Agentic RAG Chatbot with Multi-Tool Integration

An Agentic Retrieval-Augmented Generation (RAG) Chatbot powered by an intelligent agent that can dynamically choose and orchestrate multiple tools (retrieval, web search, calculator, code execution, APIs, etc.) to answer complex user queries accurately and efficiently.

This project demonstrates agentic reasoning + multi-tool execution + RAG, making it highly relevant for real-world AI/ML and LLM system design.

---

✨ Key Features

• Agentic Decision-Making: The agent reasons step-by-step and decides which tool(s) to invoke

• Multi-Tool Integration:

  • Vector Retriever (RAG)
  • Web Search Tool
  • Calculator / Math Tool
  • Code / Python Tool
  • Custom API Tools (extensible)

• RAG Pipeline: Retrieves relevant document chunks before generation

• Tool Chaining: Agent can call multiple tools in sequence for a single query

• Conversation Memory: Maintains context across turns

• Hallucination Reduction: Answers are grounded in retrieved or tool-generated data

• Modular & Scalable Design

---

🧠 System Architecture

User Query
   ↓
Agent (Planner + Reasoner)
   ↓ decides
┌──────────────────────────────┐
│ Available Tools              │
│  • Vector Retriever (RAG)    │
│  • Web Search                │
│  • Calculator                │
│  • Python / Code Executor    │
│  • Custom APIs               │
└──────────────────────────────┘
   ↓
Tool Outputs / Retrieved Context
   ↓
LLM Generator
   ↓
Final Grounded Answer

---

🛠 Tech Stack

• Language: Python 3.10+
• LLM: OpenAI / Azure OpenAI / Local LLM
• Agent Framework: LangChain Agents / LlamaIndex Agents
• Embeddings: OpenAI / Sentence-Transformers
• Vector Store: FAISS / Chroma / Pinecone
• Backend: FastAPI
• Frontend (Optional): Streamlit / React

---

🧰 Integrated Tools

Tool	Purpose
Vector Retriever	Document-based question answering (RAG)
Web Search Tool	Real-time / external knowledge
Calculator Tool	Numerical & logical computations
Python Tool	Data processing, validation, transformations
Custom API Tool	Domain-specific integrations

The agent autonomously selects the appropriate tool(s) based on the query.

---

⚙️ Setup & Installation

1. Clone the repository
2. Create & activate a virtual environment
3. Install dependencies
4. Configure environment variables

---

🚀 Running the Application

Start API Server

uvicorn fastapp.main:app --reload

Optional UI

streamlit run frontend/frontend.py

---

🔮 Future Enhancements

• Feedback-based agent learning
• Docker & cloud deployment

---

📜 License

MIT License

---

👩‍💻 Author

Vaishnavi Barolia AI / ML Research Intern Evostra Ventures

---

This project highlights agentic reasoning + multi-tool orchestration, aligning strongly with modern LLM Engineer / AI Engineer roles.