Update app.py

app.py

@@ -6,11 +6,12 @@ from flask import Flask, request, Response, stream_with_context, jsonify
 
 app = Flask(__name__)
 
-# Config from Hugging Face Secrets
-ONYX_API_KEY = os.getenv("ONYX_SECRET")
+# CONFIGURATION: Set 'ONYX_API_KEY' in Hugging Face Settings > Variables and Secrets
+ONYX_API_KEY = os.getenv("ONYX_SECRET", "")
 ONYX_URL = "https://cloud.onyx.app/api/chat/send-chat-message"
 
 def transform_to_openai_chunk(content, model_name, finish_reason=None):
+    """Formats raw text into an OpenAI-compatible SSE chunk."""
     chunk = {
         "id": f"chatcmpl-{int(time.time())}",
         "object": "chat.completion.chunk",
@@ -18,7 +19,7 @@ def transform_to_openai_chunk(content, model_name, finish_reason=None):
         "model": model_name,
         "choices": [{
             "index": 0,
-            "delta": {"content": content} if content else {},
+            "delta": {"content": content} if content is not None else {},
             "finish_reason": finish_reason
         }]
     }
@@ -30,9 +31,10 @@ def chat_proxy():
     if not data or "messages" not in data:
         return jsonify({"error": "No messages provided"}), 400
 
-    # Handle "Provider / Model ID" with gaps and caps
+    # DYNAMIC PARSING: Handles "Google Vertex / Gemini 1.5 Pro" or "openai/gpt-4o"
     raw_model = data.get("model", "OpenAI / gpt-4o")
     if "/" in raw_model:
+        # Splits on first slash, cleans whitespace, preserves case
         provider, version = [part.strip() for part in raw_model.split("/", 1)]
     else:
         provider, version = "OpenAI", raw_model.strip()
@@ -40,6 +42,7 @@ def chat_proxy():
     messages = data.get("messages", [])
     user_content = messages[-1].get("content", "") if messages else ""
 
+    # ONYX PAYLOAD: Uses your specific template requirements
     onyx_payload = {
         "message": user_content,
         "llm_override": {
@@ -47,7 +50,7 @@ def chat_proxy():
             "model_version": version,
             "temperature": data.get("temperature", 0.7)
         },
-        "stream": True, # Forced ON
+        "stream": True,  # Forced ON regardless of user input
         "include_citations": True,
         "deep_research": False,
         "parent_message_id": -1,
@@ -62,9 +65,8 @@ def chat_proxy():
     def generate():
         try:
             with requests.post(ONYX_URL, json=onyx_payload, headers=headers, stream=True) as r:
-                # Basic error handling for the Onyx API response
                 if r.status_code != 200:
-                    yield f"data: {json.dumps({'error': 'Onyx API error', 'details': r.text})}\n\n"
+                    yield f"data: {json.dumps({'error': 'Onyx API Error', 'status': r.status_code})}\n\n"
                     return
 
                 for line in r.iter_lines():
@@ -73,11 +75,14 @@ def chat_proxy():
                     
                     try:
                         packet = json.loads(line.decode('utf-8'))
+                        # Onyx provides content in 'message_delta' packets
                         if packet.get("type") == "message_delta":
-                            yield transform_to_openai_chunk(packet.get("delta", ""), raw_model)
-                    except:
+                            content = packet.get("delta", "")
+                            yield transform_to_openai_chunk(content, raw_model)
+                    except (json.JSONDecodeError, KeyError):
                         continue
                 
+                # Signal stream completion
                 yield transform_to_openai_chunk(None, raw_model, finish_reason="stop")
                 yield "data: [DONE]\n\n"
         except Exception as e:
@@ -86,9 +91,9 @@ def chat_proxy():
     return Response(stream_with_context(generate()), mimetype='text/event-stream')
 
 @app.route('/')
-def health():
-    return jsonify({"status": "running", "port": 7860}), 200
+def health_check():
+    return "Onyx-OpenAI Proxy is running on HF Space (Port 7860).", 200
 
 if __name__ == '__main__':
-    # HF Spaces requires port 7860
+    # Standard HF port
     app.run(host='0.0.0.0', port=7860)