super) fn read_request(stream: &mut TcpStream) -> Result<HttpRequest, String> {
+    let mut bytes = Vec::new();
+    let header_end = loop {
+        if bytes.len() >= MAX_HEADER_BYTES {
+            return Err("HTTP headers are too large".into());
+        }
+        let mut chunk = [0_u8; 4096];
+        let read = stream.read(&mut chunk).map_err(|error| error.to_string())?;
+        if read == 0 {
+            return Err("bad HTTP request".into());
+        }
+        bytes.extend_from_slice(&chunk[..read]);
+        if let Some(end) = find_header_end(&bytes) {
+            break end;
+        }
+    };
+    let header = std::str::from_utf8(&bytes[..header_end])
+        .map_err(|_| "HTTP headers are not UTF-8".to_owned())?;
+    let mut lines = header.lines();
+    let request_line = lines.next().ok_or_else(|| "bad HTTP request".to_owned())?;
+    let mut parts = request_line.split_whitespace();
+    let method = parts
+        .next()
+        .ok_or_else(|| "bad HTTP request".to_owned())?
+        .to_owned();
+    let path = parts
+        .next()
+        .ok_or_else(|| "bad HTTP request".to_owned())?
+        .to_owned();
+    let length = lines
+        .find_map(|line| {
+            line.split_once(':').and_then(|(name, value)| {
+                name.eq_ignore_ascii_case("content-length")
+                    .then(|| value.trim().parse::<usize>().ok())
+                    .flatten()
+            })
+        })
+        .unwrap_or(0);
+    if length > MAX_BODY_BYTES {
+        return Err("HTTP body is too large".into());
+    }
+    while bytes.len() < header_end + length {
+        let mut chunk = [0_u8; 8192];
+        let read = stream.read(&mut chunk).map_err(|error| error.to_string())?;
+        if read == 0 {
+            return Err("incomplete HTTP body".into());
+        }
+        bytes.extend_from_slice(&chunk[..read]);
+    }
+    Ok(HttpRequest {
+        method,
+        path: path.split('?').next().unwrap_or(&path).to_owned(),
+        body: bytes[header_end..header_end + length].to_vec(),
+    })
+}
+
+fn find_header_end(bytes: &[u8]) -> Option<usize> {
+    bytes
+        .windows(4)
+        .position(|window| window == b"\r\n\r\n")
+        .map(|position| position + 4)
+        .or_else(|| {
+            bytes
+                .windows(2)
+                .position(|window| window == b"\n\n")
+                .map(|position| position + 2)
+        })
+}
+
+pub(super) fn unix_time() -> u64 {
+    SystemTime::now()
+        .duration_since(UNIX_EPOCH)
+        .unwrap_or_default()
+        .as_secs()
+}
+
+pub(super) fn random_id(prefix: &str) -> String {
+    random_hex_id::<12>(prefix)
+}
+
+pub(super) fn random_tool_id(prefix: &str) -> String {
+    random_hex_id::<16>(prefix)
+}
+
+fn random_hex_id<const N: usize>(prefix: &str) -> String {
+    let mut bytes = [0_u8; N];
+    if File::open("/dev/urandom")
+        .and_then(|mut file| file.read_exact(&mut bytes))
+        .is_err()
+    {
+        let fallback = SystemTime::now()
+            .duration_since(UNIX_EPOCH)
+            .unwrap_or_default()
+            .as_nanos();
+        bytes.copy_from_slice(&fallback.to_le_bytes()[..N]);
+    }
+    let mut id = String::with_capacity(prefix.len() + N * 2);
+    id.push_str(prefix);
+    for byte in bytes {
+        use std::fmt::Write as _;
+        let _ = write!(id, "{byte:02x}");
+    }
+    id
+}

src/server/request.rs

@@ -0,0 +1,595 @@
+use super::*;
+
+pub(super) fn completion_request(mut value: Value) -> Result<ChatRequest, (u16, String)> {
+    let object = value
+        .as_object_mut()
+        .ok_or_else(|| (400, "invalid JSON request".to_owned()))?;
+    let prompt = object
+        .remove("prompt")
+        .ok_or_else(|| (400, "missing prompt".to_owned()))?;
+    let prompt = match prompt {
+        Value::String(text) => text,
+        Value::Array(values) => values
+            .into_iter()
+            .next()
+            .and_then(|value| value.as_str().map(str::to_owned))
+            .unwrap_or_default(),
+        _ => String::new(),
+    };
+    object.insert(
+        "messages".into(),
+        json!([
+            {"role": "system", "content": "You are a helpful assistant"},
+            {"role": "user", "content": prompt}
+        ]),
+    );
+    decode_chat_request(value)
+}
+
+pub(super) fn anthropic_request(value: Value) -> Result<ChatRequest, (u16, String)> {
+    let object = value
+        .as_object()
+        .ok_or_else(|| (400, "invalid JSON request".to_owned()))?;
+    let source = object
+        .get("messages")
+        .and_then(Value::as_array)
+        .ok_or_else(|| (400, "missing messages".to_owned()))?;
+    let mut messages = Vec::new();
+    let system = object.get("system").map(content_text).unwrap_or_default();
+    if !system.is_empty() {
+        messages.push(json!({"role": "system", "content": system}));
+    }
+    for message in source {
+        let role = message
+            .get("role")
+            .and_then(Value::as_str)
+            .unwrap_or("user");
+        let Some(blocks) = message.get("content").and_then(Value::as_array) else {
+            messages.push(json!({"role": role, "content": message.get("content").cloned().unwrap_or(Value::Null)}));
+            continue;
+        };
+        let mut text = String::new();
+        let mut reasoning = String::new();
+        let mut calls = Vec::new();
+        for block in blocks {
+            match block.get("type").and_then(Value::as_str).unwrap_or("text") {
+                "text" => text.push_str(block.get("text").and_then(Value::as_str).unwrap_or("")),
+                "thinking" | "redacted_thinking" => reasoning.push_str(
+                    block
+                        .get("thinking")
+                        .or_else(|| block.get("data"))
+                        .and_then(Value::as_str)
+                        .unwrap_or(""),
+                ),
+                "tool_use" => calls.push(json!({
+                    "id": block.get("id").and_then(Value::as_str).unwrap_or(""),
+                    "type": "function",
+                    "function": {
+                        "name": block.get("name").and_then(Value::as_str).unwrap_or(""),
+                        "arguments": block.get("input").cloned().unwrap_or_else(|| json!({})).to_string()
+                    }
+                })),
+                "tool_result" => {
+                    if !text.is_empty() {
+                        messages.push(json!({"role": role, "content": std::mem::take(&mut text)}));
+                    }
+                    messages.push(json!({
+                        "role": "tool",
+                        "content": block.get("content").map(content_text).unwrap_or_default(),
+                        "tool_call_id": block.get("tool_use_id").and_then(Value::as_str).unwrap_or("")
+                    }));
+                }
+                _ => {}
+            }
+        }
+        if !text.is_empty() || !reasoning.is_empty() || !calls.is_empty() {
+            messages.push(json!({
+                "role": role,
+                "content": text,
+                "reasoning_content": reasoning,
+                "tool_calls": calls
+            }));
+        }
+    }
+    let mut request = Map::new();
+    copy_request_fields(object, &mut request);
+    request.insert("messages".into(), Value::Array(messages));
+    if let Some(stops) = object.get("stop_sequences") {
+        request.insert("stop".into(), stops.clone());
+    }
+    if let Some(effort) = object
+        .get("output_config")
+        .and_then(|value| value.get("effort"))
+    {
+        request.insert("reasoning_effort".into(), effort.clone());
+    }
+    request.insert(
+        "tools".into(),
+        normalize_anthropic_tools(object.get("tools")),
+    );
+    decode_chat_request(Value::Object(request))
+}
+
+pub(super) fn responses_request(value: Value) -> Result<ChatRequest, (u16, String)> {
+    let object = value
+        .as_object()
+        .ok_or_else(|| (400, "invalid JSON request".to_owned()))?;
+    for key in ["previous_response_id", "conversation"] {
+        if object.get(key).is_some_and(|value| !value.is_null()) {
+            return Err((
+                400,
+                format!("{key} is not supported; replay full input instead"),
+            ));
+        }
+    }
+    if let Some(choice) = object.get("tool_choice") {
+        match choice {
+            Value::String(choice) if choice == "none" || choice == "auto" => {}
+            Value::String(choice) => {
+                return Err((400, format!("tool_choice={choice} not supported")));
+            }
+            Value::Object(_) => return Err((400, "forced tool_choice not supported".into())),
+            _ => {}
+        }
+    }
+    let input = object
+        .get("input")
+        .ok_or_else(|| (400, "missing input".to_owned()))?;
+    let mut messages = Vec::new();
+    match input {
+        Value::String(text) => messages.push(json!({"role": "user", "content": text})),
+        Value::Array(items) => {
+            let mut pending_reasoning = String::new();
+            for item in items {
+                let item = item
+                    .as_object()
+                    .ok_or_else(|| (400, "invalid JSON request".to_owned()))?;
+                if item
+                    .get("status")
+                    .and_then(Value::as_str)
+                    .is_some_and(|status| status != "completed")
+                {
+                    return Err((400, "invalid JSON request".into()));
+                }
+                let kind = item
+                    .get("type")
+                    .and_then(Value::as_str)
+                    .unwrap_or("message");
+                let role = item.get("role").and_then(Value::as_str).unwrap_or("user");
+                let consumes_reasoning = (kind == "message" && role == "assistant")
+                    || matches!(
+                        kind,
+                        "function_call"
+                            | "custom_tool_call"
+                            | "local_shell_call"
+                            | "web_search_call"
+                            | "tool_search_call"
+                            | "image_generation_call"
+                    );
+                let bookkeeping = matches!(kind, "compaction" | "context_compaction");
+                if !consumes_reasoning && !bookkeeping && !pending_reasoning.is_empty() {
+                    messages.push(reasoning_message(std::mem::take(&mut pending_reasoning)));
+                }
+                match kind {
+                    "message" => {
+                        let mut message = json!({
+                            "role": role,
+                            "content": responses_content_text(item.get("content").unwrap_or(&Value::Null))?
