codegraph status still reports untracked files as pending after sync

[15290391025]

Environment

• OS: Windows 11
• CodeGraph: npm global install
• Backend: native
• Project type: Git repository with newly-created untracked source files

What happened

I initialized and synced a project that had newly-created untracked source files.

cd D:\agent\ai
codegraph init -i
codegraph status
codegraph sync
codegraph status

codegraph sync reported success:

Synced 21 changed files
Added: 21 - 747 nodes in 1.2s

However, codegraph status still reports the same pending changes:

Pending Changes:
  Added:     21 files
[i] Run "codegraph sync" to update the index

Expected behavior

After codegraph sync successfully indexes the added files, codegraph status should not continue reporting those same files as pending.

Actual behavior

status keeps reporting the untracked files as pending even though the symbols from those files are queryable.

For example, after sync these new symbols are found:

codegraph query AdminQuotaView --limit 5
codegraph query RagFeedbackService --limit 5

Both return results from newly-created files, including:

ai-entitlement-api/src/main/java/com/demo/entitlement/api/dto/AdminQuotaView.java
ai-rag-impl/src/main/java/com/demo/service/RagFeedbackService.java
ai-rag-impl/src/test/java/com/demo/service/RagFeedbackServiceTest.java

Notes

This looks specific to Git untracked files. The files remain ?? in git status, but they have already been indexed into .codegraph/codegraph.db.

So this appears to be a status/change-detection issue rather than an indexing issue.

[colbymchenry]

Fixed in #218 (merged to main), shipping in the next release.

Root cause: CodeGraph's git-based change detection (used by both codegraph status and codegraph sync) treated every untracked (??) file as "added" without checking whether it was already in the index. Indexing a file into .codegraph/codegraph.db doesn't make git track it — the file stays ?? in git status — so on every run it was re-reported as a pending addition and re-indexed from scratch. That's why status kept listing the files under Pending Changes and sync kept reporting Added: N even though the symbols were already queryable. Your read that this was a change-detection issue specific to git untracked files was exactly right.

Fix: untracked files are now hash-compared against the index the same way tracked files are — counted as "added" only when missing from the index, "modified" when their contents changed, and skipped otherwise.

I reproduced your exact workflow (init -i → status → sync → status) against a real repo with untracked files and confirmed the before/after at the CLI level:

Step	Before	After
status after init -i	added: N ❌	added: 0 ✅
status after sync	added: N ❌	added: 0 ✅
repeat sync	re-adds every run	no-op ✅

Once the next version is published, npm install -g @colbymchenry/codegraph (or your install method) will pull it in. Thanks for the precise, well-isolated report! 🙏