Insight Fusion Agent 🧠

Insight Fusion is an advanced research-automation agent designed to synthesize deep insights from a hybrid of user-uploaded documents and real-time online sources. Built on LangChain and LangGraph, it employs an iterative, self-correcting workflow to ensure retrieval accuracy, claim validation, and evidence-grounded summarization.

🚀 Features

• Hybrid Context Retrieval: Seamlessly blends proprietary data (PDFs/Text) with Tavily/SerpAPI web search results.
• Iterative Refinement: Uses a cyclic LangGraph workflow to critique and refine search queries until sufficient evidence is gathered.
• Fact-Checking & Validation: Dedicated graph nodes effectively cross-reference claims against retrieved sources.
• Streamlit Interface: A clean, modern UI for interacting with the agent and visualizing the synthesis process.

🏗️ Architecture

The core of the agent is a state machine defined in src/agent.py. The workflow proceeds as follows:

1. Query Decomposition: Breaks down complex user requests into sub-questions.
2. Hybrid Search: Fetches documents from vector store and web.
3. Relevance Graduation: Evaluates retrieved documents for relevance.
4. Generation & Hallucination Check: Generates an answer and verifies it against the documents.

🛠️ Usage

1. Clone the repository.
2. Install dependencies:

   pip install -r requirements.txt

3. Set up your environment variables:

   cp .env.example .env
   # Add OPENAI_API_KEY, TAVILY_API_KEY etc.

4. Run the Streamlit app:

   streamlit run app.py

📦 Tech Stack

• Orchestration: LangChain, LangGraph
• LLM: GPT-4o / Claude 3.5 Sonnet
• Vector Store: ChromaDB / FAISS
• Frontend: Streamlit
• Search: Tavily API

📝 Roadmap

• Add multi-modal support (charts/images).
• Implement persistent memory/checkpointing with SQLite.
• Add export to PDF feature for research reports.