+                        });
+                        if role == "assistant" && !pending_reasoning.is_empty() {
+                            message["reasoning_content"] =
+                                Value::String(std::mem::take(&mut pending_reasoning));
+                        }
+                        messages.push(message);
+                    }
+                    "function_call" | "custom_tool_call" => {
+                        let name = if kind == "function_call" {
+                            format!(
+                                "{}{}",
+                                item.get("namespace").and_then(Value::as_str).unwrap_or(""),
+                                item.get("name").and_then(Value::as_str).unwrap_or("")
+                            )
+                        } else {
+                            item.get("name")
+                                .and_then(Value::as_str)
+                                .unwrap_or("")
+                                .to_owned()
+                        };
+                        push_responses_call(
+                            &mut messages,
+                            &mut pending_reasoning,
+                            item,
+                            &name,
+                            item.get("arguments").or_else(|| item.get("input")),
+                        );
+                    }
+                    "function_call_output" | "custom_tool_call_output" => messages.push(json!({
+                        "role": "tool",
+                        "content": responses_output_text(item.get("output"))?,
+                        "tool_call_id": item.get("call_id").or_else(|| item.get("id")).and_then(Value::as_str).unwrap_or("")
+                    })),
+                    "reasoning" => {
+                        for value in [item.get("summary"), item.get("content")]
+                            .into_iter()
+                            .flatten()
+                        {
+                            let text = responses_content_text(value)?;
+                            if !pending_reasoning.is_empty() && !text.is_empty() {
+                                pending_reasoning.push('\n');
+                            }
+                            pending_reasoning.push_str(&text);
+                        }
+                    }
+                    "local_shell_call"
+                    | "web_search_call"
+                    | "tool_search_call"
+                    | "image_generation_call" => {
+                        let name = match kind {
+                            "local_shell_call" => "local_shell",
+                            "tool_search_call" => "tool_search",
+                            other => other,
+                        };
+                        push_responses_call(
+                            &mut messages,
+                            &mut pending_reasoning,
+                            item,
+                            name,
+                            item.get("action")
+                                .or_else(|| item.get("arguments"))
+                                .or_else(|| item.get("input")),
+                        );
+                    }
+                    "local_shell_call_output"
+                    | "web_search_call_output"
+                    | "tool_search_output"
+                    | "tool_search_call_output"
+                    | "image_generation_call_output" => {
+                        let content = if let Some(value) =
+                            item.get("output").or_else(|| item.get("result"))
+                        {
+                            responses_output_text(Some(value))?
+                        } else {
+                            item.get("tools").map(Value::to_string).unwrap_or_default()
+                        };
+                        messages.push(json!({
+                            "role": "tool",
+                            "content": content,
+                            "tool_call_id": item.get("call_id").or_else(|| item.get("id")).and_then(Value::as_str).unwrap_or("")
+                        }));
+                    }
+                    "compaction" | "context_compaction" => {}
+                    _ => return Err((400, "invalid JSON request".into())),
+                }
+            }
+            if !pending_reasoning.is_empty() {
+                messages.push(reasoning_message(pending_reasoning));
+            }
+        }
+        _ => return Err((400, "invalid JSON request".into())),
+    }
+    if let Some(instructions) = object.get("instructions") {
+        let instructions = match instructions {
+            Value::Null => String::new(),
+            Value::String(text) => text.clone(),
+            _ => return Err((400, "invalid JSON request".into())),
+        };
+        if !instructions.is_empty() {
+            messages.insert(0, json!({"role": "system", "content": instructions}));
+        }
+    }
+    let mut request = Map::new();
+    copy_request_fields(object, &mut request);
+    request.insert("messages".into(), Value::Array(messages));
+    if let Some(tokens) = object.get("max_output_tokens") {
+        request.insert("max_tokens".into(), tokens.clone());
+    }
+    if let Some(effort) = object
+        .get("reasoning")
+        .and_then(|value| value.get("effort"))
+    {
+        request.insert("reasoning_effort".into(), effort.clone());
+    }
+    request.insert(
+        "tools".into(),
+        normalize_responses_tools(object.get("tools")),
+    );
+    decode_chat_request(Value::Object(request))
+}
+
+fn reasoning_message(reasoning: String) -> Value {
+    json!({"role": "assistant", "content": "", "reasoning_content": reasoning})
+}
+
+fn push_responses_call(
+    messages: &mut Vec<Value>,
+    pending_reasoning: &mut String,
+    item: &Map<String, Value>,
+    name: &str,
+    arguments: Option<&Value>,
+) {
+    let call = json!({
+        "id": item.get("call_id").or_else(|| item.get("id")).and_then(Value::as_str).unwrap_or(""),
+        "type": "function",
+        "function": {"name": name, "arguments": raw_json_text(arguments)}
+    });
+    if let Some(last) = messages.last_mut()
+        && last.get("role").and_then(Value::as_str) == Some("assistant")
+    {
+        if !pending_reasoning.is_empty()
+            && last
+                .get("reasoning_content")
+                .and_then(Value::as_str)
+                .is_none_or(str::is_empty)
+        {
+            last["reasoning_content"] = Value::String(std::mem::take(pending_reasoning));
+        }
+        if !last.get("tool_calls").is_some_and(Value::is_array) {
+            last["tool_calls"] = json!([]);
+        }
+        last["tool_calls"].as_array_mut().unwrap().push(call);
+    } else {
+        messages.push(json!({
+            "role": "assistant",
+            "content": "",
+            "reasoning_content": std::mem::take(pending_reasoning),
+            "tool_calls": [call]
+        }));
+    }
+}
+
+fn raw_json_text(value: Option<&Value>) -> String {
+    match value {
+        Some(Value::String(text)) => text.clone(),
+        Some(value) => value.to_string(),
+        None => "{}".into(),
+    }
+}
+
+fn responses_output_text(value: Option<&Value>) -> Result<String, (u16, String)> {
+    match value {
+        Some(Value::Array(_)) => responses_content_text(value.unwrap()),
+        Some(Value::String(text)) => Ok(text.clone()),
+        Some(value) => Ok(value.to_string()),
+        None => Ok(String::new()),
+    }
+}
+
+fn responses_content_text(value: &Value) -> Result<String, (u16, String)> {
+    match value {
+        Value::String(text) => Ok(text.clone()),
+        Value::Null => Ok(String::new()),
+        Value::Array(parts) => {
+            let mut text = String::new();
+            for part in parts {
+                match part {
+                    Value::String(part) => text.push_str(part),
+                    Value::Object(part)
+                        if matches!(
+                            part.get("type").and_then(Value::as_str),
+                            Some(
+                                "input_text"
+                                    | "output_text"
+                                    | "text"
+                                    | "summary_text"
+                                    | "reasoning_text"
+                            )
+                        ) && matches!(
+                            part.get("text"),
+                            Some(Value::String(_) | Value::Null)
+                        ) =>
+                    {
+                        if let Some(value) = part.get("text").and_then(Value::as_str) {
+                            text.push_str(value);
+                        }
+                    }
+                    _ => return Err((400, "invalid JSON request".into())),
+                }
+            }
+            Ok(text)
+        }
+        _ => Err((400, "invalid JSON request".into())),
+    }
+}
+
+fn copy_request_fields(source: &Map<String, Value>, target: &mut Map<String, Value>) {
+    for key in [
+        "model",
+        "max_tokens",
+        "max_completion_tokens",
+        "temperature",
+        "top_p",
+        "min_p",
+        "top_k",
+        "seed",
+        "stream",
+        "stream_options",
+        "thinking",
+        "think",
+        "reasoning_effort",
+        "tool_choice",
+        "stop",
+    ] {
+        if let Some(value) = source.get(key) {
+            target.insert(key.into(), value.clone());
+        }
+    }
+}
+
+fn normalize_anthropic_tools(tools: Option<&Value>) -> Value {
+    Value::Array(
+        tools
+            .and_then(Value::as_array)
+            .into_iter()
+            .flatten()
+            .map(|tool| {
+                json!({"type": "function", "function": {
+                    "name": tool.get("name").and_then(Value::as_str).unwrap_or(""),
+                    "description": tool.get("description").and_then(Value::as_str).unwrap_or(""),
+                    "parameters": tool.get("input_schema").cloned().unwrap_or_else(|| json!({"type": "object"}))
+                }})
+            })
+            .collect(),
+    )
+}
+
+fn normalize_responses_tools(tools: Option<&Value>) -> Value {
+    Value::Array(
+        tools
+            .and_then(Value::as_array)
+            .into_iter()
+            .flatten()
+            .filter(|tool| tool.get("type").and_then(Value::as_str) == Some("function"))
+            .map(|tool| {
+                json!({"type": "function", "function": {
+                    "name": tool.get("name").and_then(Value::as_str).unwrap_or(""),
+                    "description": tool.get("description").and_then(Value::as_str).unwrap_or(""),
+                    "parameters": tool.get("parameters").cloned().unwrap_or_else(|| json!({"type": "object"}))
+                }})
+            })
+            .collect(),
+    )
+}
+
+pub(super) fn decode_chat_request(value: Value) -> Result<ChatRequest, (u16, String)> {
+    serde_json::from_value(value).map_err(|_| (400, "invalid JSON request".to_owned()))
+}
+
+pub(super) fn raw_tool_schemas(
+    body: &[u8],
+    protocol: Protocol,
+) -> Result<Vec<String>, (u16, String)> {
+    let request: RawToolsRequest<'_> =
+        serde_json::from_slice(body).map_err(|_| (400, "invalid JSON request".to_owned()))?;
+    request
+        .tools
+        .into_iter()
+        .map(|tool| {
+            if protocol == Protocol::Chat {
+                let wrapper: RawOpenAiTool<'_> = serde_json::from_str(tool.get())
+                    .map_err(|_| (400, "invalid JSON request".to_owned()))?;
+                Ok(wrapper.function.unwrap_or(tool).get().to_owned())
+            } else {
+                Ok(tool.get().to_owned())
+            }
+        })
+        .collect()
+}
+
+pub(super) fn parse_chat_request(
+    state: &State,
+    request: ChatRequest,
+    protocol: Protocol,
+) -> Result<ParsedRequest, (u16, String)> {
+    if request.messages.is_empty() {
+        return Err((400, "missing messages".into()));
+    }
+    let requested_id = request.model.as_deref().unwrap_or_default();
+    let preferences = state
+        .preferences
+        .read()
+        .map_err(|_| (500, "preferences are unavailable".to_owned()))?
