Skip to content
Codex Agent edited this page Jun 2, 2026 · 1 revision

Troubleshooting 🧯

This page lists practical fixes for Export2AI issues in VS Code, Cursor, ChatGPT project upload workflows, Claude code review archives, and local VSIX builds.


Zip Was Not Created

Check What to do
No workspace folder Open a folder/workspace first.
Selected path is outside a workspace Use a folder inside the current workspace.
Export was cancelled Run Export2AI → Zip Folder again.
Zip write failed Check workspace write permissions and disk space.
Empty archive error Verify the folder has readable included files after excludes.

Enable debug logging for details:

"export2ai.debug": true

Then open View → Output → Export2AI.


Important Files Are Missing

Symptom Likely cause Fix
expected build output missing built-ins exclude dist, build, out, site disable the built-in or add a narrower export folder
a fixture folder missing .gitignore or excludePaths check .gitignore, excludePatterns, excludePaths
.github/ missing credential guard or path exclusion inspect file names and explicit excludes
source file with token in name missing likely non-source extension or explicit rule use source extension or adjust excludes
dependency folder missing node_modules is intentionally excluded upload lockfiles/source instead of dependencies

Run Export2AI: Manage Built-in Exclude Patterns to inspect default excludes.


Sensitive Files Are Still Included

Export2AI blocks common credential patterns, but project-specific secrets can use unusual names.

Add explicit excludes:

{
  "export2ai.excludePaths": ["fixtures/private-data", "local-secrets"],
  "export2ai.excludePatterns": ["*.sqlite", "private-*.json"]
}

Then recreate the zip and inspect it before uploading.


Token Count Looks Wrong

Situation Explanation
~ appears before count the selected model uses an approximate offline method
Claude differs from API offline Claude estimates are approximate
Explorer badge differs from zip notification folder aggregation counts per file; zip count may count joined corpus text
unknown model is rough unsupported families use characters ÷ 4
token count disappeared export2ai.enableTokenCounting may be false

Check:

{
  "export2ai.enableTokenCounting": true,
  "export2ai.llmModel": "gpt-5.5"
}

Explorer Token Badges Do Not Show

This is usually expected.

"export2ai.showExplorerTokenBadges": false

Explorer badges are off by default. Enable them explicitly if you want compact folder badges.

After enabling, save/create/delete/rename a file or reload the window to trigger refresh. The first scan is intentionally deferred after startup to keep activation responsive.


Settings Page Opens Slowly or Not at All

Export2AI uses extension-specific settings navigation to avoid broad Settings search.

If it fails:

  1. Enable debug logging.
  2. Run Export2AI → Settings again.
  3. Check View → Output → Export2AI.
  4. Confirm the log shows an @ext: settings query for the installed extension ID.
  5. Use fallback actions such as opening the Extensions view and selecting the Export2AI gear icon.

Copy File Content Fails

Message Meaning
Copy content supports one file at a time multi-select is not supported
Copy Content to Clipboard is only for files selected item is a directory
binary/UTF-8 issue file is not readable as exact UTF-8 text
clipboard failure OS/editor clipboard API failed

Single-file copy intentionally copies exact UTF-8 text and does not apply zip ignore/comment/compression behavior.


Git Metadata Confusion

Expected behavior:

  • real .git/ internals are not exported
  • .github/, .gitignore, .gitattributes, docs, tests, tools, and AGENTS.md can stay included
  • default marker is _EXPORT2AI_PLACEHOLDERS/git/EXPORT2AI_SOFT_DELETE_PLACEHOLDER.txt
  • the archive is a source snapshot, not a Git clone

See Git Metadata Soft-Delete.


Build or VSIX Packaging Fails

Run the standard sequence:

npm install
npm run compile
npm run test:critical
npm run package

If generated menu/settings data is stale:

npm run compile

If package.json is unexpectedly huge, read Agent Chokepoints. The old token-bucket command explosion must not be reintroduced.


Debug Logging

"export2ai.debug": true

When enabled, Export2AI writes diagnostics to the Export2AI Output channel and reveals it automatically. Routine debug entries are not written when the setting is off.

Useful log areas:

  • activation
  • settings navigation
  • ignore preparation
  • file collection
  • token estimate scans
  • zip creation
  • copy project structure
  • copy file content
  • system file-manager reveal

Turn debug off again when finished.

Clone this wiki locally