🧠 FamAgent – Research Assistant Agent

FamAgent (Functional Assistant for Meaningful Research Agent) is an intelligent research assistant built using LangChain, and Models from Ollama, FAISS with Streamlit as frontend and a powerful suite of structured tools.

It helps you by:

• Searching the web for the latest information
• Retrieving concise summaries from Wikipedia
• Fetching academic papers from sources like arXiv
• Saving outputs efficiently for further use

All responses are delivered as clean, structured JSON, making integration and analysis seamless.

---

🚀 Features

• 🤖 LLM Agent powered by Langchain models (here : mistral) or any model can be served locally via Ollama
• 🔍 Multi-tool integration including:
  • Web search & Wikipedia lookup
  • Academic paper retrieval from arXiv
  • Automated Retrieval-Augmented Generation (Auto-RAG)
  • Saving and embedding outputs for later search
• 📄 Structured Output with Pydantic schemas
• ⚙️ Extensible architecture supporting custom tools

---

🧰 Tools

✅ Included Tools & Usage

Tool Name	Description	Example Commands
search_tool	Perform broad web searches to gather up-to-date information	"Search for latest AI breakthroughs"
wiki_tool	Retrieve concise, reliable summaries from Wikipedia	"Explain quantum computing from Wikipedia"
save_embed_tool	Save research notes as text files and embed them into a FAISS vector store for semantic search	(Automatically used internally when saving)
arxiv_tool	Search and retrieve academic papers from arXiv	"Find recent papers on reinforcement learning"
auto_rag_tool	Automatic Retrieval-Augmented Generation: combines retrieval and generation for in-depth answers	"Summarize recent advances in renewable energy"

---

---

⚙️ Setup Instructions

1. Clone the Repository

git clone https://github.com/your-username/famagent.git
cd famagent

2. Install Using install.sh

Run the installation script to setup your environment and dependencies:

./install.sh

OR: Install Editable Package with pip

python3 -m venv agents
source agents/bin/activate  # or `source agents/Scripts/activate` on Windows
pip install -e .

3. Configure Environment Variables

Create a .env file and add your API keys:

MISTRAL_API_KEY=your_api_key_here
COHERE_API_KEY=your_cohere_api_key

---

🛠️ Using Ollama for Local Model Serving

This project uses Ollama to serve the Mistral model locally for fast and private inference.

Steps to Setup Ollama

1. Install Ollama

Follow instructions at Ollama Install Docs

2. Download and Run the Mistral Model

ollama pull mistral/mistral-small

3. Run Ollama Server

Make sure the Ollama daemon is running locally.

4. Configure your .env

Set the model provider in your .env:

MODEL_PROVIDER=ollama
OLLAMA_MODEL_NAME=mistral/mistral-small

The code will connect to the Ollama model endpoint automatically.

---

💡 Running the Agent

streamlit run app.py

Open http://localhost:8501 in your browser to access the Streamlit UI.

---

💡 Example Output

Upon entering your query, you receive a structured JSON response like this:

{
  "topic": "Artificial Intelligence",
  "summary": "Artificial Intelligence (AI) refers to ...",
  "sources": ["https://en.wikipedia.org/wiki/Artificial_intelligence"],
  "tools_used": ["search_tool", "wiki_tool"],
}

---

📜 License

This project is licensed under the MIT License.