Supported Content
See how Zvec-Grep extracts code, documents, structured text, and images.
Zvec-Grep assigns each admitted file to an extraction path. Structure-aware extractors preserve useful source context; other text files fall back to plain chunks.
Structure-aware code
The following languages preserve symbols, signatures, breadcrumbs, and surrounding source:
- C and C++
- Go
- Java
- JavaScript and JSX
- TypeScript and TSX
- Python
- Rust
Vue and Svelte component scripts also use structural extraction where possible.
Documents and structured text
| Content | Indexed representation |
|---|---|
| Markdown and MDX | Heading sections and breadcrumbs |
| TXT, RST, HTML, XML | Plain-text chunks |
| CSV, JSON, JSONC, TOML, YAML | Plain-text chunks |
| Other non-binary files | Plain-text fallback |
Other recognized code such as Ruby, PHP, Swift, Kotlin, C#, Scala, shell, SQL, and CSS currently uses plain-text chunks when no structural grammar is available.
Images
Images are skipped by default. To index GIF, JPEG, PNG, or WebP files, include them explicitly and select an Embedding model that accepts images.
Content currently skipped
PDF and Office files, archives, compiled artifacts, audio, video, database files, empty files, binaries detected by extension or content sampling, and files above the configured size limit are not indexed.
Use zg index --debug to inspect skipped-file counts and samples. Scope discovery with --glob, file types, ignore files, depth, and maximum file size.