Maproom Semantic Code Search FAQs

Question 1

When should I use vector-search vs standard search?

Accepted Answer

Use 'search' when you know specific keywords or function names. Use 'vector-search' for conceptual questions like 'how does the auth flow work?' or when exploring unfamiliar code without knowing the exact terminology.

Question 2

How long does it take to index a large repository?

Accepted Answer

A standard scan takes only a few seconds. However, generating embeddings for vector search makes API calls and can take several minutes for large repos; it is recommended to run this process in the background.

Question 3

Can Maproom search across different projects?

Accepted Answer

Yes, by configuring a 'maproom-repos.yaml' file, you can define multiple repositories and search across them sequentially to find design rationale in specs or implementations in source code.

Question 4

What is the benefit of the 'agent' output format?

Accepted Answer

The '--format agent' flag provides a compact, one-line-per-result output that conserves Claude's context window tokens while still providing location, score, and a code preview.

Question 5

How do I see what calls a specific function?

Accepted Answer

Once you find a code chunk, use the 'maproom context --chunk-id <id> --callers' command to trace all functions that invoke that specific piece of code.

Maproom Semantic Code Search

主な機能

ユースケース

Maproom Semantic Code Search

主な機能

ユースケース