File size: 1,974 Bytes
8e9fd43 2c42748 8e9fd43 2c42748 8e9fd43 2c42748 8e9fd43 c1ba890 8e9fd43 606b2ff 8e9fd43 c1ba890 8e9fd43 2812e86 bf0cf05 8e9fd43 bf0cf05 02f3cf9 8e9fd43 2c42748 8e9fd43 5da06ea 8e9fd43 7edbb26 2c42748 8e9fd43 2c42748 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 |
# app/routes/question.py
import os
import requests
from fastapi import APIRouter
from pydantic import BaseModel
from typing import List
from redis_client import redis_client as r
from dotenv import load_dotenv
from huggingface_hub import InferenceClient
import re
load_dotenv()
GNEWS_API_KEY = os.getenv("GNEWS_API_KEY")
HF_TOKEN = os.getenv("HF_TOKEN") # Hugging Face token for private models if needed
askMe = APIRouter()
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3", token=os.getenv("HF_TOKEN"))
class QuestionInput(BaseModel):
question: str
@askMe.post("/ask")
async def ask_question(input: QuestionInput):
question = input.question
# Extract keywords (simple version)
keywords = re.findall(r"\b\w{4,}\b", question)
query_string = " AND ".join(f'"{kw}"' for kw in keywords[:7])
print("Keywords are", query_string)
gnews_url = f"https://gnews.io/api/v4/search?q={query_string}&lang=en&max=3&expand=content&token={GNEWS_API_KEY}"
try:
response = requests.get(gnews_url, timeout=10)
response.raise_for_status()
articles = response.json().get("articles", [])
except Exception as e:
return {"error": f"GNews API error: {str(e)}"}
# Combine article content for context
context = "\n\n".join([
article.get("content") or article.get("description") or ""
for article in articles
])[:1500] # truncate to keep model input size safe
print("And context is", context)
# Build prompt
prompt = f"<s>[INST] Use the context below to answer the question. If the context is insufficient, say 'I am unable to answer'.\n\nContext:\n{context}\n\nQuestion: {question} [/INST]"
result = client.text_generation(prompt, max_new_tokens=256, temperature=0.7)
return {
"question": question,
"answer": result.strip(),
"sources": [
{"title": a["title"], "url": a["url"]}
for a in articles
]
}
|