ffreemt
commited on
Commit
·
2493442
1
Parent(s):
d9591bc
Update loguru
Browse files- app.py +128 -65
- requirements.txt +2 -1
app.py
CHANGED
@@ -1,67 +1,102 @@
|
|
|
|
|
|
1 |
import json
|
|
|
|
|
|
|
2 |
import time
|
3 |
-
import requests
|
4 |
-
from flask import Flask, request, Response, stream_with_context
|
5 |
-
from bs4 import BeautifulSoup
|
6 |
from datetime import datetime, timedelta
|
7 |
-
import logging
|
8 |
-
import os
|
9 |
|
|
|
|
|
|
|
|
|
|
|
10 |
from ycecream import y
|
|
|
11 |
y.configure(sln=1)
|
12 |
|
13 |
app = Flask(__name__)
|
14 |
|
15 |
APP_SESSION_VALIDITY = timedelta(days=3)
|
16 |
ACCESS_TOKEN_VALIDITY = timedelta(hours=1)
|
17 |
-
USERNAME = os.environ.get(
|
18 |
-
PASSWORD = os.environ.get(
|
19 |
-
AUTHKEY = os.environ.get(
|
20 |
|
21 |
y(USERNAME[:3], PASSWORD[:2], AUTHKEY[:2])
|
22 |
|
23 |
-
cache = {
|
|
|
|
|
|
|
|
|
|
|
24 |
|
25 |
y(cache)
|
26 |
|
27 |
# 配置日志记录
|
28 |
-
logging.basicConfig(level=logging.DEBUG)
|
|
|
29 |
|
30 |
def fetch_tokens():
|
31 |
session = requests.Session()
|
32 |
|
33 |
# 检查并获取 appSession
|
34 |
-
if
|
35 |
-
|
36 |
-
|
|
|
|
|
|
|
|
|
|
|
37 |
if login_page_response.status_code != 200:
|
38 |
-
|
39 |
return None
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
46 |
cache["app_session_time"] = datetime.now()
|
47 |
|
48 |
# 检查并获取 accessToken
|
49 |
-
if
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
|
|
|
|
54 |
if response.status_code != 200:
|
55 |
-
|
56 |
return None
|
57 |
-
cache["access_token"] = response.json().get(
|
58 |
cache["access_token_time"] = datetime.now()
|
59 |
|
60 |
y(cache)
|
61 |
|
62 |
return cache["access_token"]
|
63 |
|
64 |
-
|
|
|
65 |
def landing():
|
66 |
return """
|
67 |
query /hf/v1/chat/completions for a spin, e.g. curl -XPOST 127.0.0.1:7860/hf/v1/chat/completions -H "Authorization: Bearer Your_AUTHKEY"
|
@@ -71,33 +106,48 @@ curl -XPOST [127.0.0.1:7860|hf space url]/hf/v1/chat/completions -H "Authorizati
|
|
71 |
|
72 |
"""
|
73 |
|
74 |
-
|
|
|
75 |
def chat_completions():
|
76 |
if request.method == "OPTIONS":
|
77 |
-
return Response(
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
|
86 |
access_token = fetch_tokens()
|
87 |
if not access_token:
|
88 |
-
|
89 |
return Response("Failed to obtain access token.", status=500)
|
90 |
|
91 |
try:
|
92 |
request_body = request.json
|
93 |
except Exception as e:
|
94 |
-
|
95 |
return Response("Error parsing JSON body", status=400)
|
96 |
|
97 |
messages = request_body.get("messages", [])
|
98 |
model = request_body.get("model", "reka-core")
|
99 |
|
100 |
-
conversation_history = [
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
|
102 |
if conversation_history and conversation_history[0]["type"] != "human":
|
103 |
conversation_history.insert(0, {"type": "human", "text": ""})
|
@@ -107,7 +157,15 @@ def chat_completions():
|
|
107 |
i = 0
|
108 |
while i < len(conversation_history) - 1:
|
109 |
if conversation_history[i]["type"] == conversation_history[i + 1]["type"]:
|
110 |
-
conversation_history.insert(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
111 |
i += 1
|
112 |
|
113 |
new_request_body = {
|
@@ -116,21 +174,22 @@ def chat_completions():
|
|
116 |
"use_search_engine": False,
|
117 |
"use_code_interpreter": False,
|
118 |
"model_name": "reka-core",
|
119 |
-
"random_seed": int(time.time())
|
120 |
