Refactor completion request validation and enhance stream handling for real-time output

- Updated the completion request parsing logic to make the model field optional, allowing for better flexibility in handling requests.
- Enhanced the stream end handling by adding JSON parsing for stream end data and sending a structured message to the frontend.
- Modified the output writer to format data as Server-Sent Events (SSE) and ensured immediate flushing for real-time streaming in both CUI and OpenAI adapters.

agent/context/openapi.go

@@ -421,18 +421,16 @@ func parseCompletionRequestData(c *gin.Context) (*CompletionRequest, error) {
 			}
 
 			// If we got valid data from body, validate and return
-			if req.Model != "" && len(req.Messages) > 0 {
+			// Model is optional if assistant_id can be extracted later
+			if len(req.Messages) > 0 {
 				return &req, nil
 			}
 		}
 	}
 
 	// Fallback: Try to parse from query parameters
-	// Required fields
+	// Model is optional (can be extracted from assistant_id)
 	model := c.Query("model")
-	if model == "" {
-		return nil, fmt.Errorf("model field is required")
-	}
 	req.Model = model
 
 	// Messages (required, must be JSON string in query)

agent/llm/handlers/stream.go

@@ -232,6 +232,18 @@ func (s *streamState) handleGroupEnd(data []byte) int {
 
 // handleStreamEnd handles stream end event
 func (s *streamState) handleStreamEnd(data []byte) int {
+	// Parse the stream end data
+	var endData context.StreamEndData
+	if err := jsoniter.Unmarshal(data, &endData); err != nil {
+		log.Error("Failed to parse stream_end data: %v", err)
+		output.Flush(s.ctx)
+		return 0
+	}
+
+	// Send stream_end event as a message to frontend
+	msg := output.NewEventMessage("stream_end", "Stream completed", endData)
+	output.Send(s.ctx, msg)
+
 	// Flush any remaining data
 	output.Flush(s.ctx)
 	return 0 // Continue (stream will end naturally)

agent/output/adapters/cui/writer.go

@@ -99,14 +99,25 @@ func (w *Writer) sendChunk(chunk interface{}) error {
 		})
 	}
 
+	// Format as SSE (Server-Sent Events) format: "data: {json}\n\n"
+	sseData := []byte("data: ")
+	sseData = append(sseData, data...)
+	sseData = append(sseData, '\n', '\n')
+
 	// Send via context's writer
 	// The context knows how to send data based on the connection type (SSE, WebSocket, etc.)
-	if err := w.ctx.Send(data); err != nil {
+	if err := w.ctx.Send(sseData); err != nil {
 		if trace, _ := w.ctx.Trace(); trace != nil {
 			trace.Error(i18n.T(w.ctx.Locale, "output.cui.writer.send_error"), map[string]any{"error": err.Error()}) // "CUI Writer: Failed to send data to client"
 		}
 		return err
 	}
 
+	// Flush immediately to ensure real-time streaming
+	// Cast to http.ResponseWriter and call Flush if available
+	if flusher, ok := w.ctx.Writer.(interface{ Flush() }); ok {
+		flusher.Flush()
+	}
+
 	return nil
 }

agent/output/adapters/openai/writer.go

@@ -157,6 +157,12 @@ func (w *Writer) sendChunk(chunk interface{}) error {
 		return err
 	}
 
+	// Flush immediately to ensure real-time streaming
+	// Cast to http.ResponseWriter and call Flush if available
+	if flusher, ok := w.ctx.Writer.(interface{ Flush() }); ok {
+		flusher.Flush()
+	}
+
 	return nil
 }
 
@@ -176,5 +182,10 @@ func (w *Writer) sendDone() error {
 		return err
 	}
 
+	// Flush the final [DONE] message
+	if flusher, ok := w.ctx.Writer.(interface{ Flush() }); ok {
+		flusher.Flush()
+	}
+
 	return nil
 }