File size: 4,185 Bytes
69210b9 a092d54 67fbb52 e465159 69210b9 6716a7e f312f0d 1804706 0e7d7a3 69210b9 27120a6 6716a7e 69210b9 0e7d7a3 69210b9 6716a7e 69210b9 6716a7e 62a4bec 6716a7e 5c17060 6716a7e 27120a6 69210b9 27120a6 69210b9 27120a6 69210b9 27120a6 69210b9 62a4bec 27120a6 6716a7e 27120a6 62a4bec 27120a6 236d6c7 27120a6 71257bd 6716a7e 69210b9 6716a7e 69210b9 6716a7e 69210b9 6716a7e 67fbb52 e465159 69210b9 0e7d7a3 69210b9 0e7d7a3 69210b9 6716a7e 69210b9 |
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 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 |
import os
import sys
import json
import requests
import redis
from typing import List, Dict
from llama_index.core import VectorStoreIndex
from llama_index.core.query_engine import RetrieverQueryEngine
from llama_index.core.schema import Document
from llama_index.core.settings import Settings
# ✅ Disable implicit LLM usage
Settings.llm = None
# 🔐 Environment variables
REDIS_URL = os.environ.get("UPSTASH_REDIS_URL", "redis://localhost:6379")
REDIS_KEY = os.environ.get("UPSTASH_REDIS_TOKEN")
MISTRAL_URL = os.environ.get("MISTRAL_URL")
HF_TOKEN = os.environ.get("HF_TOKEN")
# ✅ Redis client
redis_client = redis.Redis.from_url(REDIS_URL, decode_responses=True)
# 📰 Topics
TOPICS = ["India news", "World news", "Tech news", "Finance news", "Sports news"]
# 📄 Headers for HF endpoint
HEADERS = {
"Authorization": f"Bearer {HF_TOKEN}",
"Content-Type": "application/json"
}
# 🧠 Build Mistral-style instruction prompt
def build_prompt(content: str, topic: str) -> str:
base_instruction = (
"You are Nuse’s official news summarizer — factual, concise, and engaging.\n"
"Summarize the following article in 25–30 words with 1–2 emojis.\n"
"The given content might contain multiple new items, so summarise each news item in 25-30 words and arranage them one line after the other starting them with a -"
"For example:"
" -India wins the biggest...."
" -The U.S trade tarrifs...."
" -Netanyahu agrees for a ceasefire...."
"Return only the summary."
)
tail = f"Topic: {topic}\n\n{content.strip()}"
return f"<s>[INST]{base_instruction}\n\n{tail}[/INST]</s>"
# 🔁 Call Mistral using HF Inference Endpoint
def call_mistral(prompt: str) -> Optional[str]:
headers = {
"Authorization": f"Bearer {HF_TOKEN}",
"Content-Type": "application/json"
}
payload = {
"inputs": prompt
}
try:
response = requests.post(MISTRAL_URL, headers=headers, json=payload, timeout=20)
response.raise_for_status()
data = response.json()
# Get the generated text
if isinstance(data, list) and data:
raw_output = data[0].get("generated_text", "")
elif isinstance(data, dict):
raw_output = data.get("generated_text", "")
else:
return None
# ✅ Extract only the portion after the [/INST]</s> marker
if "[/INST]</s>" in raw_output:
return raw_output.split("[/INST]</s>")[-1].strip()
return raw_output.strip()
except Exception as e:
print(f"⚠️ Mistral error: {e}")
return None
# ✂️ Summarize top N documents
def summarize_topic(docs: List[str], topic: str) -> List[Dict]:
feed = []
for doc in docs[:5]:
prompt = build_prompt(doc, topic)
print("\n📤 Prompt sent to Mistral:\n", prompt[:300], "...\n")
summary = call_mistral(prompt)
if summary:
feed.append({
"summary": summary,
"image_url": "https://source.unsplash.com/800x600/?news",
"article_link": "https://google.com/search?q=" + topic.replace(" ", "+")
})
return feed
# ⚡ Generate and cache daily feed
def generate_and_cache_daily_feed(documents: List[Document]):
index = VectorStoreIndex.from_documents(documents)
retriever = index.as_retriever()
query_engine = RetrieverQueryEngine(retriever=retriever)
final_feed = []
for topic in TOPICS:
print(f"\n🔍 Generating for: {topic}")
response = query_engine.query(topic)
docs = [str(node.get_content()) for node in response.source_nodes]
topic_feed = summarize_topic(docs, topic)
final_feed.append({
"topic": topic.lower().replace(" news", ""),
"feed": topic_feed
})
redis_client.set(REDIS_KEY, json.dumps(final_feed, ensure_ascii=False))
print(f"✅ Cached daily feed under key '{REDIS_KEY}'")
return final_feed
# 📦 For testing or API access
def get_cached_daily_feed():
cached = redis_client.get(REDIS_KEY)
return json.loads(cached) if cached else []
|