Spaces:
Runtime error
Runtime error
from fastapi import FastAPI, Request, HTTPException | |
from fastapi.responses import JSONResponse | |
from pydantic import BaseModel | |
from langchain.document_loaders import WebBaseLoader | |
from langchain.chains.summarize import load_summarize_chain | |
#https://python.langchain.com/docs/use_cases/summarization | |
from langchain import HuggingFaceHub | |
from huggingface_hub import InferenceClient | |
from bs4 import BeautifulSoup | |
import requests | |
import sys | |
import os | |
from dotenv import load_dotenv | |
load_dotenv() | |
hf_token = os.environ.get('HUGGINGFACEHUB_API_TOKEN') | |
repo_id=os.environ.get('repo_id') | |
#port = os.getenv('port') | |
llm = HuggingFaceHub(repo_id=repo_id, # for StarChat | |
huggingfacehub_api_token=hf_token, #这个变量huggingfacehub_api_token名称似乎没有问题! | |
model_kwargs={"min_length": 512, # for StarChat | |
"max_new_tokens": 1024, "do_sample": True, # for StarChat | |
"temperature": 0.01, | |
"top_k": 50, | |
"top_p": 0.95, "eos_token_id": 49155}) | |
#chain = load_summarize_chain(llm, chain_type="stuff") | |
chain = load_summarize_chain(llm, chain_type="refine") | |
print(f"定义处理多余的Context文本的函数") | |
def remove_context(text): | |
# 检查 'Context:' 是否存在 | |
if 'Context:' in text: | |
# 找到第一个 '\n\n' 的位置 | |
end_of_context = text.find('\n\n') | |
# 删除 'Context:' 到第一个 '\n\n' 之间的部分 | |
return text[end_of_context + 2:] # '+2' 是为了跳过两个换行符 | |
else: | |
# 如果 'Context:' 不存在,返回原始文本 | |
return text | |
print(f"处理多余的Context文本函数定义结束") | |
app = FastAPI() | |
class ChatRequest(BaseModel): | |
target_url: str | |
#似乎必须要有这个class定义??? | |
#@app.post('/') | |
#async def home_api(request: Request): | |
# data = await request.json() | |
# target_url = data['target_url'] | |
# print(target_url) | |
# return {"Message": "FastAPI Home API Deploy Success on HF"} | |
async def chat(chat_request: ChatRequest): | |
#async def chat(request: Request, chat_request: ChatRequest): | |
target_url = chat_request.target_url | |
#@app.post('/api/chat') | |
#async def chat(request: Request): | |
# data = await request.json() | |
# target_url = data['target_url'] | |
#以上四行代码,可能有问题?出在data = await request.json()???理论上似乎也没有问题,因为后面的print(target_url),能够成功打印 | |
print(target_url) | |
try: | |
loader = WebBaseLoader(target_url) | |
print(target_url) | |
docs = loader.load() | |
result = chain.run(docs) | |
print(result) | |
print() | |
result=str(result) #找到之前总是POST Error的原因:chain.run(docs)的结果,格式不是str,导致{...}的json故障所致 | |
cleaned_initial_ai_response = remove_context(result) | |
print(cleaned_initial_ai_response) | |
print() | |
final_ai_response = cleaned_initial_ai_response.partition('¿Cuál es')[0].strip() | |
final_ai_response = final_ai_response.partition('¿Cuáles')[0].strip() | |
#final_ai_response = cleaned_initial_ai_response.split('¿cuál e')[0].strip().replace('\n\n', '\n').replace('<|end|>', '').replace('<|user|>', '').replace('<|system|>', '').replace('<|assistant|>', '') | |
print("final_ai_response after cleaning/splitting/partitioning '¿cuál e'") | |
print(final_ai_response) | |
print() | |
final_ai_response = final_ai_response.split('<|end|>')[0].strip().replace('\n\n', '\n').replace('<|end|>', '').replace('<|user|>', '').replace('<|system|>', '').replace('<|assistant|>', '') | |
new_final_ai_response = final_ai_response.split('Unhelpful Answer:')[0].strip() | |
final_result = new_final_ai_response.split('Note:')[0].strip() | |
#result="Hi FastAPI" #用这个调试OK | |
print("AI Summarization: " + final_result) | |
#return {'response': result} #FastAPI方式下,这个返回形式,有问题? - NO!不是的!可以工作! | |
return JSONResponse({'response': final_result}) #这个形式也是OK的 - 只要result是字符形式即可? | |
#对比FlaskAPI的:return jsonify({'response': result}),JSONResponse vs jsonify(都需要相应的from ... import ...) | |
except Exception as e: | |
err_msg = "Wrong URL or URL not parsable." | |
print(err_msg) | |
raise HTTPException(status_code=400, detail=err_msg) |