Complete long-context chat support

src/engine.rs

@@ -324,8 +324,10 @@ impl Generator {
         settings: &TurnSettings,
         cancelled: &AtomicBool,
         mut emit: impl FnMut(bool, String),
+        mut progress: impl FnMut(u32, u32),
     ) -> Result<(), String> {
         self.executor.reset()?;
+        progress(self.executor.position(), self.executor.context());
         let tokens = self.executor.model().render_conversation(
             &settings.system_prompt,
             messages,
@@ -337,17 +339,21 @@ impl Generator {
         let max_context = self.executor.context() as usize;
         if tokens.len() >= max_context {
             return Err(format!(
-                "the conversation uses {} tokens; the current Rust attention port supports fewer than {max_context}",
+                "the conversation uses {} tokens; the configured context holds fewer than {max_context}",
                 tokens.len()
             ));
         }
         let mut rng = Rng::new(settings.seed.unwrap_or(0x4453_3453_4552_5645));
         let mut reasoning = settings.reasoning_mode != ReasoningMode::Direct;
-        for token in tokens {
+        let prompt_tokens = tokens.len();
+        for (index, token) in tokens.into_iter().enumerate() {
             if cancelled.load(Ordering::Relaxed) {
                 return Ok(());
             }
             self.executor.eval(token)?;
+            if (index + 1).is_multiple_of(16) || index + 1 == prompt_tokens {
+                progress(self.executor.position(), self.executor.context());
+            }
         }
         for _ in 0..settings
             .max_generated_tokens
@@ -380,6 +386,7 @@ impl Generator {
                 emit(reasoning, String::from_utf8_lossy(&bytes).into_owned());
             }
             self.executor.eval(token)?;
+            progress(self.executor.position(), self.executor.context());
         }
         Ok(())
     }
@@ -497,7 +504,8 @@ mod sampling_tests {
             ReasoningMode::Direct,
         );
         assert_eq!(tokens.len(), 10);
-        let mut executor = metal::Executor::open(model, 128, false).unwrap();
+        let mut executor = metal::Executor::open(model, 32_768, false).unwrap();
+        assert_eq!(executor.context(), 32_768);
         for token in tokens {
             executor.eval(token).unwrap();
         }