LangChain Assistant

A small experimental project built while learning LangChain and tool integrations.
This app uses FastAPI, LangChain, and Next.js to demonstrate how LLMs can interact with external tools like SerpAPI (for web search), LLMMathChain (for math evaluation), and simple custom logic functions.

The goal of this project was to explore multi-tool reasoning and build a minimal interactive interface using an Ollama-served LLM backend.
It’s a learning-focused implementation rather than a production-level AI assistant.

---

Features

• LLM Agent (LangChain) with multi-tool capabilities
• Web search using SerpAPI
• Custom logic tools for reasoning tasks
• Memory-enabled conversation flow
• FastAPI backend + Next.js frontend
• Easily customizable and extendable

---

Tech Stack

• Backend: FastAPI, LangChain, DeepInfra / Ollama
• Frontend: Next.js (in research_agent/ directory)
• LLMs: Mistral, LLaMA 2, or any Ollama/DeepInfra-supported model
• Tools: SerpAPI, LLMMathChain, custom logic tools

---

🛠️ Setup & Installation

1. Clone the Repository

git clone https://github.com/Rakesh-46-VR/LangChainAssistant.git
cd AutoResearchAgent

2. Create Your Environment File

Copy the .env.example file and fill in the necessary values:

cp .env.example .env

Edit .env:

SERP_API_KEY="your-serpapi-key"
MODEL_ID="mistral"
OLLAMA_BASE_URL="http://127.0.0.1:11434"
ALLOWED_ORIGIN="http://localhost:3000"

---

3. Install Backend Dependencies

cd fastapi
pip install -r req.txt

---

4. Start the FastAPI Backend

uvicorn app:app --reload

By default, it runs at http://localhost:8000.

---

5. Start the Frontend (Next.js)

The frontend is located in the research_agent/ directory:

cd ../research_agent
pnpm install
pnpm dev

Frontend runs at http://localhost:3000

Make sure the ALLOWED_ORIGIN in your .env matches this URL.

---

API Usage

Ask a Question

Send a POST request to /ask endpoint:

curl -X POST http://localhost:8000/ask \
  -H "Content-Type: application/json" \
  -d '{"query": "Evaluate the expression 2+6-9*3+6?"}'

Response:

{
  "content": "The result of the expression 2 + 6 - 9 * 3 + 6 is -13."
}

Or use the chat interface in the research_agent frontend to interact via a sleek UI.

---

Custom Tools

This project supports easily pluggable custom tools. You can create your own logic tools like this:

def frequency_count(word: str, char: str):
    return word.count(char)

And register it in the tools list via LangChain's Tool interface.

📚 Resources

• LangChain Agents Documentation
• SerpAPI Docs
• Ollama Setup Guide