+        .clone();
+    let model = if requested_id.is_empty() {
+        ModelChoice::from_id(&preferences.selected_model)
+    } else {
+        model_alias(requested_id)
+    }
+    .ok_or_else(|| (400, format!("unknown model: {requested_id}")))?;
+    if !installed_endpoint_models(&state.models_path).contains(&model) {
+        return Err((400, format!("model is not installed and verified: {model}")));
+    }
+
+    let mut generation = preferences.generation().map_err(|error| (500, error))?;
+    generation.system_prompt.clear();
+    generation.max_generated_tokens = request
+        .max_completion_tokens
+        .or(request.max_tokens)
+        .unwrap_or(generation.max_generated_tokens);
+    if generation.max_generated_tokens <= 0 {
+        return Err((400, "max_tokens must be positive".into()));
+    }
+    generation.temperature = request.temperature.or(Some(1.0));
+    generation.top_p = request.top_p.or(Some(1.0));
+    generation.min_p = request.min_p.or(Some(0.05));
+    generation.seed = request.seed.filter(|seed| *seed > 0);
+    generation.reasoning_mode = request_reasoning(&request, requested_id)?;
+
+    let tools_enabled = (!request.tools.is_empty() || !request.tool_schemas.is_empty())
+        && request.tool_choice.as_ref().and_then(Value::as_str) != Some("none");
+    let (system, messages) = render_messages(
+        state,
+        &request.messages,
+        &request.tools,
+        &request.tool_schemas,
+        tools_enabled,
+        protocol,
+    )?;
+    generation.system_prompt = system;
+    let runtime = preferences.runtime().map_err(|error| (500, error))?;
+    let mut effective = effective_settings(model, &generation, &runtime, &state.models_path)
+        .map_err(|error| (400, error))?;
+    effective.turn.top_k = request.top_k.unwrap_or(0);
+    if effective.turn.top_k < 0 {
+        return Err((400, "top_k must not be negative".into()));
+    }
+    effective.turn.stops = match request.stop {
+        Some(OneOrMany::One(stop)) => vec![stop],
+        Some(OneOrMany::Many(stops)) => stops,
+        None => Vec::new(),
+    };
+    effective.turn.stops.retain(|stop| !stop.is_empty());
+
+    Ok(ParsedRequest {
+        protocol,
+        model_id: if requested_id.is_empty() {
+            model.id().to_owned()
+        } else {
+            requested_id.to_owned()
+        },
+        messages,
+        turn: effective.turn,
+        engine: effective.engine,
+        idle_timeout: Duration::from_secs(preferences.idle_timeout_minutes.max(1) as u64 * 60),
+        stream: request.stream,
+        include_usage: request
+            .stream_options
+            .is_some_and(|options| options.include_usage),
+        has_tools: tools_enabled,
+    })
+}
+
+fn request_reasoning(
+    request: &ChatRequest,
+    model_id: &str,
+) -> Result<ReasoningMode, (u16, String)> {
+    let explicit_thinking = request
+        .think
+        .or_else(|| request.thinking.as_ref().and_then(thinking_enabled));
+    let mut reasoning = match request.reasoning_effort.as_deref() {
+        Some("max") => ReasoningMode::Max,
+        Some("none") => ReasoningMode::Direct,
+        Some("xhigh" | "high" | "medium" | "low" | "minimal") | None => ReasoningMode::High,
+        Some(value) => return Err((400, format!("unsupported reasoning_effort: {value}"))),
+    };
+    if explicit_thinking == Some(false)
+        || (explicit_thinking.is_none()
+            && matches!(
+                model_id,
+                "deepseek-chat" | "glm-5.2-chat" | "glm-5.2-no-think" | "glm-5.2-nothink"
+            ))
+    {
+        reasoning = ReasoningMode::Direct;
+    }
+    Ok(reasoning)
+}
+
+fn thinking_enabled(value: &Value) -> Option<bool> {
+    value.as_bool().or_else(|| {
+        value.as_str().map(|value| value != "disabled").or_else(|| {
+            value
+                .get("type")
+                .and_then(Value::as_str)
+                .map(|value| value != "disabled")
+        })
+    })
+}

src/server/response.rs

@@ -0,0 +1,974 @@
+use super::*;
+
+pub(super) fn final_response(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    let output = wait_for_output(active)?;
+    let (content, calls) = parse_generated_tools(state, &output.message.content, request.protocol);
+    let finish = if calls.is_empty() {
+        output.finish_reason
+    } else {
+        "tool_calls"
+    };
+    let reasoning = output.message.reasoning.filter(|value| !value.is_empty());
+    let usage = usage_json(
+        output.prompt_tokens,
+        output.cached_tokens,
+        output.completion_tokens,
+    );
+    let body = match request.protocol {
+        Protocol::Chat => {
+            let mut message = json!({"role": "assistant", "content": content});
+            if let Some(reasoning) = reasoning {
+                message["reasoning_content"] = Value::String(reasoning);
+            }
+            if !calls.is_empty() {
+                message["tool_calls"] = tool_calls_json(&calls);
+            }
+            json!({
+                "id": id, "object": "chat.completion", "created": unix_time(),
+                "model": request.model_id,
+                "choices": [{"index": 0, "message": message, "finish_reason": finish}],
+                "usage": usage,
+            })
+        }
+        Protocol::Completion => json!({
+            "id": id, "object": "text_completion", "created": unix_time(),
+            "model": request.model_id,
+            "choices": [{"text": content, "index": 0, "finish_reason": finish}],
+            "usage": usage,
+        }),
+        Protocol::Anthropic => {
+            let mut blocks = Vec::new();
+            if let Some(reasoning) = reasoning {
+                blocks.push(json!({"type": "thinking", "thinking": reasoning, "signature": id}));
+            }
+            if !content.is_empty() {
+                blocks.push(json!({"type": "text", "text": content}));
+            }
+            for call in &calls {
+                blocks.push(json!({
+                    "type": "tool_use", "id": call.id, "name": call.function.name,
+                    "input": serde_json::from_str::<Value>(&call.function.arguments).unwrap_or_else(|_| json!({}))
+                }));
+            }
+            if blocks.is_empty() || (blocks.iter().all(|block| block["type"] == "thinking")) {
+                blocks.push(json!({"type": "text", "text": ""}));
+            }
+            let cached = output.cached_tokens.min(output.prompt_tokens);
+            let written = output.prompt_tokens - cached;
+            json!({
+                "id": id, "type": "message", "role": "assistant", "model": request.model_id,
+                "content": blocks,
+                "stop_reason": if finish == "tool_calls" { "tool_use" } else if finish == "length" { "max_tokens" } else { "end_turn" },
+                "stop_sequence": Value::Null,
+                "usage": {
+                    "input_tokens": output.prompt_tokens - cached - written,
+                    "output_tokens": output.completion_tokens,
+                    "cache_read_input_tokens": cached,
+                    "cache_creation_input_tokens": written
+                }
+            })
+        }
+        Protocol::Responses => {
+            let status = if finish == "length" {
+                "incomplete"
+            } else if finish == "error" {
+                "failed"
+            } else {
+                "completed"
+            };
+            let mut items = Vec::new();
+            if let Some(reasoning) = reasoning {
+                items.push(json!({
+                    "id": random_id("rs_"), "type": "reasoning", "status": status,
+                    "summary": [{"type": "summary_text", "text": reasoning}]
+                }));
+            }
+            if !content.is_empty() {
+                items.push(json!({
+                    "id": random_id("msg_"), "type": "message", "status": status,
+                    "role": "assistant", "content": [{"type": "output_text", "text": content, "annotations": []}]
+                }));
+            }
+            for call in &calls {
+                items.push(json!({
+                    "id": random_id("fc_"), "type": "function_call", "status": status,
+                    "name": call.function.name, "call_id": call.id, "arguments": call.function.arguments
+                }));
+            }
+            json!({
+                "id": id, "object": "response", "created_at": unix_time(), "status": status,
+                "model": request.model_id, "output": items,
+                "usage": {
+                    "input_tokens": output.prompt_tokens,
+                    "input_tokens_details": {"cached_tokens": output.cached_tokens.min(output.prompt_tokens), "cache_write_tokens": output.prompt_tokens - output.cached_tokens.min(output.prompt_tokens)},
+                    "output_tokens": output.completion_tokens,
+                    "output_tokens_details": {"reasoning_tokens": 0},
+                    "total_tokens": output.prompt_tokens + output.completion_tokens
+                }
+            })
+        }
+    };
+    send_json(stream, 200, &body).map_err(|error| (500, error))
+}
+
+pub(super) fn stream_response(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    if request.protocol != Protocol::Chat {
+        return structured_stream_response(stream, state, request, active, id);
+    }
+    stream_response_with_keepalive(
+        stream,
+        state,
+        request,
+        active,
+        id,
+        PREFILL_KEEPALIVE_INTERVAL,
+    )
+}
+
+fn structured_stream_response(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    match request.protocol {
+        Protocol::Completion => completion_stream_response(stream, request, active, id),
+        Protocol::Anthropic => anthropic_stream_response(stream, state, request, active, id),
+        Protocol::Responses => responses_stream_response(stream, state, request, active, id),
+        Protocol::Chat => unreachable!(),
+    }
+}
+
+fn completion_stream_response(
+    stream: &mut TcpStream,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    send_sse_headers(stream).map_err(|error| (500, error))?;
+    while let Ok(event) = active.events.recv() {
+        match event {
+            GenerationEvent::Chunk {
+                reasoning: false,
+                content,
+            } if !content.is_empty() => send_sse(
+                stream,
+                &json!({
+                    "id": id, "object": "text_completion", "created": unix_time(),
+                    "model": request.model_id,
+                    "choices": [{"text": content, "index": 0, "finish_reason": Value::Null}]
+                }),
+            )
+            .map_err(|error| (500, error))?,
+            GenerationEvent::Finished(Ok(output)) => {
+                send_sse(
+                    stream,
+                    &json!({
+                        "id": id, "object": "text_completion", "created": unix_time(),
+                        "model": request.model_id,
+                        "choices": [{"text": "", "index": 0, "finish_reason": output.finish_reason}]
+                    }),
+                )
+                .map_err(|error| (500, error))?;
+                if request.include_usage {
+                    send_sse(
+                        stream,
+                        &json!({
+                            "id": id, "object": "text_completion", "created": unix_time(),
+                            "model": request.model_id, "choices": [],
+                            "usage": usage_json(output.prompt_tokens, output.cached_tokens, output.completion_tokens)
+                        }),
+                    )
+                    .map_err(|error| (500, error))?;
+                }
+                return stream
+                    .write_all(b"data: [DONE]\n\n")
+                    .map_err(|error| (500, error.to_string()));
+            }
+            GenerationEvent::Finished(Err(error)) => {
+                let _ = send_sse_error(stream, &error);
+                return Ok(());
+            }
+            _ => {}
+        }
+    }
+    Err((500, "The model runtime stopped unexpectedly.".into()))
+}
+
+fn anthropic_stream_response(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    send_sse_headers(stream).map_err(|error| (500, error))?;
+    let mut prompt_tokens = 0;
+    let mut started = false;
+    let mut block = None::<(usize, bool)>;
+    let mut next_index = 0;
+    let mut projector = ToolProjector::new();
+    let mut tool_indices = Vec::new();
+    while let Ok(event) = active.events.recv() {
+        match event {
+            GenerationEvent::Context {
+                used,
+                tokens_per_second,
+                ..
