AI Model
Configure chat and embedding providers, test connections, and manage the introductory AI allowance.
answerLoops uses a language model to draft answers and an embedding model to search your knowledge base. Settings → AI Model controls which provider and models an org uses.
What runs by default
| State | What's used |
|---|---|
| A new cloud org, no key configured | answerLoops' key, metered as a free trial — 5 AI‑answered tickets, lifetime, per org |
| Trial used up, still no key | AI drafting stops; connect a provider to resume |
| Self‑hosted | The OPENAI_API_KEY in your environment |
| A workspace with saved model settings | Its configured chat and embedding services |
The card shows where you stand — trial credits remaining, "trial used up", or the configured provider and model.
Bring your own provider
Included on every plan for the named providers below. A custom OpenAI‑compatible endpoint (your own base URL — Ollama, LM Studio, vLLM, a proxy) is part of Custom AI model configuration, which is included on Enterprise and in the self-hosted edition.
| Provider | Chat | Embeddings |
|---|---|---|
| OpenAI | ✓ | ✓ (same key covers both) |
| Anthropic | ✓ | — needs a separate OpenAI key for embeddings |
| Google Gemini | ✓ | — separate OpenAI key |
| Groq | ✓ | — separate OpenAI key |
| Mistral | ✓ | — separate OpenAI key |
| xAI (Grok) | ✓ | — separate OpenAI key |
| OpenAI‑compatible endpoint | ✓ (Enterprise) | ✓ (Enterprise) |
In answerLoops, OpenAI is the named provider option for embeddings. If your chat provider is anything else, enter an OpenAI API key under Embeddings — otherwise knowledge‑base search and every ingest fails after you save. The form asks for it automatically and the save is rejected with a clear message if it's missing.
Steps
- Settings → AI Model → Edit (or Save for a first‑time setup — owner or admin only).
- Pick the chat provider, enter its API key, then click Refresh models to pull that provider's current model list live — Model ID stays a dropdown of whatever that provider actually offers today instead of a list that goes stale between releases. If the live fetch fails (or before you've run it), Model ID falls back to a built‑in list of recent models, or you can type an ID directly with Custom model ID….
- Pick the embedding provider and model. Add an OpenAI embedding key if the chat provider isn't OpenAI.
- Click Test connection — this makes a one‑token call to each provider with
the values you entered (nothing is saved yet) and reports
chat: connection OK/embedding: …or a specific error (rejected key, unknown model, unreachable endpoint). - Click Save.
Changing provider later
If you switch the chat or embedding provider and leave the key field blank, the
save is rejected — the stored key belonged to the previous provider and would
fail on every call. Enter a key for the new provider. (A keyless local endpoint
is the exception: switching to openai-compatible with no key clears the old
one.)
Leaving a key field blank on an update to the same provider keeps the stored key — the field shows "Key saved".
Reset
Reset to platform default deletes the org's AI config. A cloud org falls back to the trial (or, if exhausted, to no AI); a self‑hosted deployment falls back to its environment key.