Brain Box 🧠

Brain Box is a Retrieval-Augmented Generation (RAG) chat bot. The backend ingests documents, builds a vector store, and serves a chat API that answers questions using retrieved context.

Features

• Vector embedding and storage using ChromaDB
• Semantic retrieval and prompt orchestration with LangChain
• It uses Models powered by Azure AI Foundry
• Document ingestion and vector store rebuild on upload
• Pluggable LLM / embedding configuration via environment variables

Repository Structure

• backend/ — FastAPI app and RAG logic (app.py, main.py, vector_store.py, etc.)
• frontend/ — Streamlit UI
• data/ — Uploaded documents and vector store files
• templates/ — HTML templates (if used)
• docker-compose.yml — Docker compose for local deployment

Requirements

• Python 3.10+
• See backend/requirements.txt and frontend/requirements.txt for project-specific dependencies

Quickstart - Docker

docker compose up --build

Run locally

1. Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate

2. Install backend dependencies:

pip install -r backend/requirements.txt

3. Install frontend dependencies (if using Streamlit UI):

pip install -r frontend/requirements.txt

4. Create a .env file in the project root (see next section for variables).

5. Run the backend (development):

# from project root
python backend/app.py
# or
uvicorn backend.app:app --host 0.0.0.0 --port 8000 --reload

6. Run the frontend UI (Streamlit):

streamlit run frontend/streamlit_app.py --server.port 8501

.env — recommended variables

Create a .env file at the project root. The backend reads configuration from environment variables (via os.environ in the code). Below are recommended variables and example values — adapt these to your environment and provider.

Example .env:

# Azure AI Foundry
AZURE_API_KEY
API_VERSION
AZURE_ENDPOINT 

# Models
MODEL_NAME=gpt-4.1-mini   # or any supported model name
EMBEDDING_DEPLOYMENT_NAME=text-embedding-3-small
DEPLOYMENT_NAME = "gpt-4.1-mini"

How to use the .env file:

• Place .env in the repository root. Your shell or a dotenv loader will expose these values to the Python process.
• If you rely on python-dotenv in the codebase, the app will load .env automatically; otherwise export variables before starting the app:

Notes & Troubleshooting

• Ensure OPENAI_API_KEY (or other LLM provider keys) are valid and have required permissions.
• If the app cannot initialize the bot, check logs for missing env vars or missing dependencies.
• Uploaded documents are stored in data/ — if vector files are stored there too, keep backups before deleting.

Next steps / customization

• Swap embedding or LLM models by changing EMBEDDING_MODEL and LLM_MODEL in .env.
• Add provider-specific secrets (Pinecone, Azure, etc.) to .env and update backend/vector_store.py accordingly.
• Add authentication to the API in front of the endpoints for production.

Images