+            } => {
+                if tokens_per_second.is_none() {
+                    prompt_tokens = used;
+                } else if !started {
+                    anthropic_stream_start(stream, &request, id, prompt_tokens, 0)?;
+                    started = true;
+                }
+            }
+            GenerationEvent::Chunk { reasoning, content } => {
+                if !started {
+                    anthropic_stream_start(stream, &request, id, prompt_tokens, 0)?;
+                    started = true;
+                }
+                if !reasoning && request.has_tools {
+                    let events = projector.push(&content, false, "toolu_");
+                    send_anthropic_projection_events(
+                        stream,
+                        events,
+                        &mut block,
+                        &mut next_index,
+                        &mut tool_indices,
+                    )?;
+                    continue;
+                }
+                if block.is_some_and(|(_, current_reasoning)| current_reasoning != reasoning) {
+                    let (index, _) = block.take().unwrap();
+                    send_named_sse(
+                        stream,
+                        "content_block_stop",
+                        &json!({"type": "content_block_stop", "index": index}),
+                    )?;
+                }
+                let index = if let Some((index, _)) = block {
+                    index
+                } else {
+                    let index = next_index;
+                    next_index += 1;
+                    send_named_sse(
+                        stream,
+                        "content_block_start",
+                        &json!({
+                            "type": "content_block_start", "index": index,
+                            "content_block": if reasoning { json!({"type": "thinking", "thinking": "", "signature": ""}) } else { json!({"type": "text", "text": ""}) }
+                        }),
+                    )?;
+                    block = Some((index, reasoning));
+                    index
+                };
+                let delta = if reasoning {
+                    json!({"type": "thinking_delta", "thinking": content})
+                } else {
+                    json!({"type": "text_delta", "text": content})
+                };
+                send_named_sse(
+                    stream,
+                    "content_block_delta",
+                    &json!({"type": "content_block_delta", "index": index, "delta": delta}),
+                )?;
+            }
+            GenerationEvent::Finished(Ok(output)) => {
+                if !started {
+                    anthropic_stream_start(
+                        stream,
+                        &request,
+                        id,
+                        output.prompt_tokens,
+                        output.cached_tokens,
+                    )?;
+                }
+                if request.has_tools {
+                    let events = projector.push("", true, "toolu_");
+                    send_anthropic_projection_events(
+                        stream,
+                        events,
+                        &mut block,
+                        &mut next_index,
+                        &mut tool_indices,
+                    )?;
+                }
+                if let Some((index, _)) = block.take() {
+                    send_named_sse(
+                        stream,
+                        "content_block_stop",
+                        &json!({"type": "content_block_stop", "index": index}),
+                    )?;
+                }
+                let (_, calls) = parse_generated_tools_with_ids(
+                    state,
+                    &output.message.content,
+                    Protocol::Anthropic,
+                    &projector.ids,
+                );
+                if projector.ids.is_empty() {
+                    for call in &calls {
+                        send_named_sse(
+                            stream,
+                            "content_block_start",
+                            &json!({"type": "content_block_start", "index": next_index, "content_block": {"type": "tool_use", "id": call.id, "name": call.function.name, "input": {}}}),
+                        )?;
+                        send_named_sse(
+                            stream,
+                            "content_block_delta",
+                            &json!({"type": "content_block_delta", "index": next_index, "delta": {"type": "input_json_delta", "partial_json": call.function.arguments}}),
+                        )?;
+                        send_named_sse(
+                            stream,
+                            "content_block_stop",
+                            &json!({"type": "content_block_stop", "index": next_index}),
+                        )?;
+                        next_index += 1;
+                    }
+                }
+                let finish = if calls.is_empty() {
+                    output.finish_reason
+                } else {
+                    "tool_calls"
+                };
+                send_named_sse(
+                    stream,
+                    "message_delta",
+                    &json!({"type": "message_delta", "delta": {"stop_reason": if finish == "tool_calls" { "tool_use" } else if finish == "length" { "max_tokens" } else { "end_turn" }, "stop_sequence": Value::Null}, "usage": {"output_tokens": output.completion_tokens}}),
+                )?;
+                return send_named_sse(stream, "message_stop", &json!({"type": "message_stop"}));
+            }
+            GenerationEvent::Finished(Err(error)) => {
+                return send_named_sse(
+                    stream,
+                    "error",
+                    &json!({"type": "error", "error": {"type": "api_error", "message": error}}),
+                );
+            }
+            _ => {}
+        }
+    }
+    Err((500, "The model runtime stopped unexpectedly.".into()))
+}
+
+fn send_anthropic_projection_events(
+    stream: &mut impl Write,
+    events: Vec<ToolProjectionEvent>,
+    block: &mut Option<(usize, bool)>,
+    next_index: &mut usize,
+    tool_indices: &mut Vec<usize>,
+) -> Result<(), (u16, String)> {
+    for event in events {
+        match event {
+            ToolProjectionEvent::Text(text) if !text.is_empty() => {
+                if block.is_some_and(|(_, reasoning)| reasoning) {
+                    let (index, _) = block.take().unwrap();
+                    send_named_sse(
+                        stream,
+                        "content_block_stop",
+                        &json!({"type": "content_block_stop", "index": index}),
+                    )?;
+                }
+                let index = if let Some((index, _)) = *block {
+                    index
+                } else {
+                    let index = *next_index;
+                    *next_index += 1;
+                    send_named_sse(
+                        stream,
+                        "content_block_start",
+                        &json!({"type": "content_block_start", "index": index, "content_block": {"type": "text", "text": ""}}),
+                    )?;
+                    *block = Some((index, false));
+                    index
+                };
+                send_named_sse(
+                    stream,
+                    "content_block_delta",
+                    &json!({"type": "content_block_delta", "index": index, "delta": {"type": "text_delta", "text": text}}),
+                )?;
+            }
+            ToolProjectionEvent::Start { index, id, name } => {
+                if let Some((open_index, _)) = block.take() {
+                    send_named_sse(
+                        stream,
+                        "content_block_stop",
+                        &json!({"type": "content_block_stop", "index": open_index}),
+                    )?;
+                }
+                let content_index = *next_index;
+                *next_index += 1;
+                if tool_indices.len() == index {
+                    tool_indices.push(content_index);
+                }
+                send_named_sse(
+                    stream,
+                    "content_block_start",
+                    &json!({"type": "content_block_start", "index": content_index, "content_block": {"type": "tool_use", "id": id, "name": name, "input": {}}}),
+                )?;
+            }
+            ToolProjectionEvent::Arguments { index, fragment } => {
+                if let Some(content_index) = tool_indices.get(index) {
+                    send_named_sse(
+                        stream,
+                        "content_block_delta",
+                        &json!({"type": "content_block_delta", "index": content_index, "delta": {"type": "input_json_delta", "partial_json": fragment}}),
+                    )?;
+                }
+            }
+            ToolProjectionEvent::End { index } => {
+                if let Some(content_index) = tool_indices.get(index) {
+                    send_named_sse(
+                        stream,
+                        "content_block_stop",
+                        &json!({"type": "content_block_stop", "index": content_index}),
+                    )?;
+                }
+            }
+            ToolProjectionEvent::Text(_) => {}
+        }
+    }
+    Ok(())
+}
+
+fn anthropic_stream_start(
+    stream: &mut impl Write,
+    request: &ResponseOptions,
+    id: &str,
+    prompt_tokens: u32,
+    cached_tokens: u32,
+) -> Result<(), (u16, String)> {
+    let cached = cached_tokens.min(prompt_tokens);
+    let written = prompt_tokens - cached;
+    send_named_sse(
+        stream,
+        "message_start",
+        &json!({"type": "message_start", "message": {"id": id, "type": "message", "role": "assistant", "model": request.model_id, "content": [], "stop_reason": Value::Null, "stop_sequence": Value::Null, "usage": {"input_tokens": prompt_tokens - cached - written, "output_tokens": 0, "cache_read_input_tokens": cached, "cache_creation_input_tokens": written}}}),
+    )
+}
+
+fn responses_stream_response(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+) -> Result<(), (u16, String)> {
+    send_sse_headers(stream).map_err(|error| (500, error))?;
+    let created = unix_time();
+    let message_id = random_id("msg_");
+    let reasoning_id = random_id("rs_");
+    let mut sequence = 0;
+    send_responses_sse(
+        stream,
+        &mut sequence,
+        json!({"type": "response.created", "response": {"id": id, "object": "response", "created_at": created, "status": "in_progress", "model": request.model_id, "output": []}}),
+    )?;
+    let mut reasoning_open = false;
+    let mut message_open = false;
+    let mut reasoning = String::new();
+    let mut content = String::new();
