judahwu/familiar is a local-first AI coding agent that reads code inside a repository, plans changes, runs commands, and then hands the result over as a pull request. The subject of this introduction is the repository itself rather than a specific version, and the provided information lists no release version. It is written in Python and uses the MIT license. The README describes the flow in which the agent reviews the repository and proposes changes.
The knowledge base bundled with it organizes PDFs into document material for searching and produces answers with sources attached. According to the README, answers include a human approval step, and it supports semantic memory that carries over even when sessions change. A structural feature of this project is that code change proposals and document questions are both handled by the tools of a single repository. The default setup uses SQLite and a local shell. An extended setup using Postgres, Redis, Milvus and isolated containers is also described. Repository review, posting PR comments, Gitee, GitHub and GitLab integration, and observability features are likewise listed in the README.
The verifiable scope is limited to the features given in the provided repository description and the README summary. There is no public latest release, and no figures showing performance or answer accuracy were given. The repository creation date is 2026-11-03 and it has 3 stars. Therefore, rather than judging the maturity of the features or their actual effect, it is more appropriate to view this as an early-stage project that bundles a coding agent and a document-search knowledge base in one place.