llm-fastapi-1 / main.py
huedaya's picture
Update main.py
27abc54 verified
from fastapi import FastAPI, Depends, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from huggingface_hub import InferenceClient
import os
from fastapi.security import OAuth2PasswordBearer
from typing import Annotated
oauth2_scheme = OAuth2PasswordBearer(tokenUrl="token")
app = FastAPI()
# Allow all CORS
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
# client = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407")
class Item(BaseModel):
prompt: str
history: list
system_prompt: str
temperature: float = 0.0
max_new_tokens: int = 2048
top_p: float = 0.15
repetition_penalty: float = 1.0
def format_prompt(message, history):
return message
def generate(item: Item):
temperature = float(item.temperature)
if temperature < 1e-2:
temperature = 1e-2
top_p = float(item.top_p)
generate_kwargs = dict(
temperature=temperature,
max_new_tokens=item.max_new_tokens,
top_p=top_p,
repetition_penalty=item.repetition_penalty,
do_sample=True,
seed=42,
)
formatted_prompt = format_prompt(f"{item.system_prompt}, {item.prompt}", item.history)
stream = client.text_generation(formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
output += response.token.text
return output
@app.get("/")
async def root():
return {"status": "ok"}
@app.head("/")
async def root():
return {"status": "ok"}
@app.post("/generate")
async def generate_text(item: Item, token: Annotated[str, Depends(oauth2_scheme)]):
# Reject if not authenticated
apiKey = os.environ.get("API_KEY")
if apiKey != token:
raise HTTPException(status_code=403, detail="Invalid API key")
return {
"response": generate(item)
}