}
|
121 |
|
122 |
response = requests.post(
|
123 |
"https://chat.reka.ai/api/chat",
|
124 |
headers={
|
125 |
"authorization": f"bearer {access_token}",
|
126 |
-
"content-type": "application/json"
|
127 |
},
|
128 |
data=json.dumps(new_request_body),
|
129 |
-
stream=True
|
|
|
130 |
)
|
131 |
|
132 |
if response.status_code != 200:
|
133 |
-
|
134 |
return Response(response.text, status=response.status_code)
|
135 |
|
136 |
created = int(time.time())
|
@@ -145,11 +204,11 @@ def chat_completions():
|
|
145 |
|
146 |
for line in response.iter_lines():
|
147 |
if line:
|
148 |
-
content_buffer += line.decode(
|
149 |
while "\n" in content_buffer:
|
150 |
newline_index = content_buffer.index("\n")
|
151 |
line = content_buffer[:newline_index]
|
152 |
-
content_buffer = content_buffer[newline_index + 1:]
|
153 |
|
154 |
if not line.startswith("data:"):
|
155 |
continue
|
@@ -164,25 +223,29 @@ def chat_completions():
|
|
164 |
last_four_texts.pop(0)
|
165 |
|
166 |
if len(last_four_texts) == 4 and (
|
167 |
-
|
168 |
-
|
169 |
-
|
|
|
170 |
break
|
171 |
|
172 |
full_content = data["text"]
|
173 |
-
new_content = full_content[len(prev_content):]
|
174 |
prev_content = full_content
|
175 |
|
176 |
formatted_data = {
|
177 |
-
"id": "chatcmpl-"
|
|
|
178 |
"object": "chat.completion.chunk",
|
179 |
"created": created,
|
180 |
"model": model,
|
181 |
-
"choices": [
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
|
|
|
|
|
186 |
}
|
187 |
yield f"data: {encoder.encode(formatted_data)}\n\n"
|
188 |
|
@@ -191,16 +254,16 @@ def chat_completions():
|
|
191 |
"object": "chat.completion.chunk",
|
192 |
"created": created,
|
193 |
"model": model,
|
194 |
-
"choices": [{
|
195 |
-
"index": 0,
|
196 |
-
"delta": {},
|
197 |
-
"finish_reason": "stop"
|
198 |
-
}]
|
199 |
}
|
200 |
yield f"data: {json.dumps(done_data)}\n\n"
|
201 |
yield "data: [DONE]\n\n"
|
202 |
|
203 |
-
return Response(
|
|
|
|
|
|
|
|
|
204 |
|
205 |
-
if __name__ ==
|
206 |
-
app.run(host=
|
|
|
1 |
+
# pylint: disable=line-too-long, missing-module-docstring, missing-function-docstring,broad-exception-caught, too-many-statements
|
2 |
+
|
3 |
import json
|
4 |
+
|
5 |
+
import os
|
6 |
+
import re
|
7 |
import time
|
|
|
|
|
|
|
8 |
from datetime import datetime, timedelta
|
|
|
|
|
9 |
|
10 |
+
import requests
|
11 |
+
|
12 |
+
# from bs4 import BeautifulSoup
|
13 |
+
from flask import Flask, Response, request, stream_with_context
|
14 |
+
from loguru import logger
|
15 |
from ycecream import y
|
16 |
+
|
17 |
y.configure(sln=1)
|
18 |
|
19 |
app = Flask(__name__)
|
20 |
|
21 |
APP_SESSION_VALIDITY = timedelta(days=3)
|
22 |
ACCESS_TOKEN_VALIDITY = timedelta(hours=1)
|
23 |
+
USERNAME = os.environ.get("USERNAME", "")
|
24 |
+
PASSWORD = os.environ.get("PASSWORD", "")
|
25 |
+
AUTHKEY = os.environ.get("AUTHKEY", "")
|
26 |
|
27 |
y(USERNAME[:3], PASSWORD[:2], AUTHKEY[:2])
|
28 |
|
29 |
+
cache = {
|
30 |
+
"app_session": None,
|
31 |
+
"app_session_time": None,
|
32 |
+
"access_token": None,
|
33 |
+
"access_token_time": None,
|
34 |
+
}
|
35 |
|
36 |
y(cache)
|
37 |
|
38 |
# 配置日志记录
|
39 |
+
# logging.basicConfig(level=logging.DEBUG)
|
40 |
+
|
41 |
|
42 |
def fetch_tokens():
|
43 |
session = requests.Session()
|
44 |
|
45 |
# 检查并获取 appSession
|
46 |
+
if (
|
47 |
+
not cache["app_session_time"]
|
48 |
+
