Environment: ODC (OutSystems Developer Cloud) — Development stage
Problem:
Every AI model call from my ODC AI Agent fails at runtime with OS-ABRS-FM-40001. The provider is reported as "Unknown Provider" and "There are no healthy deployments for this model", even though the endpoint configuration passes "Test endpoint" successfully.
Error message:
OS-BERT-00000 - Failed to call AI model. Status code: BadRequest. The Unknown Provider AI model provider returned this message: "You passed in model=f6ec25b6-bfaf-400b-8d7a-be1963551b6c. There are no healthy deployments for this model." (OS-ABRS-FM-40001)
What I have already verified / tried:
- The endpoint configuration validates ("Endpoint configuration validated" / Test endpoint passes).
- Tried multiple providers: Google Gemini (gemini-2.5-flash, URL https://generativelanguage.googleapis.com/v1beta) and Anthropic Claude (claude-sonnet-5, claude-haiku-4-5-20251001, URL https://api.anthropic.com/v1).
- Model IDs confirmed correct against each provider's official docs.
- Provider accounts have sufficient credits and valid API keys.
- Deleted and recreated the model several times — a new model GUID is generated each time, but the same error occurs.
- Model GUIDs observed: f4c9eec8-d9a7-425f-bb0f-4f80d33af44a, 7ff28f6b-b455-49bf-be61-ea183973fcf0, f6ec25b6-bfaf-400b-8d7a-be1963551b6c.
- Occurs regardless of provider (both Gemini and Anthropic) and affects both the test app and the agent (OS-BERT-00000 and OS-BERT-62000).
Latest occurrence:
- Date/Time: Jul 10, 2026, 04:19:41 PM
- App: Gemini test app (Revision 2)
- Environment: Development
My questions:
1. What causes the "Unknown Provider" / "no healthy deployments for this model" (OS-ABRS-FM-40001) error when the endpoint itself validates successfully?
2. Since this happens for every provider and every newly created model, is this an environment/platform-level routing issue (OS-ABRS) rather than a per-model configuration problem?
3. Is there an additional step (e.g., deploying/enabling the AI model to the Development environment) required for the model to be registered as a "healthy deployment"?
4. How can I resolve this so my agent can successfully call the model?
- The model's own built-in Test (Server Action Call<Model> in ODC Studio) also fails with the same error — not just the agent. This confirms the issue is at the model connection / platform routing level, not the agent. - TraceId: 00-f1b95bb83343d62b54bc99b71c03d56f-7e1510b44a89e5b8-01 - Error at model test: Http 500, OS-ABPS-50006 → OS-ABRS-FM-40001, model=f6ec25b6-bfaf-400b-8d7a-be1963551b6c
Hello @YUNYOUNG NOH
Could be that you are configuring the Model Name with an ID?From here, it seems that the Model ID is actually the model name, not the GUID...
I never configured a model myself, so I am not sure.
Cheers!
Hi Eduardo,
Thank you so much for taking the time to look into this — I really appreciate your help and the suggestion. That's a great thing to double-check, so I went back and verified the configuration carefully.
The fields do appear to be set correctly:
- Name (display name): "Gemini"
- Model ID: "gemini-2.5-flash" — the actual Google model ID, entered separately from the display name.
For Anthropic I used Name "Claude" with Model ID "claude-sonnet-5" / "claude-haiku-4-5-20251001", so the Model ID field holds the real provider model identifier rather than the friendly name. The GUID that shows up in the error (model=f6ec25b6-...) isn't something I entered — it seems to be ODC's internal model identifier that gets generated automatically.
One more detail that might be helpful: I also tried the OutSystems-provided trial model, which I did not configure at all (no Model ID or Name set by me), and it fails with the exact same OS-ABRS-FM-40001 error. On top of that, the model's own built-in Test (the Call Server Action in ODC Studio) fails too, and the "Unknown Provider" message appears for every provider I try (both Google Gemini and Anthropic Claude) and for every newly created model.
Because even the trial model — which involves no configuration on my side — runs into the same "Unknown Provider / no healthy deployments" error, I'm starting to wonder if this might be an environment or platform-level routing issue (OS-ABRS) rather than a Model ID setup problem. Does that seem plausible to you?
For reference, here is the TraceId: 00-f1b95bb83343d62b54bc99b71c03d56f-7e1510b44a89e5b8-01
Thanks again for your time and guidance — it's genuinely appreciated. Please let me know if there's anything else I can check on my end.
Best regards,
YUNYOUNG
In which region are you?
OutSystems is having a major Outage there right now?https://status.outsystems.com/
You say the endpoint is correct and that you tried multiple providers, including the Trial ones. It's not the agent app, as you tested directly into the model.Did you tried connecting to those models from your Personal Edition? If they work, it is something related to the platform itself.
Desyncs happen and usually redoing the connection / redeploying the agent works. But it's not the case, it seems.
I would say you should open a ticket with the Support.
Sorry for not being able to help more. Maybe someone else will be able to give you a solution (if it is something missing or misconfigured).
Thank you again for following up — I really appreciate you sticking with this.
Quick answers to your questions first:
- Region: I'm on Asia Pacific / Seoul. I checked the status page, and all ODC components for Seoul show "Operational". The current Major Outage is limited to the Middle East (UAE) region, so it doesn't affect mine.
- Yes — I did test it on ODC Personal, which was a great suggestion. The same error occurs there. A colleague also tried it on their own separate ODC Personal organization, and it fails in exactly the same way.
So the identical OS-ABRS-FM-40001 / "Unknown Provider / no healthy deployments" error reproduces across three independent ODC organizations:
1. Our company partner demo environment (where I first hit it)
2. My own ODC Personal edition
3. A colleague's ODC Personal edition
It also fails regardless of provider (Google Gemini and Anthropic Claude), for every newly created model (a new GUID each time), and even the OutSystems-provided trial model — which I didn't configure at all — fails the same way. The model's own built-in Test (Call in ODC Studio) fails too, not just the agent.
Update: I opened a ticket with Support as you suggested, and they have already been able to reproduce the issue on their side and are currently validating it. So it does appear to be a platform-level issue rather than a configuration one, which lines up with what we were seeing.
I'll share the resolution here once Support gets back to me, in case it helps anyone else who runs into the same thing.
Thank you so much again for your time and guidance — your questions really helped me narrow this down. Cheers!
I also experienced the same in Personal Environment.
It's usually resolved on it's own, or trying re-publish might help sometime.
Sometimes if run the debugger, it seems to work fine as well.
This is most likely an ODC platform/routing issue, not a model configuration problem.
Since it also fails with trial models, multiple providers, and across different ODC environments, there is no extra deployment step missing on your side.
Keep the Support ticket open and share the TraceIds, region, and timestamps. Republishing may help temporarily, but it is not a reliable fix.