iblai-api-agent-llm

Mirrored from iblai/vibe · skills/agents/iblai-api-agent-llm/SKILL.md. This page is generated — edit it in the repository, not here.

Set an agent's LLM provider and model: list the available provider/model cards and write the chosen provider + model through the same settings/ endpoint that backs the rest of the agent configuration. Use when changing which model an agent runs on.

Auth & conventions

Base URL: https://api.iblai.app
Header: Authorization: Api-Token $IBLAI_API_KEY on every request.
Path vars: {org} = $IBLAI_ORG, {username} = $IBLAI_USERNAME, {mentor} = the agent's unique id (e.g. d17dc729-60fd-4363-81a0-f67d9318b03e).
Provider/model selection saves through the same settings/ endpoint — PUT …/users/{username}/mentors/{mentor}/settings/.
Not connected yet? Run /iblai-api-login first to populate IBLAI_ORG, IBLAI_USERNAME, and IBLAI_API_KEY.

Reads

GET https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/mentors/{mentor}/settings/ — current provider/model.
GET https://api.iblai.app/dm/api/ai-mentor/orgs/{org}/users/{username}/mentor-llms/?mentor_unique_id={mentor} — provider/model cards. mentor_unique_id runs the permission check in the agent's context (mentor_id is only a legacy filter). Each provider row carries name (the llm_provider value), logo, and chat_models[]; each model's llm_name is the llm_name value. Rows also report whether the organization holds a usable key (has_credentials, can_use_main_keys, main_has_credentials); a provider without one cannot be selected. Optional filters: provider, search, is_multimodal, supports_function_calling, min_context_window.

Writes

PUT …/users/{username}/mentors/{mentor}/settings/ — set provider/model:

{
  "llm_provider": "string (required)",
  "llm_name": "string (required)"
}

PUT …/users/{username}/mentors/{mentor}/settings/ — let users switch the model in their own chat session (Allow In-Chat Model Change; it resets to the agent's model on a new chat):
```
{
  "enable_in_chat_llm_change": "boolean"
}
```

Example

Switch an agent to OpenAI's GPT-4o:

curl -X PUT \
  "https://api.iblai.app/dm/api/ai-mentor/orgs/$IBLAI_ORG/users/$IBLAI_USERNAME/mentors/$MENTOR/settings/" \
  -H "Authorization: Api-Token $IBLAI_API_KEY" \
  -F "llm_provider=openai" \
  -F "llm_name=gpt-4o"

Notes

Pull the provider/model cards first, then send exactly one llm_provider + llm_name pair that appears in that list — values must match a real card.
Each card carries an availability status: byok_configured (the org's own key pays, on whichever gateway that key opens, e.g. its Vertex key serving Claude), byok_main_fallback (ibl.ai's key pays), gateway_credits (served on ibl.ai credits, including an org-held iblai key), byok_missing (add a key first). Only byok_missing sets requires_credential.
Each card also carries gateway / gateway_label (e.g. "vertex" / "Google Vertex AI"): the gateway that will serve the model by the org's gateway priority, set only when the org's own key pays; null on ibl.ai credits. The status comes from the same gateway resolver chat uses, so the picker and runtime agree. The org's model allow-list hides models only when a key ibl.ai lends would pay, never when the org's own key does. It is a different list from User Models (/iblai-api-rbac llm-access), which limits the models users who are not admins can pick in the chat.
Gateway rule: of the gateways that serve the model, the org's own key first, then ibl.ai's key if the org may borrow it. Credential names are in /iblai-api-integration.
Selection persists through the shared settings/ endpoint, so it sits alongside the other configuration fields on the same agent.
Which models users may pick when in-chat model change is on is an organization-wide list, not an agent setting: GET/POST /api/core/rbac/llm-access/… in /iblai-api-rbac. The LLM tab's User Models dialog (/iblai-vibe-agent-llm) edits it.