Glossary
The words Garage uses, each in a sentence or two. Across the site, a word with a dotted underline shows its definition when you point at it, tab to it or tap it.
- Corpus
- Everything Garage has indexed from your sources — documents, code and messages — searchable as one collection.
- Source
- A folder, git repository, or Messages or Mail library you add to Garage to be indexed.
- Chunk
- A passage of a document. Garage embeds and searches chunks, so search returns the part that matches, not the whole file.
- Embedding
- A list of numbers a model computes for a chunk, so passages with similar meaning sit close together. Garage keeps one per chunk for each model.
- Corpus class
- What a document is — document, code or communication. Communications never leave your computer.
- Trust tier
- How much a document is trusted — authored by you, reference material, or received from someone else.
- Attribution
- Working out who wrote each document, from git history, document metadata and path rules, with the evidence recorded.
- MCP
- The Model Context Protocol, the open standard AI assistants use to call tools such as Garage's search.
- Hybrid search
- Vector similarity and keyword full-text search run together, with the two rankings fused into one list.
- Reciprocal Rank Fusion
- A way to merge rankings — each result scores 1 ÷ (60 + its rank) in every list, and the scores are added up.
- Backfill
- Embedding the chunks you already have with a newly added model, without reading your files again.
- Fact
- A short, self-contained statement a local model distills from a document, linked to the passage it came from.
- Egress
- Anything Garage sends off your computer. All of it passes one tested choke point with an allowlist of destinations.