AI Providers

Connect a model provider for log and deployment analysis.

The Ask the AI actions in the dashboard — one per deployment failure and per log view — send the relevant context to an AI provider and get a plain-language diagnosis back.

Providers#

Under Settings → Providers, connect one of:

  • Groq
  • Gemini
  • OpenRouter
  • any OpenAI-compatible endpoint (model + base URL)

Each provider has optional per-surface defaults: which model handles log analysis, which handles deployment analysis, and which is the project-wide default. Keys are stored encrypted and used only by the control plane.

How the analysis works#

  • Logs are fetched server-side and treated as untrusted data.
  • The question, the logs and the result remain transient — nothing is stored or re-sent after the analysis.
  • Truncation and confidence are surfaced, so a diagnosis over a 30-second window is labeled as such, not presented as gospel.

No provider or key is required to run the platform — the AI actions simply stay disabled until one is connected.