AI
Per-tenant AI engine configuration. The chat / embedding pair is bound to a stored Credential (see the Credentials tag); models are picked from a capability-aware catalogue surfaced by /api/ai/models. Admin-only.
The quickstart tour state
The post-sign-up tour (E7.10, doc 59 §5): business answers → channel →
Save the business answers the AI needs
Writes the company card (name, description, website, phone, email) and a
Switch the AI on and finish the tour
Requires `confirm: true`. Turns the AI on (platform key with the
Prepare the AI playground for 'try it first'
Only while the tour runs. Puts the workspace on the key activation
End the tour without switching the AI on
Requires `confirm: true`. The tour is never offered again.
Run the setup tour again from the first step
Re-opens the tour on purpose (owner decision 2026-10-13), whatever its
Run the setup tour again where it stands
Like `restart`, but every step keeps its done mark, so the clients land
How far the workspace is set up
The 'your setup is X% complete — next: …' card (dashboard, mobile
Hide the setup progress card for the caller
Per person — other admins still see it. Idempotent.
Get AI settings
Returns the current tenant's effective AI configuration. Admin only.
Update AI settings
Partial update of the tenant AI configuration. Admin only. `api_key` is
The AI tool module catalogue (doc 97)
Every AI tool module the workspace's plan allows, the caller may configure (feature `ai_tools` and the
A tool module's workspace settings (manage-ai-settings)
Every path of the module's `config_schema` with its effective value. Credentials appear as ids only.
Change a tool module's workspace settings (manage-ai-settings)
Change a tool module's workspace settings (manage-ai-settings)
A tool module's use cases (profiles) (manage-ai-settings)
A tool module's use cases (profiles) (manage-ai-settings)
Create or change a use case (profile) (manage-ai-settings)
`config` is validated against the module's `profile_schema` (nested or dot-path keys). With
Delete a use case (profile) (manage-ai-settings)
Its runs stay (they keep the profile id); a conversation bound to it no longer finds a live profile.
Recent runs of a tool module (manage-ai-settings)
Recent runs of a tool module (manage-ai-settings)
Try a tool module's saved settings once (manage-ai-settings)
One real, metered call with the saved settings (today: web search — one search). Answers with what came
Get the tenant bot web-search settings (AI manager only)
Kept for existing clients. Web search is the `web_search` tool module now (doc 97 §12): this reads the
Update the tenant bot web-search settings (AI manager only)
Kept for existing clients; writes the `web_search` tool module (use `PUT /api/tool-modules/web_search/settings`).
Run a probe web search to validate the credentials (AI manager only)
Runs a search with the stored provider credentials and returns the result titles.
Send a stateless playground turn to the configured AI bot
Runs the real chat pipeline (AiManager → VectorSearchService →
List catalogue models for a credential
Returns the catalogue entries that match the given credential's type
The platform model marketplace and what each model costs per message
Whether the workspace can run its assistant on ConnectWiz's own AI keys
Rewrite the assistant's system prompt with AI
Sends the given system prompt to the workspace's chat model and returns
Teach the assistant from this reply ('use in training')
Turns an agent's (or the assistant's) reply into a question-and-answer
What the assistant has learned — the training examples
Every learned question-and-answer pair, newest first. Admin only
Approve a learned answer waiting for review
In review mode automatically learned answers wait as `pending` and the
Remove a learned answer ('remove from training')
The assistant stops using the example at once; a pending one is
AI usage broken down by credential
Returns one aggregate per credential for the requested billing period