+    while let Ok(event) = active.events.recv() {
+        match event {
+            GenerationEvent::Chunk {
+                reasoning: true,
+                content: chunk,
+            } => {
+                if !request.reasoning_summary {
+                    continue;
+                }
+                if !reasoning_open {
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.added", "output_index": 0, "item": {"id": reasoning_id, "type": "reasoning", "status": "in_progress", "summary": []}}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.reasoning_summary_part.added", "item_id": reasoning_id, "output_index": 0, "summary_index": 0, "part": {"type": "summary_text", "text": ""}}),
+                    )?;
+                    reasoning_open = true;
+                }
+                reasoning.push_str(&chunk);
+                send_responses_sse(
+                    stream,
+                    &mut sequence,
+                    json!({"type": "response.reasoning_summary_text.delta", "item_id": reasoning_id, "output_index": 0, "summary_index": 0, "delta": chunk}),
+                )?;
+            }
+            GenerationEvent::Chunk {
+                reasoning: false,
+                content: chunk,
+            } => {
+                if request.has_tools {
+                    content.push_str(&chunk);
+                    continue;
+                }
+                if !message_open {
+                    let output_index = usize::from(reasoning_open);
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.added", "output_index": output_index, "item": {"id": message_id, "type": "message", "status": "in_progress", "role": "assistant", "content": []}}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.content_part.added", "item_id": message_id, "output_index": output_index, "content_index": 0, "part": {"type": "output_text", "text": "", "annotations": []}}),
+                    )?;
+                    message_open = true;
+                }
+                content.push_str(&chunk);
+                let output_index = usize::from(reasoning_open);
+                send_responses_sse(
+                    stream,
+                    &mut sequence,
+                    json!({"type": "response.output_text.delta", "item_id": message_id, "output_index": output_index, "content_index": 0, "delta": chunk}),
+                )?;
+            }
+            GenerationEvent::Finished(Ok(output)) => {
+                let (parsed_content, calls) =
+                    parse_generated_tools(state, &output.message.content, Protocol::Responses);
+                if request.has_tools {
+                    content = parsed_content;
+                }
+                let finish = if calls.is_empty() {
+                    output.finish_reason
+                } else {
+                    "tool_calls"
+                };
+                let status = if finish == "length" {
+                    "incomplete"
+                } else if finish == "error" {
+                    "failed"
+                } else {
+                    "completed"
+                };
+                let mut terminal_items = Vec::new();
+                let mut output_index = 0;
+                if reasoning_open {
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.reasoning_summary_text.done", "item_id": reasoning_id, "output_index": output_index, "summary_index": 0, "text": reasoning}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.reasoning_summary_part.done", "item_id": reasoning_id, "output_index": output_index, "summary_index": 0, "part": {"type": "summary_text", "text": reasoning}}),
+                    )?;
+                    let item = json!({"id": reasoning_id, "type": "reasoning", "status": status, "summary": [{"type": "summary_text", "text": reasoning}]});
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.done", "output_index": output_index, "item": item}),
+                    )?;
+                    terminal_items.push(item);
+                    output_index += 1;
+                }
+                if !content.is_empty() {
+                    if !message_open {
+                        send_responses_sse(
+                            stream,
+                            &mut sequence,
+                            json!({"type": "response.output_item.added", "output_index": output_index, "item": {"id": message_id, "type": "message", "status": "in_progress", "role": "assistant", "content": []}}),
+                        )?;
+                        send_responses_sse(
+                            stream,
+                            &mut sequence,
+                            json!({"type": "response.content_part.added", "item_id": message_id, "output_index": output_index, "content_index": 0, "part": {"type": "output_text", "text": "", "annotations": []}}),
+                        )?;
+                        send_responses_sse(
+                            stream,
+                            &mut sequence,
+                            json!({"type": "response.output_text.delta", "item_id": message_id, "output_index": output_index, "content_index": 0, "delta": content}),
+                        )?;
+                    }
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_text.done", "item_id": message_id, "output_index": output_index, "content_index": 0, "text": content}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.content_part.done", "item_id": message_id, "output_index": output_index, "content_index": 0, "part": {"type": "output_text", "text": content, "annotations": []}}),
+                    )?;
+                    let item = json!({"id": message_id, "type": "message", "status": status, "role": "assistant", "content": [{"type": "output_text", "text": content, "annotations": []}]});
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.done", "output_index": output_index, "item": item}),
+                    )?;
+                    terminal_items.push(item);
+                    output_index += 1;
+                }
+                for call in &calls {
+                    let item_id = random_id("fc_");
+                    let mut item = json!({"id": item_id, "type": "function_call", "status": status, "name": call.function.name, "call_id": call.id, "arguments": call.function.arguments});
+                    let mut added = item.clone();
+                    added["status"] = Value::String("in_progress".into());
+                    added["arguments"] = Value::String(String::new());
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.added", "output_index": output_index, "item": added}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.function_call_arguments.delta", "item_id": item_id, "output_index": output_index, "delta": call.function.arguments}),
+                    )?;
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.function_call_arguments.done", "item_id": item_id, "output_index": output_index, "name": call.function.name, "arguments": call.function.arguments}),
+                    )?;
+                    item["id"] = Value::String(item_id);
+                    send_responses_sse(
+                        stream,
+                        &mut sequence,
+                        json!({"type": "response.output_item.done", "output_index": output_index, "item": item}),
+                    )?;
+                    terminal_items.push(item);
+                    output_index += 1;
+                }
+                let event_type = if finish == "length" {
+                    "response.incomplete"
+                } else if finish == "error" {
+                    "response.failed"
+                } else {
+                    "response.completed"
+                };
+                let cached = output.cached_tokens.min(output.prompt_tokens);
+                return send_responses_sse(
+                    stream,
+                    &mut sequence,
+                    json!({"type": event_type, "response": {"id": id, "object": "response", "created_at": created, "status": status, "model": request.model_id, "output": terminal_items, "usage": {"input_tokens": output.prompt_tokens, "input_tokens_details": {"cached_tokens": cached, "cache_write_tokens": output.prompt_tokens - cached}, "output_tokens": output.completion_tokens, "output_tokens_details": {"reasoning_tokens": 0}, "total_tokens": output.prompt_tokens + output.completion_tokens}}}),
+                );
+            }
+            GenerationEvent::Finished(Err(error)) => {
+                let _ = send_sse_error(stream, &error);
+                return Ok(());
+            }
+            _ => {}
+        }
+    }
+    Err((500, "The model runtime stopped unexpectedly.".into()))
+}
+
+#[allow(clippy::too_many_arguments)]
+fn send_named_sse(
+    stream: &mut impl Write,
+    event: &str,
+    value: &Value,
+) -> Result<(), (u16, String)> {
+    let body = serde_json::to_vec(value).map_err(|error| (500, error.to_string()))?;
+    stream
+        .write_all(b"event: ")
+        .and_then(|()| stream.write_all(event.as_bytes()))
+        .and_then(|()| stream.write_all(b"\ndata: "))
+        .and_then(|()| stream.write_all(&body))
+        .and_then(|()| stream.write_all(b"\n\n"))
+        .map_err(|error| (500, error.to_string()))
+}
+
+fn send_responses_sse(
+    stream: &mut impl Write,
+    sequence: &mut u32,
+    value: Value,
+) -> Result<(), (u16, String)> {
+    let mut object = value
+        .as_object()
+        .cloned()
+        .ok_or_else(|| (500, "Responses event is not an object".to_owned()))?;
+    let event_type = object.shift_remove("type").unwrap_or(Value::Null);
+    let mut ordered = Map::new();
+    ordered.insert("type".into(), event_type);
+    ordered.insert("sequence_number".into(), Value::from(*sequence));
+    ordered.extend(object);
+    *sequence += 1;
+    send_sse(stream, &Value::Object(ordered)).map_err(|error| (500, error))
+}
+
+fn stream_response_with_keepalive(
+    stream: &mut TcpStream,
+    state: &State,
+    request: ResponseOptions,
+    active: crate::runtime::ActiveGeneration,
+    id: &str,
+    keepalive_interval: Duration,
+) -> Result<(), (u16, String)> {
+    let mut projector = ToolProjector::new();
+    let mut output = None;
+    let mut prefilling = true;
+    let mut headers_sent = false;
+    let mut role_sent = false;
+    let mut last_keepalive = Instant::now();
+    loop {
+        let event = match receive_stream_event(
+            stream,
+            &active,
+            prefilling && headers_sent,
+            &mut last_keepalive,
+            keepalive_interval,
+        ) {
+            Ok(Some(event)) => event,
+            Ok(None) => break,
+            Err(_) => return Ok(()),
+        };
+        match event {
+            GenerationEvent::Loading => {}
+            GenerationEvent::Context {
+                tokens_per_second, ..
