Verba

Verba is a self-hosted document chatbot for people who want to ask questions across their files and knowledge bases. The project is archived and no longer maintained, so it won't receive bug fixes or security patches.

• Questions across documents: Use retrieval augmented generation (RAG) to answer questions using your data and cross-reference information across documents.
• Local models: Ollama supports local embeddings and answer generation with models such as Llama 3 and Mistral. HuggingFace provides local embedding models.
• Cloud model connections: Connect to providers including OpenAI, Anthropic, Cohere and Groq, or use custom OpenAI-compatible endpoints such as LiteLLM.
• Document and audio imports: Import text, PDFs and CSV files through Unstructured, or use AssemblyAI to turn audio files and audio from videos into material you can query. These integrations require external services and API keys.

Verba runs locally or in a cloud deployment and supports Docker. It uses Weaviate for its database; the embedded local Weaviate option doesn't support Windows, which requires Docker or cloud deployment. Local model processing runs on your hardware, while cloud providers process requests sent to them. HuggingFace embedding models download at startup when used.

The code is open source under the BSD-3-Clause license.