hf token and llm change

.env

@@ -1,2 +0,0 @@
-GNEWS_API_KEY=6c61f5da1b24fa83fbf964f8b280c438
-UPSTASH_REDIS_URL=rediss://:AU6-AAIjcDEyZDY4Njk3OGUwNzg0NTczODUxYmRmMDUyZDlmZWNiZXAxMA@clever-turtle-20158.upstash.io:6379

.gitignore

@@ -0,0 +1 @@
+.env 

question.py

@@ -1,21 +1,22 @@
 # app/routes/question.py
 import os
 import requests
-from fastapi import APIRouter, Query
+from fastapi import APIRouter
 from pydantic import BaseModel
 from typing import List
 from redis_client import redis_client as r
-from transformers import pipeline
-import re
 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()
 
-qa_model = pipeline("text2text-generation", model="google/flan-t5-base")  # replace with your preferred model
+client = InferenceClient("mistralai/Mistral-7B-Instruct", token=HF_TOKEN)
 
 class QuestionInput(BaseModel):
     question: str
@@ -42,13 +43,14 @@ async def ask_question(input: QuestionInput):
     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"Context:\n{context}\n\nQuestion: {question}\nAnswer:"
-    result = qa_model(prompt, max_length=256, do_sample=False)[0]['generated_text']
+    prompt = f"<s>[INST] Use the context below to answer the question. If the context is insufficient, say 'I don't know'.\n\nContext:\n{context}\n\nQuestion: {question} [/INST]"
+
+    result = client.text_generation(prompt, max_new_tokens=256, temperature=0.7)
 
     return {
         "question": question,
@@ -57,4 +59,4 @@ async def ask_question(input: QuestionInput):
             {"title": a["title"], "url": a["url"]}
             for a in articles
         ]
-    }
+    }