llm: set done reason at server level (#9830)

No functional change. Many different done reasons can be set at the runner
level, so rather than obsuring them we should return them to the server
process and let it choose what to do with the done reason. This separates
the API concerns from the runner.

llm/server.go

@@ -675,9 +675,32 @@ type CompletionRequest struct {
 	Grammar string // set before sending the request to the subprocess
 }
 
+// DoneReason represents the reason why a completion response is done
+type DoneReason int
+
+const (
+	// DoneReasonStop indicates the completion stopped naturally
+	DoneReasonStop DoneReason = iota
+	// DoneReasonLength indicates the completion stopped due to length limits
+	DoneReasonLength
+	// DoneReasonConnectionClosed indicates the completion stopped due to the connection being closed
+	DoneReasonConnectionClosed
+)
+
+func (d DoneReason) String() string {
+	switch d {
+	case DoneReasonLength:
+		return "length"
+	case DoneReasonStop:
+		return "stop"
+	default:
+		return "" // closed
+	}
+}
+
 type CompletionResponse struct {
 	Content            string        `json:"content"`
-	DoneReason         string        `json:"done_reason"`
+	DoneReason         DoneReason    `json:"done_reason"`
 	Done               bool          `json:"done"`
 	PromptEvalCount    int           `json:"prompt_eval_count"`
 	PromptEvalDuration time.Duration `json:"prompt_eval_duration"`
@@ -786,7 +809,6 @@ func (s *llmServer) Completion(ctx context.Context, req CompletionRequest, fn fu
 				continue
 			}
 
-			// slog.Debug("got line", "line", string(line))
 			evt, ok := bytes.CutPrefix(line, []byte("data: "))
 			if !ok {
 				evt = line