Ignore rules (.gitignore / .arbor config) not honored outside a git repo — venv/vendored trees bloat the graph

[vortxhq]

Summary

In a non-git directory, arbor appears to honor no exclusion mechanism, so dependency and
vendored trees (e.g. venv/, node_modules/, site-packages/, decompiled/vendored code) get fully
indexed and dominate the graph — inflating index size, status, and centrality ranking.

What I observed (sandbox, arbor 2.2.0 and re-checked on 2.6.0)

Indexing a plain (non-git) directory, verified by reading .arbor/graph.json:

mechanism	excluded?
built-in excludes (venv/node_modules/site-packages)	no
.gitignore (dir is not a git repo)	no
.arborignore	no
ignore list in .arbor/config.json	no
git init + .gitignore	yes
moving files out of the tree (symlink back)	yes (symlinks not followed by default)

Concretely, a venv/ with ~27k files ended up in the graph, producing a multi-GB graph.json, while
the real first-party code was a few thousand files.

Repro

1. mkdir demo && cd demo (do not git init)
2. add src/main.py plus a venv/lib/.../pkg/mod.py and a .gitignore containing venv/
3. arbor index .
4. inspect .arbor/graph.json → the venv symbols are present

Expected
Outside a git repo, honor .gitignore (root + nested) and/or the ignore list in .arbor/config.json,
so users can exclude dependencies without needing git init.

Workarounds today

• git init + .gitignore (arbor then honors it), or
• relocate the heavy trees out of the indexed directory and symlink them back.

Either would be nicer as native config. Happy to help repro further. Thanks for arbor!