fix: use JSON encoded SSE chunks to preserve multi-line responses and prevent stream breaking on newline chunks

src/api/v1/chat.py

@@ -127,20 +127,21 @@ async def chat_stream(request: Request, payload: ChatRequest, uid: str = Depends
 
         gemini = GeminiService()
 
+        import json
         async def generate():
             full_response = ""
             try:
                 async for chunk in gemini.stream(payload.query, context=context, chat_history=chat_history):
                     full_response += chunk
                     # Format as SSE
-                    yield f"data: {chunk}\n\n"
+                    yield f"data: {json.dumps({'content': chunk, 'type': 'token'})}\n\n"
                 
                 # Save full response to memory after streaming completes
                 engine.memory.add_message("assistant", full_response)
                 yield "data: [DONE]\n\n"
             except Exception as e:
                 logger.error(f"Streaming error: {e}")
-                yield f"data: ⚠️ Streaming error: {e}\n\n"
+                yield f"data: {json.dumps({'content': f'\\n\\n⚠️ Streaming error: {e}', 'type': 'token'})}\n\n"
                 yield "data: [DONE]\n\n"
 
         return StreamingResponse(generate(), media_type="text/event-stream")

src/services/gemini_service.py

@@ -163,9 +163,32 @@ class GeminiService:
                 if chunk.text:
                     yield chunk.text
         except Exception as e:
-            logger.error(f"Streaming error: {e}")
-            yield f"\n\n⚠️ Streaming error: {e}"
-
+            logger.error(f"Primary model {self.primary_model} streaming failed: {e}")
+            logger.info(f"Falling back to {self.fallback_model}...")
+            try:
+                response_stream = client.models.generate_content_stream(
+                    model=self.fallback_model,
+                    contents=prompt,
+                    config=config
+                )
+                for chunk in response_stream:
+                    if chunk.text:
+                        yield chunk.text
+            except Exception as e2:
+                logger.error(f"Fallback model {self.fallback_model} streaming also failed: {e2}")
+                logger.info("Falling back to gemini-1.5-flash-8b...")
+                try:
+                    response_stream = client.models.generate_content_stream(
+                        model="gemini-1.5-flash-8b",
+                        contents=prompt,
+                        config=config
+                    )
+                    for chunk in response_stream:
+                        if chunk.text:
+                            yield chunk.text
+                except Exception as e3:
+                    logger.error(f"All Gemini streaming models failed: {e3}")
+                    yield f"\n\n⚠️ Streaming error: All Gemini models failed. You might be out of API quota."
 
 class GeminiVisionService:
     """Gemini Vision for extracting math from images."""