Is an LLM provider configured for our Archestra instance? My app's built-in archestra.llm.complete() call is failing 100% of the time with error code llm_unavailable ("The LLM completion could not be produced") — it's not a quota issue (llm_quota), and it happens on every request.

My app code is correct (the call is caught and the error surfaced cleanly). I also checked and there's no LLM configuration tool or model-serving tool exposed to app authors, so I can't enable it myself.

Can someone confirm whether an LLM provider/API key + model is set up at the instance/admin level for app-level llm.complete calls? If it needs enabling, could you turn it on (or tell me who can)?

@user could you please check?

I'll dig into the Archestra docs to understand how the built-in llm.complete() capability is configured and what could cause a consistent llm_unavailable error. Let me check. Let me read the most relevant docs: adding LLM providers, apps, and the LLM proxy. I now have a clear picture. The docs confirm exactly how archestra.llm.complete() works and what llm_unavailable means. Let me post the answer. Here's what's going on, Matvey — and Sathvika's diagnosis is essentially correct. *How archestra.llm.complete() works (per the [MCP Apps docs](https://github.com/archestra-ai/archestra/blob/main/docs/pages/platform-apps.md)): - It runs one host LLM completion as the viewer, using the organization's configured model — the app can't choose the model, and there is no app-author-facing tool to configure it. That matches Sathvika finding no LLM/model-serving tool exposed to app authors — by design. - The call goes through the LLM Proxy, counts against the viewer's usage limits, and is recorded like any other interaction. - It has exactly two typed failure modes: - { code: "llm_quota" } → usage limits reached - { code: "llm_unavailable" } → everything else So llm_unavailable on 100% of requests (and confirmed not llm_quota) points at the org-level LLM setup behind the proxy, not the app code. This is an admin/instance-level thing, not something an app author can fix. What to check on the instance (admin side): 1. Is a provider actually configured for the org's built-in model? Since llm.complete() uses the organization's configured model routed through the proxy, there needs to be a working provider + API key + model behind it. See [Adding/Supported LLM Providers](https://github.com/archestra-ai/archestra/blob/main/docs/pages/platform-supported-llm-providers.md) and the chat provider config (ARCHESTRA_CHAT_{PROVIDER}_API_KEY / ARCHESTRA_{PROVIDER}_BASE_URL). 2. Is the backing proxy/route healthy? A provider whose route is registered but not enabled at runtime returns a 500 ("not configured"), which would surface to the app as llm_unavailable. Worth checking the provider is enabled (has a baseUrl) and the API key is valid/not expired. 3. Check the LLM Proxy Logs UI (/llm/logs) — since these calls are recorded like any other interaction, the failed completions should appear there with the real upstream error (bad key, model not found, provider down), which will tell you exactly why. Bottom line:* Sathvika can't turn this on from app-author tooling — it needs someone with instance/admin access to confirm the org's configured LLM provider + key + model are set up and healthy behind the proxy. Matvey, since you're on the Archestra team, you (or whoever administers this instance) would be the one to verify the provider config and check /llm/logs for the underlying error. Want me to pull up the specific env vars / provider setup steps from the supported-providers or deployment d…