+            } => {
+                prefilling = tokens_per_second.is_none();
+                if prefilling && !headers_sent {
+                    send_sse_headers(stream).map_err(|error| (500, error))?;
+                    headers_sent = true;
+                    last_keepalive = Instant::now();
+                } else if !prefilling {
+                    send_stream_start(stream, &request, id, &mut headers_sent, &mut role_sent)?;
+                }
+            }
+            GenerationEvent::Chunk { reasoning, content } => {
+                prefilling = false;
+                send_stream_start(stream, &request, id, &mut headers_sent, &mut role_sent)?;
+                if request.has_tools && !reasoning {
+                    let events = projector.push(&content, false, "call_");
+                    if send_chat_projection_events(stream, &request, id, events).is_err() {
+                        active.cancel.store(true, Ordering::Relaxed);
+                        return Ok(());
+                    }
+                    if TOOL_SYNTAXES
+                        .iter()
+                        .any(|syntax| projector.raw.contains(syntax.tool_end))
+                    {
+                        active.cancel.store(true, Ordering::Relaxed);
+                    }
+                } else if !content.is_empty() {
+                    let field = if reasoning {
+                        "reasoning_content"
+                    } else {
+                        "content"
+                    };
+                    let chunk = chunk_json(id, &request.model_id, json!({field: content}), None);
+                    if send_sse(stream, &chunk).is_err() {
+                        active.cancel.store(true, Ordering::Relaxed);
+                        return Ok(());
+                    }
+                }
+            }
+            GenerationEvent::Finished(result) => {
+                match result {
+                    Ok(result) => {
+                        send_stream_start(stream, &request, id, &mut headers_sent, &mut role_sent)?;
+                        output = Some(result);
+                    }
+                    Err(error) => {
+                        if headers_sent {
+                            let _ = send_sse_error(stream, &error);
+                            return Ok(());
+                        }
+                        return Err((500, error));
+                    }
+                }
+                break;
+            }
+        }
+    }
+    let output = match output {
+        Some(output) => output,
+        None if headers_sent => {
+            let _ = send_sse_error(stream, "The model runtime stopped unexpectedly.");
+            return Ok(());
+        }
+        None => return Err((500, "The model runtime stopped unexpectedly.".into())),
+    };
+    if request.has_tools {
+        let events = projector.push("", true, "call_");
+        send_chat_projection_events(stream, &request, id, events)?;
+    }
+    let (_, calls) = parse_generated_tools_with_ids(
+        state,
+        &output.message.content,
+        Protocol::Chat,
+        &projector.ids,
+    );
+    if request.has_tools && projector.ids.is_empty() && !calls.is_empty() {
+        send_sse(
+            stream,
+            &chunk_json(
+                id,
+                &request.model_id,
+                json!({"tool_calls": tool_calls_json(&calls)}),
+                None,
+            ),
+        )
+        .map_err(|error| (500, error))?;
+    }
+    let finish = if calls.is_empty() {
+        output.finish_reason
+    } else {
+        "tool_calls"
+    };
+    send_sse(
+        stream,
+        &chunk_json(id, &request.model_id, json!({}), Some(finish)),
+    )
+    .map_err(|error| (500, error))?;
+    if request.include_usage {
+        let usage = json!({
+            "id": id,
+            "object": "chat.completion.chunk",
+            "created": unix_time(),
+            "model": request.model_id,
+            "choices": [],
+            "usage": usage_json(output.prompt_tokens, output.cached_tokens, output.completion_tokens),
+        });
+        send_sse(stream, &usage).map_err(|error| (500, error))?;
+    }
+    stream
+        .write_all(b"data: [DONE]\n\n")
+        .map_err(|error| (500, error.to_string()))
+}
+
+fn send_chat_projection_events(
+    stream: &mut impl Write,
+    request: &ResponseOptions,
+    response_id: &str,
+    events: Vec<ToolProjectionEvent>,
+) -> Result<(), (u16, String)> {
+    for event in events {
+        let delta = match event {
+            ToolProjectionEvent::Text(content) if !content.is_empty() => {
+                json!({"content": content})
+            }
+            ToolProjectionEvent::Start { index, id, name } => json!({"tool_calls": [{
+                "index": index, "id": id, "type": "function",
+                "function": {"name": name, "arguments": ""}
+            }]}),
+            ToolProjectionEvent::Arguments { index, fragment } => json!({
+                "tool_calls": [{"index": index, "function": {"arguments": fragment}}]
+            }),
+            ToolProjectionEvent::End { .. } | ToolProjectionEvent::Text(_) => continue,
+        };
+        send_sse(
+            stream,
+            &chunk_json(response_id, &request.model_id, delta, None),
+        )
+        .map_err(|error| (500, error))?;
+    }
+    Ok(())
+}
+
+pub(super) fn send_stream_start(
+    stream: &mut impl Write,
+    request: &ResponseOptions,
+    id: &str,
+    headers_sent: &mut bool,
+    role_sent: &mut bool,
+) -> Result<(), (u16, String)> {
+    if !*headers_sent {
+        send_sse_headers(stream).map_err(|error| (500, error))?;
+        *headers_sent = true;
+    }
+    if !*role_sent {
+        let role = chunk_json(id, &request.model_id, json!({"role": "assistant"}), None);
+        send_sse(stream, &role).map_err(|error| (500, error))?;
+        *role_sent = true;
+    }
+    Ok(())
+}
+
+pub(super) fn receive_stream_event(
+    stream: &mut impl Write,
+    active: &crate::runtime::ActiveGeneration,
+    prefilling: bool,
+    last_keepalive: &mut Instant,
+    keepalive_interval: Duration,
+) -> Result<Option<GenerationEvent>, String> {
+    if !prefilling {
+        return Ok(active.events.recv().ok());
+    }
+    loop {
+        if last_keepalive.elapsed() >= keepalive_interval {
+            if let Err(error) = stream.write_all(b": prefill\n\n") {
+                active.cancel.store(true, Ordering::Relaxed);
+                return Err(error.to_string());
+            }
+            *last_keepalive = Instant::now();
+        }
+        let remaining = keepalive_interval.saturating_sub(last_keepalive.elapsed());
+        match active.events.recv_timeout(remaining) {
+            Ok(event) => return Ok(Some(event)),
+            Err(std::sync::mpsc::RecvTimeoutError::Timeout) => {}
+            Err(std::sync::mpsc::RecvTimeoutError::Disconnected) => return Ok(None),
+        }
+    }
+}
+
+fn wait_for_output(
+    active: crate::runtime::ActiveGeneration,
+) -> Result<crate::engine::GenerationOutput, (u16, String)> {
+    let mut content = String::new();
+    while let Ok(event) = active.events.recv() {
+        match event {
+            GenerationEvent::Chunk {
+                reasoning: false,
+                content: chunk,
+            } => {
+                content.push_str(&chunk);
+                if TOOL_SYNTAXES
+                    .iter()
+                    .any(|syntax| content.contains(syntax.tool_end))
+                {
+                    active.cancel.store(true, Ordering::Relaxed);
+                }
+            }
+            GenerationEvent::Finished(result) => {
+                return result.map_err(|error| (500, error));
+            }
+            _ => {}
+        }
+    }
+    Err((500, "The model runtime stopped unexpectedly.".into()))
+}
+
+pub(super) fn usage_json(prompt: u32, cached: u32, completion: u32) -> Value {
+    let cached = cached.min(prompt);
+    json!({
+        "prompt_tokens": prompt,
+        "completion_tokens": completion,
+        "total_tokens": prompt + completion,
+        "prompt_tokens_details": {
+            "cached_tokens": cached,
+            "cache_write_tokens": prompt - cached
+        }
+    })
+}
+
+pub(super) fn chunk_json(id: &str, model: &str, delta: Value, finish: Option<&str>) -> Value {
+    json!({
+        "id": id,
+        "object": "chat.completion.chunk",
+        "created": unix_time(),
+        "model": model,
+        "choices": [{"index": 0, "delta": delta, "finish_reason": finish}]
+    })
+}

src/server/tools.rs

@@ -0,0 +1,733 @@
+use super::*;
+
+enum ToolProjectionState {
+    Seeking,
+    Invokes,
+    Parameters,
+    Value,
+    Done,
+    Failed,
+}
+
+pub(super) enum ToolProjectionEvent {
+    Text(String),
+    Start {
+        index: usize,
+        id: String,
+        name: String,
+    },
+    Arguments {
+        index: usize,
+        fragment: String,
+    },
+    End {
+        index: usize,
+    },
+}
+
+pub(super) struct ToolProjector {
+    pub(super) raw: String,
+    position: usize,
+    text_emitted: usize,
+    state: ToolProjectionState,
+    index: usize,
+    pub(super) ids: Vec<String>,
+    first_parameter: bool,
+    string_parameter: bool,
+    syntax: Option<ToolSyntax>,
+}
+
+impl ToolProjector {
+    pub(super) fn new() -> Self {
+        Self {
+            raw: String::new(),
+            position: 0,
+            text_emitted: 0,
+            state: ToolProjectionState::Seeking,
+            index: 0,
+            ids: Vec::new(),
+            first_parameter: true,
+            string_parameter: false,
+            syntax: None,
+        }
+    }
+
+    pub(super) fn push(
+        &mut self,
+        chunk: &str,
+        final_chunk: bool,
+        prefix: &str,
+    ) -> Vec<ToolProjectionEvent> {
+        self.raw.push_str(chunk);
+        let mut events = Vec::new();
+        loop {
+            match self.state {
+                ToolProjectionState::Seeking => {
+                    if let Some((start, syntax)) = TOOL_SYNTAXES
+                        .iter()
+                        .filter_map(|syntax| {
+                            self.raw
+                                .find(syntax.tool_start)
+                                .map(|start| (start, *syntax))
+                        })
+                        .min_by_key(|(start, _)| *start)
+                    {
+                        if start > self.text_emitted {
+                            let text = self.raw[self.text_emitted..start].trim_end();
+                            if !text.is_empty() {
+                                events.push(ToolProjectionEvent::Text(text.to_owned()));
+                            }
+                        }
+                        self.position = start + syntax.tool_start.len();
+                        self.text_emitted = start;
+                        self.syntax = Some(syntax);
+                        self.state = ToolProjectionState::Invokes;
+                    } else {
+                        let limit = if final_chunk {
+                            self.raw.len()
+                        } else {
+                            TOOL_SYNTAXES
+                                .iter()
+                                .map(|syntax| {
+                                    safe_before_partial_marker(&self.raw, syntax.tool_start)
+                                })
+                                .min()
+                                .unwrap_or(self.raw.len())
+                        };
+                        if limit > self.text_emitted {
+                            let text = &self.raw[self.text_emitted..limit];
+                            if !text.trim().is_empty() {
