Follow up to #1187
OpenAI specification has a WebSocket mode for /v1/responses API.
https://developers.openai.com/api/docs/guides/websocket-mode.
Websocket mode is also supported in agentic-api.
aiperf should support the following:
- Conversations maps one-to-one with WS connections. concurrency parameter still limits the number of in-flight requests at once.
- Can be used with traces and other modes that do not override server responses.
Note: stream_id multiplexing (running forked conversations in parallel over a single WebSocket connection) is not part of this issue. It has been split out into a follow-up: #1390. This issue covers the one-connection-per-conversation model only.
Follow up to #1187
OpenAI specification has a WebSocket mode for /v1/responses API.
https://developers.openai.com/api/docs/guides/websocket-mode.
Websocket mode is also supported in agentic-api.
aiperf should support the following: