Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,137 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
import random
|
3 |
+
import string
|
4 |
+
import time
|
5 |
+
#from typing import Any
|
6 |
+
|
7 |
+
import g4f
|
8 |
+
from fastapi import FastAPI, Request
|
9 |
+
from fastapi.responses import StreamingResponse
|
10 |
+
|
11 |
+
from g4f import ChatCompletion
|
12 |
+
from loguru import logger
|
13 |
+
from starlette.middleware.cors import CORSMiddleware
|
14 |
+
|
15 |
+
import nest_asyncio
|
16 |
+
import os # Importo el módulo os para usar la variable de entorno
|
17 |
+
|
18 |
+
nest_asyncio.apply()
|
19 |
+
|
20 |
+
app = FastAPI()
|
21 |
+
|
22 |
+
app.add_middleware(
|
23 |
+
CORSMiddleware,
|
24 |
+
allow_origins=["*"],
|
25 |
+
allow_methods=["*"],
|
26 |
+
allow_headers=["*"],
|
27 |
+
)
|
28 |
+
|
29 |
+
|
30 |
+
@app.post("/chat/completions")
|
31 |
+
@app.post("/v1/chat/completions")
|
32 |
+
async def chat_completions(request: Request):
|
33 |
+
req_data = await request.json()
|
34 |
+
stream = req_data.get("stream", False)
|
35 |
+
model = req_data.get("model", "gpt-3.5-turbo")
|
36 |
+
messages = req_data.get("messages")
|
37 |
+
temperature = req_data.get("temperature", 1.0)
|
38 |
+
top_p = req_data.get("top_p", 1.0)
|
39 |
+
max_tokens = req_data.get("max_tokens", 0)
|
40 |
+
|
41 |
+
logger.info(
|
42 |
+
f"chat_completions: stream: {stream}, model: {model}, temperature: {temperature}, top_p: {top_p}, max_tokens: {max_tokens}"
|
43 |
+
)
|
44 |
+
|
45 |
+
response = await gen_resp(max_tokens, messages, model, stream, temperature, top_p)
|
46 |
+
|
47 |
+
completion_id = "".join(random.choices(string.ascii_letters + string.digits, k=28))
|
48 |
+
completion_timestamp = int(time.time())
|
49 |
+
|
50 |
+
if not stream:
|
51 |
+
logger.info(f"chat_completions: response: {response}")
|
52 |
+
return {
|
53 |
+
"id": f"chatcmpl-{completion_id}",
|
54 |
+
"object": "chat.completion",
|
55 |
+
"created": completion_timestamp,
|
56 |
+
"model": model,
|
57 |
+
"choices": [
|
58 |
+
{
|
59 |
+
"index": 0,
|
60 |
+
"message": {
|
61 |
+
"role": "assistant",
|
62 |
+
"content": response,
|
63 |
+
},
|
64 |
+
"finish_reason": "stop",
|
65 |
+
}
|
66 |
+
],
|
67 |
+
"usage": {
|
68 |
+
"prompt_tokens": None,
|
69 |
+
"completion_tokens": None,
|
70 |
+
"total_tokens": None,
|
71 |
+
},
|
72 |
+
}
|
73 |
+
|
74 |
+
def streaming():
|
75 |
+
for chunk in response:
|
76 |
+
completion_data = {
|
77 |
+
"id": f"chatcmpl-{completion_id}",
|
78 |
+
"object": "chat.completion.chunk",
|
79 |
+
"created": completion_timestamp,
|
80 |
+
"model": model,
|
81 |
+
"choices": [
|
82 |
+
{
|
83 |
+
"index": 0,
|
84 |
+
"delta": {
|
85 |
+
"content": chunk,
|
86 |
+
},
|
87 |
+
"finish_reason": None,
|
88 |
+
}
|
89 |
+
],
|
90 |
+
}
|
91 |
+
|
92 |
+
content = json.dumps(completion_data, separators=(",", ":"))
|
93 |
+
yield f"data: {content}\n\n"
|
94 |
+
time.sleep(0)
|
95 |
+
|
96 |
+
end_completion_data: dict[str, Any] = {
|
97 |
+
"id": f"chatcmpl-{completion_id}",
|
98 |
+
"object": "chat.completion.chunk",
|
99 |
+
"created": completion_timestamp,
|
100 |
+
"model": model,
|
101 |
+
"choices": [
|
102 |
+
{
|
103 |
+
"index": 0,
|
104 |
+
"delta": {},
|
105 |
+
"finish_reason": "stop",
|
106 |
+
}
|
107 |
+
],
|
108 |
+
}
|
109 |
+
content = json.dumps(end_completion_data, separators=(",", ":"))
|
110 |
+
yield f"data: {content}\n\n"
|
111 |
+
|
112 |
+
return StreamingResponse(streaming(), media_type="text/event-stream")
|
113 |
+
|
114 |
+
|
115 |
+
async def gen_resp(max_tokens, messages, model, stream, temperature, top_p):
|
116 |
+
# Obtengo el valor de MAX_ATTEMPTS desde la variable de entorno o uso un valor por defecto de 10
|
117 |
+
MAX_ATTEMPTS = int(os.getenv("MAX_ATTEMPTS", 10))
|
118 |
+
attempts = 0
|
119 |
+
while True:
|
120 |
+
try:
|
121 |
+
response = ChatCompletion.create(
|
122 |
+
model=model,
|
123 |
+
stream=stream,
|
124 |
+
messages=messages,
|
125 |
+
temperature=temperature,
|
126 |
+
top_p=top_p,
|
127 |
+
max_tokens=max_tokens,
|
128 |
+
system_prompt="",
|
129 |
+
provider=g4f.Provider.Bing,
|
130 |
+
)
|
131 |
+
return response
|
132 |
+
except Exception as e:
|
133 |
+
logger.error(f"gen_resp: Exception: {e}")
|
134 |
+
attempts += 1
|
135 |
+
if attempts >= MAX_ATTEMPTS:
|
136 |
+
return "Lo siento, no he podido generar una respuesta de chat. Por favor, revisa tu conexión a Internet y la configuración de la API y vuelve a intentarlo."
|
137 |
+
|