+                                events.push(ToolProjectionEvent::Text(text.to_owned()));
+                                self.text_emitted = limit;
+                            }
+                        }
+                        break;
+                    }
+                }
+                ToolProjectionState::Invokes => {
+                    let syntax = self.syntax.unwrap();
+                    self.skip_whitespace();
+                    if self.full_at(syntax.tool_end) {
+                        self.position += syntax.tool_end.len();
+                        self.state = ToolProjectionState::Done;
+                        break;
+                    }
+                    if self.partial_at(syntax.tool_end) || self.partial_at(syntax.invoke_start) {
+                        break;
+                    }
+                    if !self.full_at(syntax.invoke_start) {
+                        self.state = ToolProjectionState::Failed;
+                        break;
+                    }
+                    let Some(tag_end) = self.raw[self.position..].find('>') else {
+                        break;
+                    };
+                    let tag_end = self.position + tag_end + 1;
+                    let Some(name) = dsml_attribute(&self.raw[self.position..tag_end], "name")
+                    else {
+                        self.state = ToolProjectionState::Failed;
+                        break;
+                    };
+                    let id = random_tool_id(prefix);
+                    self.ids.push(id.clone());
+                    events.push(ToolProjectionEvent::Start {
+                        index: self.index,
+                        id,
+                        name,
+                    });
+                    events.push(ToolProjectionEvent::Arguments {
+                        index: self.index,
+                        fragment: "{".into(),
+                    });
+                    self.position = tag_end;
+                    self.first_parameter = true;
+                    self.state = ToolProjectionState::Parameters;
+                }
+                ToolProjectionState::Parameters => {
+                    let syntax = self.syntax.unwrap();
+                    self.skip_whitespace();
+                    if self.full_at(syntax.invoke_end) {
+                        events.push(ToolProjectionEvent::Arguments {
+                            index: self.index,
+                            fragment: "}".into(),
+                        });
+                        events.push(ToolProjectionEvent::End { index: self.index });
+                        self.position += syntax.invoke_end.len();
+                        self.index += 1;
+                        self.state = ToolProjectionState::Invokes;
+                        continue;
+                    }
+                    if self.partial_at(syntax.invoke_end) || self.partial_at(syntax.parameter_start)
+                    {
+                        break;
+                    }
+                    if !self.full_at(syntax.parameter_start) {
+                        self.state = ToolProjectionState::Failed;
+                        break;
+                    }
+                    let Some(tag_end) = self.raw[self.position..].find('>') else {
+                        break;
+                    };
+                    let tag_end = self.position + tag_end + 1;
+                    let tag = &self.raw[self.position..tag_end];
+                    let Some(name) = dsml_attribute(tag, "name") else {
+                        self.state = ToolProjectionState::Failed;
+                        break;
+                    };
+                    self.string_parameter =
+                        dsml_attribute(tag, "string").as_deref() != Some("false");
+                    let mut fragment = if self.first_parameter {
+                        String::new()
+                    } else {
+                        ",".into()
+                    };
+                    self.first_parameter = false;
+                    fragment
+                        .push_str(&serde_json::to_string(&name).unwrap_or_else(|_| "\"\"".into()));
+                    fragment.push(':');
+                    if self.string_parameter {
+                        fragment.push('"');
+                    }
+                    events.push(ToolProjectionEvent::Arguments {
+                        index: self.index,
+                        fragment,
+                    });
+                    self.position = tag_end;
+                    self.state = ToolProjectionState::Value;
+                }
+                ToolProjectionState::Value => {
+                    let syntax = self.syntax.unwrap();
+                    if let Some(relative_end) = self.raw[self.position..].find(syntax.parameter_end)
+                    {
+                        let end = self.position + relative_end;
+                        self.emit_value(end, &mut events);
+                        if self.string_parameter {
+                            events.push(ToolProjectionEvent::Arguments {
+                                index: self.index,
+                                fragment: "\"".into(),
+                            });
+                        }
+                        self.position = end + syntax.parameter_end.len();
+                        self.state = ToolProjectionState::Parameters;
+                        continue;
+                    }
+                    let limit = safe_parameter_value_limit(
+                        &self.raw,
+                        self.position,
+                        syntax.parameter_end,
+                        self.string_parameter,
+                    );
+                    self.emit_value(limit, &mut events);
+                    break;
+                }
+                ToolProjectionState::Done | ToolProjectionState::Failed => break,
+            }
+        }
+        events
+    }
+
+    fn emit_value(&mut self, end: usize, events: &mut Vec<ToolProjectionEvent>) {
+        if end <= self.position {
+            return;
+        }
+        let raw = &self.raw[self.position..end];
+        let fragment = if self.string_parameter {
+            let value = unescape_dsml(raw);
+            let encoded = serde_json::to_string(&value).unwrap_or_else(|_| "\"\"".into());
+            encoded[1..encoded.len() - 1].to_owned()
+        } else {
+            raw.to_owned()
+        };
+        events.push(ToolProjectionEvent::Arguments {
+            index: self.index,
+            fragment,
+        });
+        self.position = end;
+    }
+
+    fn skip_whitespace(&mut self) {
+        while self.raw[self.position..]
+            .chars()
+            .next()
+            .is_some_and(char::is_whitespace)
+        {
+            self.position += self.raw[self.position..].chars().next().unwrap().len_utf8();
+        }
+    }
+
+    fn full_at(&self, marker: &str) -> bool {
+        self.raw.as_bytes()[self.position..].starts_with(marker.as_bytes())
+    }
+
+    fn partial_at(&self, marker: &str) -> bool {
+        let tail = &self.raw.as_bytes()[self.position..];
+        tail.len() < marker.len() && marker.as_bytes().starts_with(tail)
+    }
+}
+
+fn dsml_attribute(tag: &str, name: &str) -> Option<String> {
+    let start = tag.find(&format!("{name}=\""))? + name.len() + 2;
+    let end = start + tag[start..].find('"')?;
+    Some(unescape_dsml(&tag[start..end]))
+}
+
+fn safe_before_partial_marker(text: &str, marker: &str) -> usize {
+    let mut limit = text.len().saturating_sub(marker.len().saturating_sub(1));
+    while !text.is_char_boundary(limit) {
+        limit -= 1;
+    }
+    limit
+}
+
+fn safe_parameter_value_limit(text: &str, start: usize, end_marker: &str, string: bool) -> usize {
+    let bytes = text.as_bytes();
+    let marker = end_marker.as_bytes();
+    let mut limit = bytes.len();
+    for length in (1..marker.len().min(bytes.len().saturating_sub(start) + 1)).rev() {
+        if bytes[start..].ends_with(&marker[..length]) {
+            limit -= length;
+            break;
+        }
+    }
+    if string {
+        for entity in ["&amp;", "&lt;", "&gt;", "&quot;", "&apos;"] {
+            let entity = entity.as_bytes();
+            for length in 1..entity.len() {
+                if bytes[start..limit].ends_with(&entity[..length]) {
+                    limit -= length;
+                    return limit;
+                }
+            }
+        }
+    }
+    limit
+}
+
+pub(super) fn render_messages(
+    state: &State,
+    messages: &[ApiMessage],
+    tools: &[Value],
+    tool_schemas: &[String],
+    tools_enabled: bool,
+    protocol: Protocol,
+) -> Result<(String, Vec<ChatTurn>), (u16, String)> {
+    validate_tool_results(state, messages, protocol)?;
+    let preserve_reasoning = tools_enabled
+        || messages.iter().any(|message| {
+            matches!(message.role.as_str(), "tool" | "function") || !message.tool_calls.is_empty()
+        });
+    let mut system = String::new();
+    if tools_enabled {
+        system.push_str(TOOLS_PROMPT);
+        if tool_schemas.is_empty() {
+            for tool in tools {
+                let schema = tool.get("function").unwrap_or(tool);
+                if !system.ends_with("\n\n") {
+                    system.push('\n');
+                }
+                system.push_str(
+                    &serde_json::to_string(schema)
+                        .map_err(|error| (400, format!("invalid tool schema: {error}")))?,
+                );
+                system.push('\n');
+            }
+        } else {
+            for schema in tool_schemas {
+                if !system.ends_with("\n\n") {
+                    system.push('\n');
+                }
+                system.push_str(schema);
+                system.push('\n');
+            }
+        }
+        system.push_str(
+            "\nYou MUST strictly follow the above defined tool name and parameter schemas to invoke tool calls. Use the exact parameter names from the schemas.",
+        );
+    }
+
+    let mut turns = Vec::<ChatTurn>::new();
+    for message in messages {
+        let content = content_text(&message.content);
+        match message.role.as_str() {
+            "system" | "developer" => {
+                if !system.is_empty() {
+                    system.push_str("\n\n");
+                }
+                system.push_str(&content);
+            }
+            "user" => turns.push(ChatTurn {
+                user: true,
+                skip_previous_eos: false,
+                reasoning: None,
+                reasoning_complete: true,
+                content,
+            }),
+            "tool" | "function" => {
+                let wrapped = format!(
+                    "<tool_result>{}</tool_result>",
+                    escape_tool_result(&content)
+                );
+                if let Some(previous) = turns.last_mut()
+                    && previous.user
+                    && previous.content.starts_with("<tool_result>")
+                {
+                    previous.content.push_str(&wrapped);
+                } else {
+                    turns.push(ChatTurn {
+                        user: true,
+                        skip_previous_eos: protocol == Protocol::Responses,
+                        reasoning: None,
+                        reasoning_complete: true,
+                        content: wrapped,
+                    });
+                }
+            }
+            "assistant" => {
+                let mut content = content;
+                if !message.tool_calls.is_empty() {
+                    content.push_str(&replayed_or_canonical_tools(state, &message.tool_calls));
+                }
+                let reasoning = content_text(&message.reasoning_content);
+                turns.push(ChatTurn {
