OpenHuman Guide

Problem solving

OpenHuman Troubleshooting

Diagnose first, then escalate. Capture the error, check the README, search existing issues, then file a report with full context.

Install guidePrivacy checklist

OpenHuman Troubleshooting

Practical notes for evaluating a fast-moving open-source AI assistant.

Practical, source-linked OpenHuman guidance

First Checks

  • Use the official download or script URL only.
  • Capture the exact command, OS, architecture, app version, and error text.
  • Search open and closed GitHub issues before filing a new report.
  • Test on a clean profile before connecting sensitive accounts.

Installer Problems

  • Download fails: check network access and company proxies.
  • Windows SmartScreen or antivirus blocks the binary: verify the source is official before bypassing.
  • Windows crash or restart loop: fixed in a recent rebuild — check latest release notes.
  • macOS Gatekeeper blocks the app: verify the installer is signed and from the official page.
  • Linux blank window: check desktop libraries and Wayland compatibility; try the AppImage.

First Launch and Onboarding

  • OAuth hangs: check network, proxy, and system clock accuracy.
  • Auto-fetch does not complete: wait one 20-minute cycle after connecting the first account.
  • Memory tab is empty: confirm the Obsidian vault path is writable and the first sync finished.
  • Missing permissions: on macOS, grant Accessibility, Screen Recording, and Microphone in System Settings.

Integration and Local AI Issues

  • Integration does not sync: recheck OAuth permissions and whether the token was revoked at the source.
  • Connector accesses more than expected: read the OAuth scope carefully before accepting.
  • Ollama URL read-only in the UI: known Windows 11 issue with an existing Ollama server; edit config.toml directly.
  • Ollama not detected: confirm port 11434 is open and Ollama is running.
  • Common Ollama issues: PATH not picked up, antivirus quarantining, non-standard install location, port conflict, or insufficient memory for the loaded model.