Login

Troubleshooting & FAQ

Protocol for troubleshooting and FAQ from user feedback

Troubleshooting

If you are experiencing any issues, please follow these steps to troubleshoot:

  • Check if you are on the latest version of Copilot.
  • Disable all other plugins to see if Copilot is causing the issue.
  • Turn on debug mode in Copilot settings.
  • Open the dev console in Obsidian by pressing cmd + option + i on Mac or ctrl + shift + i on Windows.
  • Try to understand the error message.
  • If you can't solve the problem and believe it's a bug, check GitHub to see if it's already reported. If not, please open an issue following the bug report template.

Frequently Asked Questions

🔸 RangeError: invalid string length

Q: My vault files are too big. How do I fix indexing issues caused by file size?

A: Split the vault into smaller partitions using the Partitioning setting in QA options. Rule of thumb: the first partition file should be less than 400mb.

🔸 Referencing notes.

Q: Can I reference a note in Copilot chat?

A: Yes—use [ [Note Name] ] syntax to reference notes.

🔸 Truncated AI responses.

Q: My AI response is getting cut off. How do I fix that?

A: Raise the Max Token limit in Copilot Settings -> Model Settings -> Token Limit.

🔸 Indexing doesn't work.

Q: Indexing isn't working. What should I do?

A:

  1. Use the 3-dot menu → Run Copilot: Index (refresh) vault or Copilot: Force reindex vault in Command Palette (Mac: Cmd + P, Windows: Ctrl + P).
  2. Switch to a larger text embedding model and tune batch size + RPM if using OpenAI's small model.
  3. If you see a "RangeError: invalid string length", increase the number of partitions in QA settings if the first index file exceeds ~300MB. Check .obsidian/ for copilot-index files for their sizes.

🔸 Indexing but not retrieving.

Q: Copilot isn't finding notes I know are indexed. What should I check?

A:

  1. If you're not a Copilot Plus subscriber, use Vault QA. If you are, switch to Copilot Plus mode and use the @vault tool in chat.
  2. Raise "Max Sources" in QA settings.
  3. Try the multilingual model for non-English notes.
  4. Review QA filters for inclusion/exclusion.
  5. Run the "List Indexed Files" command.
  6. Run "Copilot:Force reindex vault" in Command Palette (Mac: Cmd + P, Windows: Ctrl + P).

🔸 API Basics.

Q: What is an API and how do I set it up for Copilot?

A:

  • Use your OpenAI API key under Copilot Settings > Basic Settings -> API Keys -> Set Key.
  • For Copilot Plus, enter your license key in the Plus License field. Obtain your license key in your dashboard on https://www.obsidiancopilot.com/en/dashboard.

🔸 Models with Vision.

Q: Why can't the model understand images?

A: Only models marked with a vision icon support image understanding. Enable via Copilot Settings → Model → Capabilities -> Vision icon is shown.

🔸 Enforcing English Responses.

Q: How can I make sure Copilot always replies in English?

A:

  • Copilot Settings -> Advanced Settings -> Set "User System Prompt" to "Always respond in English".

🔸 Image as context.

Q: Can Copilot understand images in my vault?

A:

  • Only supported in Copilot Plus.
  • Ensure Copilot Settings -> Model Settings -> Select a model with "vision capabilities" enabled
  • Ensure Copilot Settings -> Advanced Settings -> "Images in markdown" turned on
  • 💡 Tip: Set a longer context window for your selected models.

🔸 PDF reading issues.

Q: Why can't Copilot read my PDF?

A:

  • Large PDFs (>10MB) should be converted to Markdown.
  • PDF parsing is costly; Project Mode is better suited for large PDFs, coming soon.
  • 💡 Tip: Reduce size by converting embedded images to text.

🔸 Mobile/iPad support.

Q: Why doesn't Copilot work fully on mobile or iPad?

A: Mobile/iPad support is experimental and may lack full feature parity with desktop.

🔸 Privacy Concerns.

Q: Which AI model should I use with Copilot to minimize the risk of exposing sensitive data, while still being able to process non-private data?

A: Gemini, when accessed via its paid API (which is the foundation for copilot-plus-flash), does not use user input data for training its models. This makes it a safe option for processing your data. For maximum privacy, consider using local or self-hosted models. We are going to ship an "offline mode" for Believers with maximum privacy in mind.

Copilot Settings