+                    user: false,
+                    skip_previous_eos: false,
+                    reasoning: (preserve_reasoning && !reasoning.is_empty()).then_some(reasoning),
+                    reasoning_complete: true,
+                    content,
+                });
+            }
+            role => return Err((400, format!("unsupported message role: {role}"))),
+        }
+    }
+    Ok((system, turns))
+}
+
+pub(super) fn validate_tool_results(
+    state: &State,
+    messages: &[ApiMessage],
+    protocol: Protocol,
+) -> Result<(), (u16, String)> {
+    if !matches!(protocol, Protocol::Anthropic | Protocol::Responses) {
+        return Ok(());
+    }
+    let memory = state.tool_memory.lock().ok();
+    for (index, message) in messages.iter().enumerate() {
+        if !matches!(message.role.as_str(), "tool" | "function") || message.tool_call_id.is_empty()
+        {
+            continue;
+        }
+        let id = &message.tool_call_id;
+        let live = memory
+            .as_ref()
+            .is_some_and(|memory| memory.contains_key(id));
+        let replayed = messages[..index].iter().any(|message| {
+            message.role == "assistant" && message.tool_calls.iter().any(|call| call.id == *id)
+        });
+        if live || replayed {
+            continue;
+        }
+        let message = match protocol {
+            Protocol::Anthropic => format!(
+                "Anthropic continuation state is not available for tool_use_id {id}; retry by replaying the full messages history"
+            ),
+            Protocol::Responses => format!(
+                "Responses continuation state is not available for call_id {id}; retry by replaying the full input history"
+            ),
+            _ => unreachable!(),
+        };
+        return Err((400, message));
+    }
+    Ok(())
+}
+
+fn replayed_or_canonical_tools(state: &State, calls: &[ApiToolCall]) -> String {
+    if let Ok(memory) = state.tool_memory.lock()
+        && let Some(raw) = calls.iter().find_map(|call| {
+            (!call.id.is_empty())
+                .then(|| memory.get(&call.id))
+                .flatten()
+        })
+    {
+        return raw.clone();
+    }
+    canonical_tools(calls)
+}
+
+pub(super) fn canonical_tools(calls: &[ApiToolCall]) -> String {
+    let mut output = String::from("\n\n<｜DSML｜tool_calls>\n");
+    for call in calls {
+        output.push_str("<｜DSML｜invoke name=\"");
+        output.push_str(&escape_attribute(&call.function.name));
+        output.push_str("\">\n");
+        match serde_json::from_str::<Value>(&call.function.arguments) {
+            Ok(Value::Object(arguments)) => {
+                for (name, value) in arguments {
+                    output.push_str("<｜DSML｜parameter name=\"");
+                    output.push_str(&escape_attribute(&name));
+                    let string = value.as_str();
+                    output.push_str(if string.is_some() {
+                        "\" string=\"true\">"
+                    } else {
+                        "\" string=\"false\">"
+                    });
+                    if let Some(value) = string {
+                        output.push_str(&escape_parameter(value));
+                    } else {
+                        output.push_str(&escape_json_parameter(&value.to_string()));
+                    }
+                    output.push_str("</｜DSML｜parameter>\n");
+                }
+            }
+            _ => {
+                output.push_str("<｜DSML｜parameter name=\"arguments\" string=\"true\">");
+                output.push_str(&escape_parameter(&call.function.arguments));
+                output.push_str("</｜DSML｜parameter>\n");
+            }
+        }
+        output.push_str("</｜DSML｜invoke>\n");
+    }
+    output.push_str("</｜DSML｜tool_calls>");
+    output
+}
+
+pub(super) fn parse_generated_tools(
+    state: &State,
+    text: &str,
+    protocol: Protocol,
+) -> (String, Vec<ApiToolCall>) {
+    parse_generated_tools_with_ids(state, text, protocol, &[])
+}
+
+pub(super) fn parse_generated_tools_with_ids(
+    state: &State,
+    text: &str,
+    protocol: Protocol,
+    streamed_ids: &[String],
+) -> (String, Vec<ApiToolCall>) {
+    let Some((start, syntax)) = TOOL_SYNTAXES
+        .iter()
+        .filter_map(|syntax| text.find(syntax.tool_start).map(|start| (start, *syntax)))
+        .min_by_key(|(start, _)| *start)
+    else {
+        return (text.to_owned(), Vec::new());
+    };
+    let Some(relative_end) = text[start..].find(syntax.tool_end) else {
+        return (text.to_owned(), Vec::new());
+    };
+    let end = start + relative_end + syntax.tool_end.len();
+    let content = text[..start].trim_end();
+    let raw = &text[content.len()..end];
+    let mut calls = Vec::new();
+    let mut cursor = raw.find(syntax.tool_start).unwrap() + syntax.tool_start.len();
+    loop {
+        skip_text_whitespace(raw, &mut cursor);
+        if raw[cursor..].starts_with(syntax.tool_end) {
+            break;
+        }
+        if !raw[cursor..].starts_with(syntax.invoke_start) {
+            return (text.to_owned(), Vec::new());
+        }
+        let Some(tag_end) = raw[cursor..].find('>').map(|end| cursor + end + 1) else {
+            return (text.to_owned(), Vec::new());
+        };
+        let Some(name) = dsml_attribute(&raw[cursor..tag_end], "name") else {
+            return (text.to_owned(), Vec::new());
+        };
+        cursor = tag_end;
+        let mut arguments = Map::new();
+        loop {
+            skip_text_whitespace(raw, &mut cursor);
+            if raw[cursor..].starts_with(syntax.invoke_end) {
+                cursor += syntax.invoke_end.len();
+                break;
+            }
+            let Some((name, value)) = parse_tool_parameter(raw, &mut cursor, syntax) else {
+                return (text.to_owned(), Vec::new());
+            };
+            arguments.insert(name, value);
+        }
+        calls.push(ApiToolCall {
+            id: String::new(),
+            function: ApiFunction {
+                name,
+                arguments: Value::Object(arguments).to_string(),
+            },
+        });
+    }
+    if calls.is_empty() {
+        return (text.to_owned(), Vec::new());
+    }
+    let prefix = if protocol == Protocol::Anthropic {
+        "toolu_"
+    } else {
+        "call_"
+    };
+    for (index, call) in calls.iter_mut().enumerate() {
+        call.id = streamed_ids
+            .get(index)
+            .cloned()
+            .unwrap_or_else(|| random_tool_id(prefix));
+    }
+    if let Ok(mut memory) = state.tool_memory.lock() {
+        // ponytail: one process-local replay table; add LRU eviction if 100k live tool ids is measured insufficient.
+        if memory.len() >= 100_000 {
+            memory.clear();
+        }
+        for call in &calls {
+            memory.insert(call.id.clone(), raw.to_owned());
+        }
+    }
+    (content.to_owned(), calls)
+}
+
+#[derive(Clone, Copy)]
+pub(super) struct ToolSyntax {
+    pub(super) tool_start: &'static str,
+    pub(super) tool_end: &'static str,
+    pub(super) invoke_start: &'static str,
+    pub(super) invoke_end: &'static str,
+    pub(super) parameter_start: &'static str,
+    pub(super) parameter_end: &'static str,
+}
+
+pub(super) const TOOL_SYNTAXES: [ToolSyntax; 3] = [
+    ToolSyntax {
+        tool_start: "<｜DSML｜tool_calls>",
+        tool_end: "</｜DSML｜tool_calls>",
+        invoke_start: "<｜DSML｜invoke",
+        invoke_end: "</｜DSML｜invoke>",
+        parameter_start: "<｜DSML｜parameter",
+        parameter_end: "</｜DSML｜parameter>",
+    },
+    ToolSyntax {
+        tool_start: "<DSML｜tool_calls>",
+        tool_end: "</DSML｜tool_calls>",
+        invoke_start: "<DSML｜invoke",
+        invoke_end: "</DSML｜invoke>",
+        parameter_start: "<DSML｜parameter",
+        parameter_end: "</DSML｜parameter>",
+    },
+    ToolSyntax {
+        tool_start: "<tool_calls>",
+        tool_end: "</tool_calls>",
+        invoke_start: "<invoke",
+        invoke_end: "</invoke>",
+        parameter_start: "<parameter",
+        parameter_end: "</parameter>",
+    },
+];
+
+fn parse_tool_parameter(
+    text: &str,
+    cursor: &mut usize,
+    syntax: ToolSyntax,
+) -> Option<(String, Value)> {
+    if !text[*cursor..].starts_with(syntax.parameter_start) {
+        return None;
+    }
+    let tag_end = text[*cursor..].find('>').map(|end| *cursor + end + 1)?;
+    let tag = &text[*cursor..tag_end];
+    let name = dsml_attribute(tag, "name")?;
+    let is_string = dsml_attribute(tag, "string");
+    *cursor = tag_end;
+    let mut nested_start = *cursor;
+    skip_text_whitespace(text, &mut nested_start);
+    if is_string.is_none() && text[nested_start..].starts_with(syntax.parameter_start) {
+        *cursor = nested_start;
+        let mut nested = Map::new();
+        loop {
+            skip_text_whitespace(text, cursor);
+            if !text[*cursor..].starts_with(syntax.parameter_start) {
+                break;
+            }
+            let (name, value) = parse_tool_parameter(text, cursor, syntax)?;
+            nested.insert(name, value);
+        }
+        skip_text_whitespace(text, cursor);
+        if !text[*cursor..].starts_with(syntax.parameter_end) {
+            return None;
+        }
+        *cursor += syntax.parameter_end.len();
+        return Some((name, Value::Object(nested)));
+    }
+    let value_end = text[*cursor..]
+        .find(syntax.parameter_end)
+        .map(|end| *cursor + end)?;
+    let raw = &text[*cursor..value_end];
+    *cursor = value_end + syntax.parameter_end.len();
+    let value = if is_string.as_deref().unwrap_or("true") == "true" {
+        Value::String(unescape_dsml(raw))
+    } else {
+        serde_json::from_str(raw).unwrap_or(Value::Null)
+    };
+    Some((name, value))
+}
+
+fn skip_text_whitespace(text: &str, cursor: &mut usize) {
+    while text[*cursor..]
+        .chars()
+        .next()
+        .is_some_and(char::is_whitespace)
+    {
+        *cursor += text[*cursor..].chars().next().unwrap().len_utf8();
+    }
+}
+
+pub(super) fn tool_calls_json(calls: &[ApiToolCall]) -> Value {
+    Value::Array(
+        calls
+            .iter()
+            .map(|call| {
+                json!({
+                    "id": call.id,
+                    "type": "function",
+                    "function": {
+                        "name": call.function.name,
+                        "arguments": call.function.arguments,
+                    }
+                })
+            })
+            .collect(),
+    )
+}
+
+pub(super) fn content_text(value: &Value) -> String {
+    match value {
+        Value::String(text) => text.clone(),
+        Value::Array(parts) => parts
+            .iter()
+            .filter_map(|part| match part {
+                Value::String(text) => Some(text.as_str()),
+                Value::Object(object) => object.get("text").and_then(Value::as_str),
+                _ => None,
+            })
+            .collect(),
+        _ => String::new(),
+    }
+}
+
+pub(super) fn escape_attribute(text: &str) -> String {
+    text.replace('&', "&amp;")
+        .replace('<', "&lt;")
+        .replace('>', "&gt;")
+        .replace('"', "&quot;")
+}
+
+pub(super) fn escape_parameter(text: &str) -> String {
+    text.replace("</｜DSML｜parameter>", "&lt;/｜DSML｜parameter>")
+}
+
+pub(super) fn escape_json_parameter(text: &str) -> String {
+    text.replace("</｜DSML｜parameter>", "\\u003c/｜DSML｜parameter>")
+}
+
+pub(super) fn escape_tool_result(text: &str) -> String {
+    text.replace("</tool_result>", "&lt;/tool_result>")
+}
+
+pub(super) fn unescape_dsml(text: &str) -> String {
+    text.replace("&quot;", "\"")
+        .replace("&gt;", ">")
+        .replace("&lt;", "<")
+        .replace("&amp;", "&")
+}