NoteGenNOTEGEN.

Troubleshooting

Fix common model, sync, MCP, OCR, audio, workspace, and mobile issues.

Model connection fails

  1. Enter the base URL through its version segment, such as https://api.openai.com/v1; do not append /chat/completions.
  2. Verify the API key, model name, and model type.
  3. Run the connection check in Model Configuration.
  4. Ollama or LM Studio must be running and reachable from the device running NoteGen.
  5. Check whether the provider uses inherited, direct, or custom proxy mode.
  6. For older compatible APIs, change the token-limit parameter from max_completion_tokens to max_tokens.

Sync fails or conflicts

  • Confirm that the provider is connected and verify token permissions, repository/bucket, and path prefix.
  • Existing GitHub repositories usually need Contents read/write; repository creation also needs Administration read/write.
  • GitLab's REST API requires the api scope. Self-hosted Gitea requires the correct instance URL.
  • A conflict can be merged, replaced with remote data, or overwritten with local data. The overwrite choices can discard unsynced changes.
  • On a new device, choose the initial Records and Settings sync direction carefully.
  • Device-specific values such as workspace paths are excluded by default.

MCP server does not start

  • Mobile supports HTTP MCP only; stdio commands are desktop-only.
  • Use the runtime check for Node.js, Python, uv, npx, and related launchers.
  • Keep commands and arguments separate. Environment variables and headers must be valid JSON.
  • Test the connection, then select the server from the chat toolbar.
  • Do not import command-based configurations from untrusted sources.

OCR or image recognition fails

  • OCR requires the selected language packages and may download them on first use.
  • Use valid language codes, for example chi_sim,eng.
  • macOS and mobile builds may use native OCR; other environments may use Tesseract.
  • VLM recognition needs a compatible online model. Prefer OCR for clean text and VLM for complex layouts.

Recording or read-aloud has no audio

  • Check microphone permissions and the selected input device.
  • Transcription requires an STT model.
  • Use Auto for read-aloud; configure a TTS model when the system voice is unavailable.
  • For imported audio, verify the format and file integrity.

Workspace files are missing

  • Restart after changing the workspace so all components reload the path.
  • Confirm that NoteGen has read/write permission for the directory.
  • Unsupported files may appear as non-editable.
  • Switching workspaces does not move files. Switch back through history or migrate them in the system file manager.

Still stuck?

When opening an Issue, include the NoteGen version, OS version, reproduction steps, expected result, and actual result. Remove API keys, tokens, private paths, and note content from logs and screenshots.