Move preferences from the database to a YAML config file

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

src/settings.rs

@@ -1,4 +1,5 @@
 use crate::model::{self, EngineArtifacts, ModelChoice};
+use serde::{Deserialize, Serialize};
 use std::fmt;
 use std::path::Path;
 
@@ -17,7 +18,8 @@ const DEFAULT_KV_MIN_TOKENS: u32 = 512;
 const DEFAULT_KV_COLD_MAX_TOKENS: u32 = 30_000;
 const DEFAULT_KV_CONTINUED_INTERVAL_TOKENS: u32 = 10_000;
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct SpeculativePreferences {
     pub(crate) mtp_draft_tokens: i32,
     pub(crate) mtp_margin: f32,
@@ -94,13 +96,70 @@ pub(crate) struct EngineSpeculativeSettings {
     pub(crate) dspark_strict: bool,
 }
 
-#[derive(Clone, Copy, Debug, Eq, PartialEq)]
+/// An expert count, or a whole GiB budget. Written as `4` or `64GB`, the same
+/// spelling the preferences window accepts.
+#[derive(Clone, Copy, Debug, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(into = "String", try_from = "String")]
 pub(crate) enum StreamingCacheBudget {
     Experts(u32),
     Gib(u64),
 }
 
-#[derive(Clone, Debug, Default, Eq, PartialEq)]
+impl fmt::Display for StreamingCacheBudget {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        match self {
+            Self::Experts(experts) => write!(formatter, "{experts}"),
+            Self::Gib(gib) => write!(formatter, "{gib}GB"),
+        }
+    }
+}
+
+impl From<StreamingCacheBudget> for String {
+    fn from(budget: StreamingCacheBudget) -> Self {
+        budget.to_string()
+    }
+}
+
+impl TryFrom<String> for StreamingCacheBudget {
+    type Error = String;
+
+    fn try_from(value: String) -> Result<Self, Self::Error> {
+        value.parse()
+    }
+}
+
+impl std::str::FromStr for StreamingCacheBudget {
+    type Err = String;
+
+    fn from_str(value: &str) -> Result<Self, Self::Err> {
+        let value = value.trim();
+        let gib = value
+            .get(value.len().saturating_sub(2)..)
+            .is_some_and(|suffix| suffix.eq_ignore_ascii_case("gb"));
+        let digits = if gib {
+            &value[..value.len() - 2]
+        } else {
+            value
+        };
+        let number = (!digits.is_empty() && digits.bytes().all(|byte| byte.is_ascii_digit()))
+            .then(|| digits.parse::<u64>().ok())
+            .flatten()
+            .filter(|number| *number > 0);
+        match number {
+            Some(number) if gib && number <= u64::MAX / GIB => Ok(Self::Gib(number)),
+            Some(number) if !gib && number <= u64::from(u32::MAX) => {
+                Ok(Self::Experts(number as u32))
+            }
+            _ => Err(
+                "SSD cache budget must be a positive expert count or whole GiB value such as 64GB."
+                    .into(),
+            ),
+        }
+    }
+}
+
+#[derive(Clone, Debug, Default, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct SsdPreferences {
     pub(crate) enabled: bool,
     pub(crate) cold: bool,
@@ -166,7 +225,8 @@ pub(crate) struct EngineSsdSettings {
 
 /// Disk KV cache management, mirroring ds4's `--kv-disk-space-mb` budget and
 /// its store-side gates. Unset values keep the DS4 defaults.
-#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
+#[derive(Clone, Copy, Debug, Default, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct KvCachePreferences {
     pub(crate) budget_gib: Option<u64>,
     pub(crate) min_tokens: Option<u32>,
@@ -219,7 +279,8 @@ impl KvCacheSettings {
     }
 }
 
-#[derive(Clone, Debug, Default, PartialEq)]
+#[derive(Clone, Debug, Default, Deserialize, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct SteeringPreferences {
     pub(crate) file: Option<String>,
     pub(crate) ffn_scale: Option<f32>,
@@ -270,7 +331,8 @@ pub(crate) struct EngineSteeringSettings {
     pub(crate) attention_scale: f32,
 }
 
-#[derive(Clone, Debug, Default, Eq, PartialEq)]
+#[derive(Clone, Debug, Default, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct DiagnosticPreferences {
     pub(crate) simulated_used_memory_gib: Option<u64>,
     pub(crate) expert_profile_path: Option<String>,
@@ -298,7 +360,8 @@ pub(crate) struct EngineDiagnosticSettings {
     pub(crate) expert_profile_path: Option<String>,
 }
 
-#[derive(Clone, Debug, Default, PartialEq)]
+#[derive(Clone, Debug, Default, Deserialize, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct RuntimePreferences {
     pub(crate) execution: ExecutionPreferences,
     pub(crate) speculative: SpeculativePreferences,
@@ -354,7 +417,8 @@ pub(crate) struct EngineSettings {
     pub(crate) diagnostics: EngineDiagnosticSettings,
 }
 
-#[derive(Clone, Debug, Default, Eq, PartialEq)]
+#[derive(Clone, Debug, Default, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct ExecutionPreferences {
     pub(crate) cpu_threads: Option<u32>,
     pub(crate) power_percent: Option<u8>,
@@ -420,28 +484,16 @@ pub(crate) struct EngineExecutionSettings {
     pub(crate) warm_weights: bool,
 }
 
-#[derive(Clone, Copy, Debug, Default, Eq, PartialEq)]
+#[derive(Clone, Copy, Debug, Default, Deserialize, Eq, PartialEq, Serialize)]
+#[serde(rename_all = "lowercase")]
 pub(crate) enum ReasoningMode {
+    #[serde(rename = "none")]
     Direct,
     #[default]
     High,
     Max,
 }
 
-impl ReasoningMode {
-    pub(crate) fn id(self) -> &'static str {
-        match self {
-            Self::Direct => "none",
-            Self::High => "high",
-            Self::Max => "max",
-        }
-    }
-
-    pub(crate) fn from_id(id: &str) -> Option<Self> {
-        REASONING_MODES.into_iter().find(|mode| mode.id() == id)
-    }
-}
-
 impl fmt::Display for ReasoningMode {
     fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
         formatter.write_str(match self {
@@ -452,7 +504,8 @@ impl fmt::Display for ReasoningMode {
     }
 }
 
-#[derive(Clone, Debug, PartialEq)]
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+#[serde(default, deny_unknown_fields)]
 pub(crate) struct GenerationPreferences {
     pub(crate) context_tokens: i32,
     pub(crate) max_generated_tokens: i32,