Verba
Self-hosted RAG chatbot with Ollama and cloud model support, licensed BSD-3-Clause. The project is archived and no longer maintained.
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 The project is archived and no longer maintained. It uses Weaviate to find relevant passages and gives those passages to a language model to generate answers.
You can run the application and Weaviate locally through Docker, connect to your own Weaviate server, or use Weaviate Cloud for storage. Ollama handles local embeddings and answer generation, with support for models such as Llama 3 and Mistral. HuggingFace supplies local embedding models. Choosing local storage and models keeps those parts of the workflow on your hardware; cloud storage, model providers and external import services process the data sent to them.
Verba accepts PDFs, Word documents and CSV or XLSX tables, plus files from GitHub and GitLab. Firecrawl brings in website content, while AssemblyAI transcribes audio for use in document queries. Its chat interface shows relevant source passages alongside generated answers.
Search combines semantic matching with keywords. Document filters and custom metadata help narrow results, and configurable text splitting accommodates prose, HTML, Markdown and code. A 3D vector viewer lets you inspect the stored data. Model connections include OpenAI, Anthropic and Cohere, as well as custom OpenAI-compatible endpoints such as LiteLLM.
The code is open source under BSD-3-Clause. Verba is designed for a single user and doesn't provide multi-user collaboration or role-based access. Existing Weaviate collections must be imported through Verba's interface to use its required document format.
