Skip to main content

Settings - OpenAI Connections

Bring-your-own AI provider configuration.

πŸ“„οΈ Save the whole BYOAI configuration (fast, smart and embeddings)

Validate and save the three blocks a BYOAI tenant must provide: the fast chat model, the smart chat model and the embeddings endpoint. All three are tested before anything is written, and they are written together β€” a partial BYOAI would break every AI SAST scan. The embedding model id is optional (platform default) and must serve the index's vector width; its endpoint and key come from the request. Returns 400 on any failed test.

πŸ“„οΈ Test a BYOAI block from the form without saving

Dry-run the connectivity probe for one BYOAI block β€” 'validation' (fast), 'remediator' (smart) or 'embedding' β€” against the request body (the current form values), without persisting anything. Mirrors PUT /connections/byoai: every connection is the same generic OpenAI-compatible client, an empty api_key reuses the stored one, and an omitted embeddings model falls back to the platform default. Returns 400 on a failed probe. Requires the ai_byoai plan feature.

πŸ“„οΈ Reset the AI connections to the Plexicus defaults

Clear the client's own model overrides so every connection falls back to the platform (environment-configured) defaults. ai_type is 'validation' (fast), 'remediator' (smart) or 'embedding'; whichever is named, the THREE blocks are reset together, because a partial BYOAI breaks every scan. Deliberately NOT plan-gated: disabling BYOAI must work on every tier, so a downgraded tenant is never locked onto a provider it can no longer configure.