Update app.py

app.py

@@ -1,11 +1,11 @@
 import json
+import sseclient
 import random
-import asyncio
 import aiohttp
-from fastapi import FastAPI, Request, Response
-from fastapi.responses import StreamingResponse
+import asyncio
+from flask import Flask, request, Response, stream_with_context
 
-app = FastAPI()
+app = Flask(__name__)
 
 def generate_random_ip():
     return f"{random.randint(1, 255)}.{random.randint(0, 255)}.{random.randint(0, 255)}.{random.randint(0, 255)}"
@@ -43,29 +43,15 @@ def format_openai_response(content, finish_reason=None):
         }]
     }
 
-def sse_parser():
-    """Generator function to parse SSE messages."""
-    data = ''
-    while True:
-        line = yield
-        if line == '':
-            if data:
-                yield data
-                data = ''
-        elif line.startswith('data:'):
-            data += line[5:].strip()
-        else:
-            continue
-
-@app.post('/hf/v1/chat/completions')
-async def chat_completions(request: Request):
-    data = await request.json()
+@app.route('/hf/v1/chat/completions', methods=['POST'])
+async def chat_completions():
+    data = request.json
     messages = data.get('messages', [])
     stream = data.get('stream', False)
-    
+
     if not messages:
         return {"error": "No messages provided"}, 400
-    
+
     model = data.get('model', 'gpt-4o')
     if model.startswith('gpt'):
         endpoint = "openAI"
@@ -88,7 +74,7 @@ async def chat_completions(request: Request):
         while True:
             conversation = "\n".join([f"{msg['role']}: {msg['content']}" for msg in messages])
             conversation += "\nPlease follow and reply to the user’s recent messages and avoid answers that summarize the conversation history."
-            
+
             payload = {
                 "text": conversation,
                 "endpoint": endpoint,
@@ -96,65 +82,32 @@ async def chat_completions(request: Request):
             }
 
             async with aiohttp.ClientSession() as session:
-                async with session.post(original_api_url, headers=headers, json=payload) as resp:
-                    if resp.status != 200:
-                        yield f"data: {json.dumps({'error': 'Failed to connect to upstream server'})}\n\n"
-                        return
-
-                    parser = sse_parser()
-                    next(parser)  # Initialize the generator
-
-                    async for line in resp.content:
-                        line = line.decode('utf-8').strip()
-                        if line == '':
-                            continue
-
-                        parser.send(line)
-                        try:
-                            event_data = parser.send(None)
-                            if event_data:
-                                # Process the SSE event
-                                event_json = json.loads(event_data)
-                                if 'text' in event_json:
-                                    new_content = event_json['text'][len(full_response):]
-                                    full_response = event_json['text']
-                                    if new_content:
-                                        yield f"data: {json.dumps(format_openai_response(new_content))}\n\n"
-                                elif '"final":true' in event_data:
-                                    final_data = event_json
-                                    response_message = final_data.get('responseMessage', {})
-                                    finish_reason = response_message.get('finish_reason', 'stop')
-                                    if finish_reason == 'length':
-                                        messages.append({"role": "assistant", "content": full_response})
-                                        messages.append({"role": "user", "content": "Please continue your output and do not repeat the previous content"})
-                                        break  # Continue with the next request
-                                    else:
-                                        last_content = response_message.get('text', '')
-                                        if last_content and last_content != full_response:
-                                            yield f"data: {json.dumps(format_openai_response(last_content[len(full_response):]))}\n\n"
-                                        yield f"data: {json.dumps(format_openai_response('', finish_reason))}\n\n"
-                                        yield "data: [DONE]\n\n"
-                                        return
-                        except StopIteration:
-                            pass  # No complete event yet
-
-            yield f"data: {json.dumps(format_openai_response('', 'stop'))}\n\n"
-            yield "data: [DONE]\n\n"
-
+                async with session.post(original_api_url, headers=headers, json=payload) as response:
+                    async for line in response.content:
+                        if line.startswith(b'data: '):
+                            event_data = line.decode().lstrip('data: ')
+                            event = json.loads(event_data)
+                            if event.get("final"):
+                                yield f"data: {json.dumps(format_openai_response('', 'stop'))}\n\n"
+                                return
+                            new_content = event.get('text', '')[len(full_response):]
+                            if new_content:
+                                full_response = event.get('text', '')
+                                yield f"data: {json.dumps(format_openai_response(new_content))}\n\n"
+    
     if stream:
-        return StreamingResponse(generate(), media_type='text/event-stream')
+        return Response(stream_with_context(generate()), content_type='text/event-stream')
     else:
         full_response = ""
         finish_reason = "stop"
         async for chunk in generate():
-            if chunk.startswith("data: ") and not chunk.strip() == "data: [DONE]":
-                response_data = json.loads(chunk[6:])
-                if 'choices' in response_data and response_data['choices']:
-                    delta = response_data['choices'][0].get('delta', {})
-                    if 'content' in delta:
-                        full_response += delta['content']
-                    if 'finish_reason' in delta:
-                        finish_reason = delta['finish_reason']
+            response_data = json.loads(chunk[6:])
+            if 'choices' in response_data and response_data['choices']:
+                delta = response_data['choices'][0].get('delta', {})
+                if 'content' in delta:
+                    full_response += delta['content']
+                if 'finish_reason' in delta:
+                    finish_reason = delta['finish_reason']
 
         return {
             "id": "chatcmpl-123",
@@ -175,3 +128,6 @@ async def chat_completions(request: Request):
                 "total_tokens": 0
             }
         }
+
+if __name__ == '__main__':
+    app.run(debug=True, port=5000)