Update app.py

app.py

@@ -1,12 +1,50 @@
-from flask import Flask
+import requests
+from flask import Flask, request, Response, jsonify
 
 app = Flask(__name__)
 
-# Sample endpoint
+# Authentication headers (this could be made dynamic too)
+headers = {
+    'accept': 'application/json',
+    'content-type': 'application/json',
+    'cookie': 'session=eyJzZXNzaW9uX2lkIjogIjgyMGM2OTJmLTRhMzItNGVhMi05MTNhLTBjNzk1YTYwN2MyMSIsICJ1c2VyX2lkIjogIjA2ODkxMzVhLWRhZTUtNDk0OS05NTE3LWQ3MTJlMmZkZDU4YiJ9.ZykspA.CLUCSuyHoG8S1uruiVKNkHRjw5Q'
+}
+
+# Base URL for proxying the requests
+base_url = 'https://openui.fly.dev'
+
 @app.route('/')
 def hello():
     return 'Hello, World!'
 
+# /v1/models endpoint
+@app.route('/v1/models', methods=['GET'])
+def get_models():
+    url = f'{base_url}/v1/models'
+    # Forward the GET request to the target API
+    response = requests.get(url, headers=headers)
+    
+    # Return the JSON response from the target API
+    return jsonify(response.json())
+
+# /v1/chat/completions endpoint with streaming
+@app.route('/v1/chat/completions', methods=['POST'])
+def chat_completions():
+    # Extract request data (JSON payload)
+    payload = request.json
+    
+    # Forward the POST request to the target API with streaming responses
+    url = f'{base_url}/v1/chat/completions'
+    response = requests.post(url, headers=headers, json=payload, stream=True)
+    
+    # Streaming response back to the client
+    def stream_response():
+        for chunk in response.iter_content(chunk_size=1024):
+            if chunk:
+                yield chunk.decode('utf-8')
+    
+    return Response(stream_response(), content_type='application/json')
 
 if __name__ == '__main__':
-    app.run(debug=True)  # Run the Flask app in debug mode if executed directly
+    app.run(debug=True, host='0.0.0.0')  # To run it in docker container
+