Troubleshooting (beta-era)
- · Pick the symptom that matches—you’ll bounce to the richer doc only when necessary.
- · Keep exact error text plus OpenHuman/OS versions when escalating to Discord or GitHub.
- · 'Empty briefing' complaints are timing more often than the model—wait one auto-fetch lap (~20 min) first.
Answers feel blank or shallow
- Finish at least one connector and wait ~20 minutes for the first scheduled pull—retry prompts after. Integrations primer
- If something never improves, skim First prompts expectations before swapping models blindly.
Installer or script blocked
- Windows SmartScreen / Defender: confirm publisher + download URL, optionally compare hashes when TinyHumans publishes them, then use “More info → Run anyway” only if satisfied.
- curl | bash · irm | iex: reread upstream scripts anytime they change—corporate proxies sometimes rewrite GitHub raw downloads.
- CPU mismatch: README cites Linux x64; Apple Silicon installers must explicitly match ARM builds.
Permissions won’t stick (macOS / Windows)
- After toggling macOS Accessibility, fully quit OpenHuman before re-testing—the OS caches stale bundles.
- Meeting-style features need microphone/camera; silent failures often mean denying them at the OS privacy pane—confirm toggles remain on.
Accounts & integrations misbehave
- Infinite OAuth loops: clear cookies on the SaaS login domain, revoke stale tokens vendor-side, re-open the connector wizard from OpenHuman after signing out/in.
- Missing context despite waiting: cross-check Gmail/Slack scopes actually granted—not just desktop login—per integrations guidance.
Still stuck?
Discord + GitHub issues linked from TinyHumans’ repository ↗ move fastest when you paste logs exactly as templates request—screenshots labeled with OS/build help too.