detecting-languages

Detecting languages

tree-sitter-language-pack maps a file to one of 371 supported languages. Detection works from a file path, a bare extension, or — via the SDK — the file content itself.

On the CLI: detection is implicit

There is no standalone ts-pack detect command. parse and process auto-detect the language from the file extension and act on it:

ts-pack parse src/app.ts      # detects "typescript", parses
ts-pack process app.py        # detects "python", extracts

When detection fails (no extension, ambiguous, or stdin), pass --language:

cat snippet | ts-pack parse - --language go

To check whether a language name is supported and cached without parsing anything, use list and info:

ts-pack list --filter type      # languages whose name contains "type"
ts-pack list --manifest         # every language in the remote manifest
ts-pack info typescript         # is it known? is it downloaded? cache path?

In the SDK: explicit detection

The core exposes three detection functions, including content-based detection that the CLI does not surface:

Function	Detects from
detect_language_from_path(path)	Full path (uses the extension).
detect_language_from_extension(ext)	A bare extension like "rs".
detect_language_from_content(content)	Raw source text — no path needed.

Each returns the language name or null/None when no grammar matches.

Python

from tree_sitter_language_pack import (
    detect_language_from_path,
    detect_language_from_extension,
    detect_language_from_content,
)

detect_language_from_path("src/app.py")      # "python"
detect_language_from_extension("rs")          # "rust"
detect_language_from_content(open("x").read())  # from the shebang line, if present

Node.js / TypeScript

import { detectLanguageFromPath, detectLanguageFromContent } from "@xberg-io/tree-sitter-language-pack";

const lang = detectLanguageFromPath("src/app.ts") ?? detectLanguageFromContent(source);

Choosing CLI vs SDK

• Path or extension is enough, one-shot → just run ts-pack parse/process and let it auto-detect.
• Only raw content (a pasted snippet, a stream, a file with no extension) → use the SDK's detect_language_from_content, then pass the result as --language to the CLI if needed.

When to reach for the other skills

Once the language is known, hand off to parsing-source (syntax tree) or extracting-code-structure (metadata). To see or prefetch the parser for a detected language, see managing-parsers.