Skip to content
Guide v1

Provider Availability and Known AI Limitations

Text AI, image generation, web search, embeddings and speech are separate capabilities. One working feature does not prove all providers are healthy.

Members can submit requests and use their available personal settings. Server administrators control channel/tool policy. Operator credentials, provider routing and budgets are not configured by posting keys in Discord or by using these docs.

The thinking preference accepts minimal, low, medium, high, xhigh, max and off. Support depends on the configured provider/model. Off omits the reasoning parameter; it does not promise the model will stop reasoning. Higher effort is not a guarantee of a better or faster answer.

  • AI can be wrong, omit context or produce unsafe code. Verify consequential output.
  • Stored memory and working semantic retrieval are different capabilities.
  • A fallback provider may have different parameter support.
  • Tools can be disabled, refused or unavailable despite normal text replies.
  • Speech uses a separately configured TTS service; direct Fish voice selection does not require users to submit Fish credentials.

No public docs page is a live provider health probe. Use Service status for observed service status and AI troubleshooting for symptom checks.