or datetime.now() - cache["app_session_time"] >= APP_SESSION_VALIDITY
|
49 |
+
):
|
50 |
+
logger.info("Fetching new appSession")
|
51 |
+
login_page_response = session.get(
|
52 |
+
"https://chat.reka.ai/bff/auth/login", allow_redirects=True
|
53 |
+
)
|
54 |
if login_page_response.status_code != 200:
|
55 |
+
logger.error("Failed to load login page")
|
56 |
return None
|
57 |
+
|
58 |
+
# soup = BeautifulSoup(login_page_response.text, "html.parser")
|
59 |
+
# state_value = soup.find("input", {"name": "state"})["value"]
|
60 |
+
|
61 |
+
state_value = ""
|
62 |
+
_ = re.search(r"\w{20,}", login_page_response.text)
|
63 |
+
if _:
|
64 |
+
state_value = _.group()
|
65 |
+
|
66 |
+
session.post(
|
67 |
+
"https://auth.reka.ai/u/login",
|
68 |
+
data={
|
69 |
+
"state": state_value,
|
70 |
+
"username": USERNAME,
|
71 |
+
"password": PASSWORD,
|
72 |
+
"action": "default",
|
73 |
+
},
|
74 |
+
)
|
75 |
+
cache["app_session"] = session.cookies.get("appSession")
|
76 |
cache["app_session_time"] = datetime.now()
|
77 |
|
78 |
# 检查并获取 accessToken
|
79 |
+
if (
|
80 |
+
not cache["access_token_time"]
|
81 |
+
or datetime.now() - cache["access_token_time"] >= ACCESS_TOKEN_VALIDITY
|
82 |
+
):
|
83 |
+
logger.info("Fetching new accessToken")
|
84 |
+
response = session.get(
|
85 |
+
"https://chat.reka.ai/bff/auth/access_token",
|
86 |
+
headers={"Cookie": f'appSession={cache["app_session"]}'},
|
87 |
+
)
|
88 |
if response.status_code != 200:
|
89 |
+
logger.error("Failed to get access token")
|
90 |
return None
|
91 |
+
cache["access_token"] = response.json().get("accessToken")
|
92 |
cache["access_token_time"] = datetime.now()
|
93 |
|
94 |
y(cache)
|
95 |
|
96 |
return cache["access_token"]
|
97 |
|
98 |
+
|
99 |
+
@app.route("/")
|
100 |
def landing():
|
101 |
return """
|
102 |
query /hf/v1/chat/completions for a spin, e.g. curl -XPOST 127.0.0.1:7860/hf/v1/chat/completions -H "Authorization: Bearer Your_AUTHKEY"
|
|
|
106 |
|
107 |
"""
|
108 |
|
109 |
+
|
110 |
+
@app.route("/hf/v1/chat/completions", methods=["POST", "OPTIONS"])
|
111 |
def chat_completions():
|
112 |
if request.method == "OPTIONS":
|
113 |
+
return Response(
|
114 |
+
"",
|
115 |
+
status=204,
|
116 |
+
headers={
|
117 |
+
"Access-Control-Allow-Origin": "*",
|
118 |
+
"Access-Control-Allow-Headers": "*",
|
119 |
+
},
|
120 |
+
)
|
121 |
+
|
122 |
+
if (
|
123 |
+
request.method != "POST"
|
124 |
+
or request.path != "/hf/v1/chat/completions"
|
125 |
+
or request.headers.get("Authorization") != f"Bearer {AUTHKEY}"
|
126 |
+
):
|
127 |
+
logger.error("Unauthorized access attempt")
|
128 |
+
return Response("Unauthorized", status=401)
|
129 |
|
130 |
access_token = fetch_tokens()
|
131 |
if not access_token:
|
132 |
+
logger.error("Failed to obtain access token")
|
133 |
return Response("Failed to obtain access token.", status=500)
|
134 |
|
135 |
try:
|
136 |
request_body = request.json
|
137 |
except Exception as e:
|
138 |
+
logger.error(f"Error parsing JSON body: {e}")
|
139 |
return Response("Error parsing JSON body", status=400)
|
140 |
|
141 |
messages = request_body.get("messages", [])
|
142 |
model = request_body.get("model", "reka-core")
|
143 |
|
144 |
+
conversation_history = [
|
145 |
+
{
|
146 |
+
"type": "human" if msg["role"] in ["user", "system"] else "model",
|
147 |
+
"text": msg["content"],
|
148 |
+
}
|
149 |
+
for msg in messages
|
