Update app.py
Browse files
app.py
CHANGED
@@ -1,60 +1,37 @@
|
|
|
|
1 |
from flask import Flask, request, Response
|
2 |
import requests
|
3 |
import json
|
4 |
-
import
|
5 |
-
import
|
6 |
-
import logging
|
7 |
-
from helper import create_jwt
|
8 |
|
9 |
app = Flask(__name__)
|
10 |
|
11 |
-
# 配置日志
|
12 |
-
logging.basicConfig(
|
13 |
-
level=logging.INFO,
|
14 |
-
format='%(asctime)s [%(levelname)s] %(message)s',
|
15 |
-
handlers=[
|
16 |
-
logging.StreamHandler(sys.stdout)
|
17 |
-
]
|
18 |
-
)
|
19 |
-
logger = logging.getLogger(__name__)
|
20 |
-
|
21 |
@app.route('/chat/completions', methods=['POST'])
|
22 |
-
def chat():
|
23 |
"""
|
24 |
Handle chat completion requests.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
"""
|
26 |
-
logger.info("=== Received chat completion request ===")
|
27 |
-
|
28 |
-
# 记录请求信息
|
29 |
-
logger.info(f"Request method: {request.method}")
|
30 |
-
logger.info(f"Request URL: {request.url}")
|
31 |
-
logger.info(f"Request headers: {json.dumps(dict(request.headers), indent=2)}")
|
32 |
-
|
33 |
# Get the payload from the request
|
34 |
payload = request.json
|
35 |
-
logger.info(f"Request payload: {json.dumps(payload, indent=2)}")
|
36 |
|
37 |
# Get the model from the payload, defaulting to "claude-3-5-sonnet-20240620"
|
38 |
model = payload.get('model', 'claude-3-5-sonnet-20240620')
|
39 |
-
logger.info(f"Using model: {model}")
|
40 |
-
|
41 |
-
# Extract GitHub username and Zed user ID from Authorization header
|
42 |
-
auth_header = request.headers.get('Authorization')
|
43 |
-
if not auth_header or not auth_header.startswith('Bearer '):
|
44 |
-
logger.error("Invalid Authorization header")
|
45 |
-
return Response('Invalid Authorization header', status=401)
|
46 |
-
|
47 |
-
try:
|
48 |
-
github_username, zed_user_id = auth_header[7:].split(',')
|
49 |
-
logger.info(f"Extracted GitHub username: {github_username}")
|
50 |
-
logger.info(f"Extracted Zed user ID: {zed_user_id}")
|
51 |
-
except ValueError:
|
52 |
-
logger.error("Invalid Authorization header format")
|
53 |
-
return Response('Invalid Authorization header format', status=401)
|
54 |
|
55 |
# Prepare the request for the LLM API
|
56 |
-
url = "https://llm.zed.dev/completion"
|
57 |
-
logger.info(f"LLM API URL: {url}")
|
58 |
|
59 |
llm_payload = {
|
60 |
"provider": "anthropic",
|
@@ -68,10 +45,9 @@ def chat():
|
|
68 |
"system": ""
|
69 |
}
|
70 |
}
|
71 |
-
logger.info(f"LLM API payload: {json.dumps(llm_payload, indent=2)}")
|
72 |
|
73 |
-
|
74 |
-
|
75 |
|
76 |
headers = {
|
77 |
'Host': 'llm.zed.dev',
|
@@ -80,40 +56,22 @@ def chat():
|
|
80 |
'authorization': f'Bearer {jwt}',
|
81 |
'user-agent': 'Zed/0.149.3 (macos; aarch64)'
|
82 |
}
|
83 |
-
logger.info(f"Request headers for LLM API: {json.dumps(headers, indent=2)}")
|
84 |
|
85 |
# Get proxy from environment variable
|
86 |
proxy = os.environ.get('HTTP_PROXY', None)
|
87 |
proxies = {'http': proxy, 'https': proxy} if proxy else None
|
88 |
-
logger.info(f"Using proxy: {proxy}")
|
89 |
|
90 |
-
def generate():
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
logger.info(f"LLM API response headers: {json.dumps(dict(response.headers), indent=2)}")
|
96 |
-
|
97 |
-
if response.status_code == 301:
|
98 |
-
new_location = response.headers.get('Location')
|
99 |
-
logger.warning(f"Received 301 redirect. New location: {new_location}")
|
100 |
-
