| 1 |
"""Curated default-exclude patterns applied unless opted out. |
| 2 |
|
| 3 |
Everyone training on a codebase tree wants the same starter set gone: |
| 4 |
VCS metadata, secrets, build artifacts, lockfiles, binaries. Shipping |
| 5 |
these as defaults removes the "why is my adapter memorizing |
| 6 |
package-lock.json" foot-gun from the common path. |
| 7 |
|
| 8 |
`training.yaml` can opt out per-subtree via `exclude_defaults: false` |
| 9 |
when a tree legitimately wants to train on (e.g.) generated code or |
| 10 |
the `.git` dir itself. Defaults are a *starting point*, not a security |
| 11 |
boundary — users with real secrets still need their own excludes. |
| 12 |
""" |
| 13 |
|
| 14 |
from __future__ import annotations |
| 15 |
|
| 16 |
from typing import Final |
| 17 |
|
| 18 |
DEFAULT_EXCLUDES: Final[tuple[str, ...]] = ( |
| 19 |
# Version control metadata |
| 20 |
".git/**", |
| 21 |
".hg/**", |
| 22 |
".svn/**", |
| 23 |
# Secrets / local config (best-effort; not a security boundary — |
| 24 |
# users with actual secrets need an explicit exclude list). |
| 25 |
".env", |
| 26 |
".env.*", |
| 27 |
"**/.env", |
| 28 |
"**/.env.*", |
| 29 |
"**/id_rsa", |
| 30 |
"**/id_ed25519", |
| 31 |
"**/*.pem", |
| 32 |
"**/*.key", |
| 33 |
"**/secrets.*", |
| 34 |
# Python artifacts |
| 35 |
"**/__pycache__/**", |
| 36 |
"**/*.pyc", |
| 37 |
"**/*.pyo", |
| 38 |
".venv/**", |
| 39 |
"venv/**", |
| 40 |
".tox/**", |
| 41 |
# Node / JS artifacts |
| 42 |
"node_modules/**", |
| 43 |
"**/*.min.js", |
| 44 |
"**/*.min.css", |
| 45 |
"**/*.map", |
| 46 |
# Rust / Go / Java / C / C++ compiled output |
| 47 |
"target/**", |
| 48 |
"**/*.rlib", |
| 49 |
"**/*.class", |
| 50 |
"**/*.jar", |
| 51 |
"**/*.o", |
| 52 |
"**/*.so", |
| 53 |
"**/*.dylib", |
| 54 |
"**/*.dll", |
| 55 |
# Build / dist trees |
| 56 |
"build/**", |
| 57 |
"dist/**", |
| 58 |
"__generated__/**", |
| 59 |
"generated/**", |
| 60 |
# Lockfiles — long, low training signal, noisy to diff |
| 61 |
"**/package-lock.json", |
| 62 |
"**/yarn.lock", |
| 63 |
"**/pnpm-lock.yaml", |
| 64 |
"**/Cargo.lock", |
| 65 |
"**/uv.lock", |
| 66 |
"**/poetry.lock", |
| 67 |
"**/Pipfile.lock", |
| 68 |
# Archives + non-image binaries. Image extensions (png/jpg/jpeg/gif/ |
| 69 |
# webp/bmp/tiff) are intentionally *not* excluded — schema v10 made |
| 70 |
# them first-class training material via `SectionType.IMAGE`, and |
| 71 |
# `expand_sources` refuses to ingest them unless the caller supplies |
| 72 |
# a `BlobStore`. Text-only workflows that pass no blob store still |
| 73 |
# see zero images because the walker tallies them under |
| 74 |
# `skipped_image_no_store` instead of emitting Sections. |
| 75 |
"**/*.ico", |
| 76 |
"**/*.pdf", |
| 77 |
"**/*.zip", |
| 78 |
"**/*.tar", |
| 79 |
"**/*.gz", |
| 80 |
"**/*.xz", |
| 81 |
"**/*.bz2", |
| 82 |
"**/*.7z", |
| 83 |
"**/*.wasm", |
| 84 |
# dlm's own config — don't train on the training config |
| 85 |
".dlm/**", |
| 86 |
"**/.dlm/**", |
| 87 |
) |