Add agent context compaction

src/server/response.rs

@@ -165,6 +165,12 @@ fn completion_stream_response(
     send_sse_headers_with_cors(stream, request.cors).map_err(|error| (500, error))?;
     while let Ok(event) = active.events.recv() {
         match event {
+            GenerationEvent::Compacted(_) => {
+                return Err((
+                    500,
+                    "The model runtime returned an unexpected compaction event.".into(),
+                ));
+            }
             GenerationEvent::Chunk {
                 reasoning: false,
                 content,
@@ -737,6 +743,12 @@ fn stream_response_with_keepalive(
             Err(_) => return Ok(()),
         };
         match event {
+            GenerationEvent::Compacted(_) => {
+                return Err((
+                    500,
+                    "The model runtime returned an unexpected compaction event.".into(),
+                ));
+            }
             GenerationEvent::Loading => {}
             GenerationEvent::Context {
                 tokens_per_second, ..