ISSUE DESCRIPTION Add reasoning.effort support for OpenAI gpt-realtime-2 Feature Type…
- User
ISSUE DESCRIPTION
Add reasoning.effort support for OpenAI gpt-realtime-2
Feature Type
Would make my life easier
Feature Description
OpenAI
gpt-realtime-2supports configurable reasoning viareasoning.effort, but the Python OpenAI Realtime plugin does not seem to expose this option onopenai.realtime.RealtimeModel.It would be useful to support something like:
from livekit.plugins import openai llm = openai.realtime.RealtimeModel( model="gpt-realtime-2", reasoning={"effort": "low"}, )Workarounds / Alternatives
I don't think there is a clean workaround through the current Python
RealtimeModelAPI.The only workaround I can see is patching the plugin locally to add
reasoningto the session update payload.Additional Context
OpenAI docs list these effort values:
minimal,low,medium,high,xhigh.References:
- https://developers.openai.com/api/docs/models/gpt-realtime-2
- https://developers.openai.com/api/docs/guides/realtime-models-prompting#set-reasoning-effort
- Related JS issue: https://github.com/livekit/agents-js/issues/1433
- Related JS PR: https://github.com/livekit/agents-js/pull/1575
Interface
No new interfaces are introduced.
ADDITIONAL ADVICE
Since you are given a git repository, you can use git commands to simplify your work. For example, if you made a mistake and want to revert your changes, you can use
git checkout <file>to restore the file to its original state. You can also reset all changes in the repository usinggit reset --hardcommand. Additionally, you can usegit stashandgit stash popto temporarily save your changes and restore them later. Finally, you can see the changes that you've made (which will be submitted when you callsubmit) usinggit statusorgit diffcommands. However you don't need to usegit addorgit commitbefore submitting your changes. In fact,submitonly submits currently unstaged changes because it usesgit diffto get the changes that need to be submitted.CHECKLIST
Before submitting your solution, please go over the following checklist and make sure you've done everything:
- If an error was reported in the issue description, I have successfully reproduced it.
- If an error was reported in the issue description, I have confirmed that my fix resolves the error.
- I have thought about edge cases and made sure my fix handles them as well.
- I have run existing tests in the repository that might have been affected by the change I made and confirmed that they pass. I want you to list every bullet from this checklist and write your reasoning for why you think you did it or didn't need to.
Repository has been uploaded and your shell is currently at the repository root. Time to solve the issue!
(Current directory: /agents, current file: none) bash-$
- Assistant
I'll start by exploring the repository structure and locating the OpenAI RealtimeModel implementation to add reasoning.effort support.
- Assistant
Next I'll inspect the OpenAI RealtimeModel source and how session updates are built.
- Assistant
rg is not available, let me use grep instead.
- Assistant
I'll read the RealtimeModel constructor and session-update building logic in detail.
- Assistant
Let me continue reading the init and session options
- Assistant
Let me continue reading the options dataclass and how session is created.
- Assistant
Now let me see the session update event creation and the with_azure method.
- Assistant
Let me look at _create_session_update_event and update_options, plus the imports at the top of the file.
- Assistant
I'll check whether the OpenAI SDK defines a reasoning type for realtime sessions.
- Assistant
I'll inspect the reasoning types and the imports used by the realtime model.
- Assistant
I'll review the full
update_optionsimplementations and how Azure session mapping handles related fields. - Assistant
I'll read the rest of session
update_optionsand how response create events are built. - Assistant
I'll check whether other OpenAI integrations already support reasoning, for consistency.
- Assistant
I'll look at how the responses LLM already wires reasoning options through.
- Assistant
I'll inspect the Tracing type and how similar optional configs are accepted in the constructor.
- Assistant
I'll verify whether RealtimeSessionCreateRequest coerces a reasoning dict.
- Assistant
I'll check the Azure session conversion and with_azure signature next.
- Assistant
I'll check how response create events are constructed and whether per-response reasoning fits.
- Assistant
Let me look at generate_reply
- Assistant
I'll quickly confirm there's no JS package in this repo to mirror.
- Assistant
I have a clear picture of the changes. I'll wire
reasoningthrough options, constructors, session updates, andupdate_options. - Assistant
I'll open the file at the import section and apply the edit there.
- Assistant
Let me apply multiple edits carefully.
- Assistant
I'll add a small helper to normalize reasoning and update
_RealtimeOptions. - Assistant
Next I'll add
reasoningto_RealtimeOptions. - Assistant
Need to navigate to the lines first.
- Assistant
Good, now I can replace.
- Assistant
I'll update both
__init__overloads to acceptreasoning. - Assistant
I'll update the Azure overload and the main
__init__signature next. - Assistant
I'll update the docstring and how
_optsis constructed. - Assistant
Now update the _opts assignment
- Assistant
I'll set
reasoningon_optsand also expose it viawith_azure. - Assistant
I'll update
with_azure, bothupdate_optionsmethods, and_create_session_update_event. - Assistant
Need to find the return RealtimeModel(...) inside with_azure
- Assistant
I'll update with_azure return and both update_options methods.
- Assistant
Large replace failed — I'll apply the change with a Python script instead.
- Assistant
Next I'll update session creation and RealtimeSession.update_options.
- Assistant
I'll verify how session updates are serialized and that the changes look correct.
- Assistant
I'll inspect send_event serialization to make sure reasoning is included when set.
- Assistant
I'll only attach reasoning when it's set, matching the truncation pattern.
- Assistant
I'll change reasoning to only be set when not None, matching truncation.
- Assistant
I'll check whether RealtimeSession shares or copies model options.
- Assistant
I'll run a quick script to confirm reasoning lands in the session update payload.
- Assistant
Session creation needs a running event loop, so I'll retest under asyncio.
- Assistant
I'll check whether converter helpers should live in
utils.pyfor consistency. - Assistant
I'll move
to_reasoninginto utils.py to match the other converters. - Assistant
I'll re-run the verification script after moving the helper.
- Assistant
I'll review the full diff and clean up any leftover issues.
- Assistant
I'll run a quick syntax/import check and any nearby unit tests.
- Assistant
I'll inspect one of the test errors to confirm they're environment-related, not from our change.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
claude-sonnet-5Compare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-35B-A3BCompare side by side →
Qwen/Qwen3.6-27BCompare side by side →
Qwen/Qwen3.5-35B-A3BCompare side by side →