Spaces:
Runtime error
Runtime error
from fastapi import FastAPI, Request, Body, HTTPException, Depends | |
from fastapi.security import APIKeyHeader | |
from typing import Optional | |
from huggingface_hub import InferenceClient | |
import random | |
API_URL = "https://api-inference.huggingface.co/models/" | |
API_KEY = "abcd12345" # Replace with your actual API key | |
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.1") | |
app = FastAPI() | |
security = APIKeyHeader(name="api_key", auto_error=False) | |
def get_api_key(api_key: Optional[str] = Depends(security)): | |
if api_key is None or api_key != API_KEY: | |
raise HTTPException(status_code=401, detail="Unauthorized access") | |
return api_key | |
def format_prompt(message, history): | |
prompt = "<s>" | |
for user_prompt, bot_response in history: | |
prompt += f"[INST] {user_prompt} [/INST]" | |
prompt += f" {bot_response}</s> " | |
prompt += f"[INST] {message} [/INST]" | |
return prompt | |
def generate_text( | |
request: Request, | |
body: dict = Body(...), | |
api_key: str = Depends(get_api_key) | |
): | |
prompt = body.get("prompt", "") | |
sys_prompt = body.get("sysPrompt", "") | |
temperature = body.get("temperature", 0.5) | |
top_p = body.get("top_p", 0.95) | |
repetition_penalty = body.get("repetition_penalty", 1.0) | |
history = [] # You might need to handle this based on your actual usage | |
formatted_prompt = format_prompt(prompt, history) | |
stream = client.text_generation( | |
formatted_prompt, | |
temperature=temperature, | |
max_new_tokens=512, | |
top_p=top_p, | |
repetition_penalty=repetition_penalty, | |
do_sample=True, | |
seed=random.randint(0, 10**7), | |
stream=False, | |
details=False, | |
return_full_text=False | |
) | |
return stream | |