# archi

A retrieval-augmented generation framework, first built at MIT for the SubMIT computing project, aimed at research and education teams who need a configurable, private and extensible assistant with a low barrier to entry.

archi combines data retrieval with language models through customizable pipelines, and now powers chat, ticketing and course-support workflows across several academic and research organizations. It ingests data from sources such as web links, git repositories, local files and JIRA, and exposes that data through interfaces including a chat application, a ticketing assistant and an email bot.

Its modular design allows custom data sources, LLM backends, including both local and API-based models, and deployment targets to be substituted, and it ships containerized services and CLI utilities for repeatable deployment. Documentation covers configuration through a YAML schema, a REST API, agent and tool specifications with MCP integration, and guides for benchmarking and evaluating retrieval and response quality.

It is written in Python and licensed under MIT.

Website: https://archi-physics.github.io/archi/