# 如果需要,可以在这里处理重定向
|
101 |
-
|
102 |
-
for chunk in response.iter_content(chunk_size=1024):
|
103 |
-
if chunk:
|
104 |
-
logger.debug(f"Received chunk of size: {len(chunk)} bytes")
|
105 |
-
yield chunk
|
106 |
-
except Exception as e:
|
107 |
-
logger.error(f"Error during API request: {str(e)}")
|
108 |
-
yield str(e).encode()
|
109 |
|
110 |
-
logger.info("Returning streaming response")
|
111 |
return Response(generate(), content_type='application/octet-stream')
|
112 |
|
113 |
-
|
114 |
-
|
115 |
-
logger.info("Received request to home page")
|
116 |
-
return "Welcome to the Chat Completion API", 200
|
117 |
|
118 |
if __name__ == '__main__':
|
119 |
-
|
|
|
|
1 |
+
from asgiref.wsgi import WsgiToAsgi
|
2 |
from flask import Flask, request, Response
|
3 |
import requests
|
4 |
import json
|
5 |
+
import random
|
6 |
+
from helper import create_jwt, get_github_username_zed_userid_list
|
|
|
|
|
7 |
|
8 |
app = Flask(__name__)
|
9 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
@app.route('/chat/completions', methods=['POST'])
|
11 |
+
async def chat():
|
12 |
"""
|
13 |
Handle chat completion requests.
|
14 |
+
|
15 |
+
This function processes incoming POST requests to the '/chat/completions' endpoint.
|
16 |
+
It prepares the payload for the LLM API, generates a JWT for authentication,
|
17 |
+
and streams the response from the LLM API back to the client.
|
18 |
+
|
19 |
+
Returns:
|
20 |
+
Response: A streaming response containing the LLM API's output.
|
21 |
+
|
22 |
+
Note:
|
23 |
+
- The function uses environment variables for proxy configuration.
|
24 |
+
- It generates random GitHub username and Zed user ID for each request.
|
25 |
+
- The LLM model defaults to "claude-3-5-sonnet-20240620" if not specified.
|
26 |
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
# Get the payload from the request
|
28 |
payload = request.json
|
|
|
29 |
|
30 |
# Get the model from the payload, defaulting to "claude-3-5-sonnet-20240620"
|
31 |
model = payload.get('model', 'claude-3-5-sonnet-20240620')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
|
33 |
# Prepare the request for the LLM API
|
34 |
+
url = "https://llm.zed.dev/completion?"
|
|
|
35 |
|
36 |
llm_payload = {
|
37 |
"provider": "anthropic",
|
|
|
45 |
"system": ""
|
46 |
}
|
47 |
}
|
|
|
48 |
|
49 |
+
github_username, zed_user_id = get_github_username_zed_userid_list()
|
50 |
+
jwt = create_jwt(github_username, zed_user_id)
|
51 |
|
52 |
headers = {
|
53 |
'Host': 'llm.zed.dev',
|
|
|
56 |
'authorization': f'Bearer {jwt}',
|
57 |
'user-agent': 'Zed/0.149.3 (macos; aarch64)'
|
58 |
}
|
|
|
59 |
|
60 |
# Get proxy from environment variable
|
61 |
proxy = os.environ.get('HTTP_PROXY', None)
|
62 |
proxies = {'http': proxy, 'https': proxy} if proxy else None
|
|
|
63 |
|
64 |
+
async def generate():
|
65 |
+
with requests.post(url, headers=headers, json=llm_payload, stream=True, proxies=proxies) as response:
|
66 |
+
for chunk in response.iter_content(chunk_size=1024):
|
67 |
+
if chunk:
|
68 |
+
yield chunk
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
|
|
|
70 |
return Response(generate(), content_type='application/octet-stream')
|
71 |
|
72 |
+
# Convert the Flask app to an ASGI app
|
73 |
+
asgi_app = WsgiToAsgi(app)
|
|
|
|
|
74 |
|
75 |
if __name__ == '__main__':
|
76 |
+
import uvicorn
|
77 |
+
uvicorn.run(asgi_app, host="0.0.0.0", port=8000)
|