-
Notifications
You must be signed in to change notification settings - Fork 121
Description
Problem Statement
grepai hardcodes supported file extensions in indexer/scanner.go. Files with unrecognized extensions are silently skipped — no chunking, no embedding, no search results.
Our workspace (milaboratory/platforma) uses Tengo (.tengo), a Go-like scripting language, for 292 workflow files across 40+ blocks. These files contain core business logic — but grepai can't see them at all.
Per-language issues (#69, #128, #155) work for common languages but don't scale for niche ones — each requires a source change and a new release.
Proposed Solution
A custom_extensions field in .grepai/config.yaml would solve this for everyone:
chunking:
custom_extensions:
- .tengo
- .prismaFiles matching these extensions would pass the scanner's extension check and follow the standard chunking/embedding pipeline. Binary detection already guards against non-text files, so this should be safe.
Other languages that would benefit: Prisma (.prisma), Starlark (.star), Jsonnet (.jsonnet), Nix (.nix).
Alternatives Considered
No response
Category
Search / Indexing
Additional Context
Really appreciate this tool — it's the only local semantic search that actually delivers. This would make it work for polyglot codebases too.
Contribution
- I would be willing to contribute this feature