txtai
Open-source AI framework for semantic search, RAG and agents. Runs locally or in Docker, with Hugging Face, llama.cpp and cloud models via LiteLLM.
txtai is a Python framework for developers building search applications, chat with their data, and AI agents on their own hardware or servers. Its embeddings database combines sparse and dense vector search with graphs and relational data, so the same system can find related content and supply context to language models. It's open source under Apache 2.0.
Semantic search finds results by meaning rather than requiring matching keywords. txtai supports SQL queries, object storage, topic modeling and graph analysis, and can index text, documents, audio, images and video. It can also place images and text in a shared search space.
Model pipelines handle question answering, summarization, transcription, translation and text-to-speech. Workflows connect these tasks and combine specialized smaller models with LLMs. Agents built on smolagents can use the database, pipelines, workflows and other agents to carry out tasks.
You can run models locally with Hugging Face or llama.cpp and keep data within your deployment instead of sending it to remote services. Models can load from local directories or the Hugging Face Hub. LiteLLM also connects to cloud services such as OpenAI, Claude and AWS Bedrock; those calls use remote providers.
txtai runs locally or in Docker and can scale through container orchestration. Its web and Model Context Protocol (MCP) APIs expose applications to other software, with language bindings for JavaScript, Java, Rust and Go.
