Agentic RAG–powered WhatsApp News Assistant

FlashFeedAgent is an Agentic RAG (Retrieval-Augmented Generation) system that delivers real-time news updates and contextual answers directly on WhatsApp.
It uses LangGraph for agent orchestration, FastAPI for backend services, FAISS for vector retrieval, and WhatsApp Cloud API (Meta) for messaging.

The system allows users to ask natural-language questions about current news and receive accurate, up-to-date responses via WhatsApp chat.

• WhatsApp-based interface (no app, no login)
• Agentic RAG pipeline using LangGraph
• Dynamic retrieval decision (agent decides when to fetch news)
• Web-scraped live news ingestion (refreshed every 30 mins)
• FAISS vector database for semantic search
• Per-user conversational memory using phone number as thread_id
• Webhook-driven real-time messaging
• FastAPI backend (production-ready)

User (WhatsApp)
     ↓
WhatsApp Cloud API (Meta)
    ↓  (Webhook POST)
FastAPI Backend (/webhook)
    ↓
LangGraph Agentic RAG
├── Retrieval Decision Agent
├── News Retrieval Agent (FAISS)
├── Answer Generation Agent (LLM)
└── Memory Update Agent
    ↓
WhatsApp Cloud API (Send Message)
    ↓
  User

Each agent is an independent node in a LangGraph state machine.

• Uses WhatsApp Cloud API (Meta)
• Messages are received via webhooks
• Replies are sent using the /messages API
• User phone number is used as:
  • Unique user identifier
  • LangGraph thread_id
  • WhatsApp reply destination

Follow-up questions are supported via conversational memory.

WhatsApp pushes incoming messages to your backend automatically — no polling required.

Environment variables required:

• GROQ_API_KEY=your_llm_api_key
• WHATSAPP_TOKEN=meta_access_token
• VERIFY_TOKEN=webhook_verify_token
• PHONE_NUMBER_ID=whatsapp_phone_number_id

• Backend: FastAPI
• Agent Framework: LangGraph
• LLM Provider: Groq (LLaMA)
• Vector Store: FAISS
• Embeddings: HuggingFace (gte-small)
• Messaging: WhatsApp Cloud API
• Scraping: BeautifulSoup + RecursiveUrlLoader
• Tunnel (Dev): ngrok

git clone https://github.com/a-anuj/flashFeedAgent.git
cd flashFeedAgent
pip install -r requirements.txt

uvicorn main:app --host 0.0.0.0 --port 8000

ngrok http 8000