How do we get typesafe/jev and livekit/clef-omni decisions enabled for our project?

Hi all,

We’re evaluating LiveKit Inference decision models for a small yes/no classifier that runs on every user turn in our voice agent, and we’d like to compare Jev and Clef Omni against GPT-6 Luna on our test set.

openai/gpt-6-luna-decisions works for us through inference.DecisionModel. The other models return a 403 on the same endpoint with the same credentials:

APIStatusError (403 Forbidden) body={'type': 'not_enabled', 'error': 'decisions are not enabled for this project', 'hint': ''}

This happens for typesafe/jev, typesafe/jev-latest, typesafe/jev-preview, and livekit/clef-omni. The 1.8.9 release notes say decisions need a project feature flag, and PR livekit/agents#7855 mentions the flag is per provider, so I’m guessing our project has the OpenAI flag but not the TypeSafe and Clef ones.

Could someone enable those for our project, or point me at where to request it? Happy to DM the project ID if needed.

  • livekit-agents 1.8.10, inference.DecisionModel with the default gateway
  • One failing request id: 25e0a80d-da13-478c-add0-1bd74f8b1383

Thanks!

Hi Callan, those status errors are a bit misleading as there isn’t a flag we can just enable to unblock you, those models are not yet available to customers through LiveKit Inference.

Here’s what you can do today:

  • Jev: call it directly with the livekit-plugins-typesafe plugin and your own Typesafe API key. You can run it side by side with Luna on the same test set.
  • Clef Omni: Watch the Inference docs for when this gets added.

Sorry for the confusion, I’m sure this won’t be the case for much longer so this answer will go out of date very quickly.

Thanks for the update @darryncampbell. If you remember, would you be able to send a note in here when Clef Omni becomes available? We’ll be tracking this as well, but it can be challenging at times to keep up-to-date given the pace of LiveKit updates.

I’ve set a calendar reminder to reply here when it’s available. :bell:

I also suspect we will make an announcement in Announcements

Thanks for all your help!

@darryncampbell , i was also integrating decisions API. i can see that these all models are supported in types, but Inference pricing | LiveKit has no details about cost and mention of any those models?

Maybe it need new section as decisions API in inference to know about available models and cost.

GPT-6 Luna (Decisions) is on the pricing page, but I noticed when I launched the page it automatically filtered some of the models - if you use https://livekit.com/pricing/inference?llm_query=openai you will see it.

The other models will be added when they are officially supported. They are currently being internally tested, which is why they’re in that matrix.