150 |
+
]
|
151 |
|
152 |
if conversation_history and conversation_history[0]["type"] != "human":
|
153 |
conversation_history.insert(0, {"type": "human", "text": ""})
|
|
|
157 |
i = 0
|
158 |
while i < len(conversation_history) - 1:
|
159 |
if conversation_history[i]["type"] == conversation_history[i + 1]["type"]:
|
160 |
+
conversation_history.insert(
|
161 |
+
i + 1,
|
162 |
+
{
|
163 |
+
"type": "model"
|
164 |
+
if conversation_history[i]["type"] == "human"
|
165 |
+
else "human",
|
166 |
+
"text": "",
|
167 |
+
},
|
168 |
+
)
|
169 |
i += 1
|
170 |
|
171 |
new_request_body = {
|
|
|
174 |
"use_search_engine": False,
|
175 |
"use_code_interpreter": False,
|
176 |
"model_name": "reka-core",
|
177 |
+
"random_seed": int(time.time()),
|
178 |
}
|
179 |
|
180 |
response = requests.post(
|
181 |
"https://chat.reka.ai/api/chat",
|
182 |
headers={
|
183 |
"authorization": f"bearer {access_token}",
|
184 |
+
"content-type": "application/json",
|
185 |
},
|
186 |
data=json.dumps(new_request_body),
|
187 |
+
stream=True,
|
188 |
+
timeout=300,
|
189 |
)
|
190 |
|
191 |
if response.status_code != 200:
|
192 |
+
logger.error(f"Error from external API: {response.status_code} {response.text}")
|
193 |
return Response(response.text, status=response.status_code)
|
194 |
|
195 |
created = int(time.time())
|
|
|
204 |
|
205 |
for line in response.iter_lines():
|
206 |
if line:
|
207 |
+
content_buffer += line.decode("utf-8") + "\n"
|
208 |
while "\n" in content_buffer:
|
209 |
newline_index = content_buffer.index("\n")
|
210 |
line = content_buffer[:newline_index]
|
211 |
+
content_buffer = content_buffer[newline_index + 1 :]
|
212 |
|
213 |
if not line.startswith("data:"):
|
214 |
continue
|
|
|
223 |
last_four_texts.pop(0)
|
224 |
|
225 |
if len(last_four_texts) == 4 and (
|
226 |
+
len(last_four_texts[3]) < len(last_four_texts[2])
|
227 |
+
or last_four_texts[3].endswith("<sep")
|
228 |
+
or last_four_texts[3].endswith("<")
|
229 |
+
):
|
230 |
break
|
231 |
|
232 |
full_content = data["text"]
|
233 |
+
new_content = full_content[len(prev_content) :]
|
234 |
prev_content = full_content
|
235 |
|
236 |
formatted_data = {
|
237 |
+
"id": "chatcmpl-"
|
238 |
+
+ "".join([str(time.time()), str(hash(new_content))]),
|
239 |
"object": "chat.completion.chunk",
|
240 |
"created": created,
|
241 |
"model": model,
|
242 |
+
"choices": [
|
243 |
+
{
|
244 |
+
"index": 0,
|
245 |
+
"delta": {"content": new_content},
|
246 |
+
"finish_reason": None,
|
247 |
+
}
|
248 |
+
],
|
249 |
}
|
250 |
yield f"data: {encoder.encode(formatted_data)}\n\n"
|
251 |
|
|
|
254 |
"object": "chat.completion.chunk",
|
255 |
"created": created,
|
256 |
"model": model,
|
257 |
+
"choices": [{"index": 0, "delta": {}, "finish_reason": "stop"}],
|
|
|
|
|
|
|
|
|
258 |
}
|
259 |
yield f"data: {json.dumps(done_data)}\n\n"
|
260 |
yield "data: [DONE]\n\n"
|
261 |
|
262 |
+
return Response(
|
263 |
+
stream_with_context(generate_stream()),
|
264 |
+
headers={"Content-Type": "text/event-stream"},
|
265 |
+
)
|
266 |
+
|
267 |
|
268 |
+
if __name__ == "__main__":
|
269 |
+
app.run(host="0.0.0.0", port=7860)
|
requirements.txt
CHANGED
@@ -1,4 +1,5 @@
|
|
1 |
Flask==3.0.3
|
2 |
requests==2.31.0
|
3 |
beautifulsoup4==4.12.3
|
4 |
-
ycecream
|
|
|
|
1 |
Flask==3.0.3
|
2 |
requests==2.31.0
|
3 |
beautifulsoup4==4.12.3
|
4 |
+
ycecream
|
5 |
+
loguru
|