What TRACE scans¶

TRACE only reads files that a parser or framework adapter can turn into routes, symbols or sinks. Everything else is skipped before it is opened, so it costs no I/O, hashing or graph nodes.

Analyzed languages¶

Languages	Extensions
Python	`.py`, `.pyi`
JavaScript / TypeScript	`.js`, `.mjs`, `.cjs`, `.jsx`, `.ts`, `.tsx`
Java / Kotlin	`.java`, `.kt`, `.kts`
Go, PHP, Ruby, Rust, C#, Dart	`.go`, `.php`, `.rb`, `.rs`, `.cs`, `.dart`

Skipped formats¶

Markup, styles, data/config files and similar formats yield no routes or sinks today and are not ingested: HTML, CSS and preprocessors (.scss, .sass, .less), JSON, YAML, TOML, SQL, Swift and C/C++. Images, fonts, archives, binaries, logs and databases were never ingested.

Also pruned without descending: .git, node_modules, virtualenvs, dist, build, target, vendor, .next, .cache, and tool caches such as .pytest_cache, .ruff_cache, .mypy_cache, .tox, .gradle, .dart_tool, Pods, site-packages. File-level skips include minified and bundled assets (*.min.js, *.bundle.js, *.chunk.js), TypeScript declarations (*.d.ts), source maps and *.egg-info. Files over 2 MB are skipped.

Customizing¶

.traceignore at the repo root: one glob per line, # for comments. Applied in every scan scope.
--scope app removes test and fixture directories from a scan; --scope fixtures scans only those. See CLI reference.
From Python, opt formats back in:

from pathlib import Path
from trace_engine.ingest.repository import RepositoryScanner

scanner = RepositoryScanner(Path("."), languages={"python", "json", "yaml"})
files = scanner.scan()
print(scanner.skipped_by_language)   # Counter of known-but-skipped formats

Config files

Endpoint discovery reads source code. API definitions in JSON or YAML (for example an OpenAPI file) are not analyzed yet; skipping them does not change today's findings.