Baddie

Baddie helps you talk to your documents. Ask questions in natural language and get context aware answers. I've created this as a demo project to toy around with LLMs on amazon bedrock, planning to evolve it to something useful. Under the hood, it uses

• FM for chat : Meta Llama3-8B on Amazon bedrock (as of now)
• FM for generating embeddings : Amazon Titan Multimodal Embeddings G1
• PostgreSQL and PGvector for storing/retrieving embeddings And everything tied together with langchain and streamlit.

Running on local with python

Two steps

1. Create a python3 virtualenv and run.

   pip install -r requirements.txt

2. Add required values to keys in .env file.

3. Finally, run it using below command.

   streamlit run streamlit_app.py --server.port=80 --server.address=0.0.0.0

Running on docker-compose

Add right values for env_file in docker-compose.yaml and then run below command.

docker-compose up -d

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change. Few things I have in mind next :

1. Use langchain classes for holding chat context
2. Vector indexing
3. Session wise doc storage (currently its global)

Other chunkier improvements

1. Auth and hosting
2. Async ingestion pipeline for docs
3. Support other sources for data

Found some good demos and docs while building this, I'll add a references section for those.