raft
A command-line tool implementing Retrieval-Augmented Fine-Tuning, fetching a person's public writing from sources such as Substack, X, or RSS to build a combined fine-tuning and retrieval dataset.
raft implements Retrieval-Augmented Fine-Tuning, combining a fine-tuning phase with a RAG-based retrieval phase, aimed at building agents that emulate a specific target's writing style from their own public material. A single fetch command handles every importer, pulling from Substack, RSS feeds, an arbitrary URL, a PDF, or public posts on X or Bluesky, live or from an export file, with a shared date-range and count cap applied consistently across sources.
Fetched material is chunked, embedded, and stored so it can be queried later, and the project describes automatically fetching, chunking, embedding, storing, and querying a database starting from a single Substack URL. The author describes it as an older, informally maintained research project, originally built around emulating a public commentator's conversational style, shared because parts of the technique remain relevant despite the code's rough edges. It installs with uv and hatchling, and is MIT licensed.