openai /v1/realtime websocket support

[guenhter]

Hi,

I appreciate this repo very much. Makes my life easier :)

Now I'm integrating audio realtime in my app and it would be nice if the mock has some baisc realtime capabilities.
What do you think? Is this something aimock should support?

I'm also more then willing to contribute this myself if you want to.

[jpr5]

I love contributions, and, if you want me to do it instead, that's also fine too. I'd just need you to explain a little bit more about what you're expecting. But - absolutely would welcome more contributors. If you do make a PR, just make sure to enable maintainers can contribute so we can collaborate on it together. Let me know!

…

On Thu, May 7, 2026 at 6:23 AM, Günther Grill < ***@***.*** > wrote: *guenhter* created an issue (CopilotKit/aimock#159) ( #159 ) Hi, I appreciate this repo very much. Makes my life easier :) Now I'm integrating audio realtime in my app and it would be nice if the mock has some baisc realtime capabilities. What do you think? Is this something aimock should support? I'm also more then willing to contribute this myself if you want to. — Reply to this email directly, view it on GitHub ( #159 ) , or unsubscribe ( https://github.com/notifications/unsubscribe-auth/AAAVHLIMN2VZXSMZG7F2ECL4ZSE3JAVCNFSM6AAAAACYUTBXOCVHI2DSMVQWIX3LMV43ASLTON2WKOZUGM4TSMBYGUYTAMA ). You are receiving this because you are subscribed to this thread. Message ID: <CopilotKit/aimock/issues/159 @ github. com>

[guenhter]

Nice.

Actually what I need is basically this endpoint: https://developers.openai.com/api/docs/guides/realtime-websocket

I'll give it a try in the next days hopefully and prepare a PR for you.

[jpr5]

Hey @guenhter — good news, aimock already supports the OpenAI /v1/realtime WebSocket API!

How to use it:

1. Start aimock with your fixtures as usual
2. Point your OpenAI client's base URL at the aimock server
3. Connect via WebSocket to ws://<aimock-url>/v1/realtime

The full protocol is implemented — session.update, conversation.item.create, response.create, and the complete response event sequence (text deltas, tool calls, etc.). Same fixtures work for both HTTP and WebSocket.

Docs: WebSocket APIs covers the setup and event flow.

One caveat: Currently supports text modality only (no audio binary frame streaming). If you specifically need raw audio frames over the WebSocket, that's not there yet — but the Realtime API protocol structure, session management, and response events are all fully mocked.

Let me know if you hit any issues!

[guenhter]

That is awesome. I completely missed that. I'll integrate that next week and close the issue then.

[jpr5]

Perfect, lmk!