# FAQ Bot with Chroma and MCP Tool

## Overview
This project implements a FAQ bot that can answer questions from a local set of FAQ markdown files and from an external course metadata service via an MCP‑style HTTP tool.

- **Local data** – loaded from `data/` and persisted in a Chroma vector store.
- **MCP tool** – HTTP GET request to `https://api.example.com/course_meta`.
- **LLM** – Ollama’s `llama3` model with `nomic-embed-text` embeddings.
- **CLI** – preset questions and interactive mode.

## Setup

### 1. Clone the repo
```bash
git clone <repo-url>
cd <repo-dir>
```

### 2. Create a virtual environment
```bash
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
```

### 3. Install dependencies
```bash
pip install -r requirements.txt
```

### 4. Download the Ollama model
```bash
ollama pull nomic-embed-text
```

### 5. Set environment variables
Create a `.env` file in the project root:
```
OLLAMA_BASE_URL=http://localhost:11434/v1
```
(If you run Ollama on a different host/port, adjust accordingly.)

## Running the Bot
```bash
python agent.py
```
You will see preset questions. Select a number or type a custom question.

## Data Folder
Place your FAQ markdown files in the `data/` directory. The bot will automatically load them on first run and persist the vector store in `chromadb/`.

## Dependencies
See `requirements.txt` for exact versions.