I contributed #6382 earlier this year (the ValueError raised while parsing streamed tool calls in the evals judge), which was merged in July. I would like to do more here, but on something you actually want rather than something I picked on my own.
CONTRIBUTING asks for an issue before starting on anything sizeable, so this is that issue.
I can commit real time over the next couple of months to one substantial piece of work, with tests and docs, and I will stay on it through review rather than dropping it after the first round. Areas I am comfortable in: the voice pipeline and its timing and interruption behaviour, tool calling and streamed LLM output, plugin implementations for STT/TTS providers, and test coverage for any of the above.
Is there something on your list that an outside contributor could take? A flaky area, a plugin you want brought up to the current interface, a subsystem that needs tests. If nothing fits right now, that is a fine answer and I will not keep pinging.
I contributed #6382 earlier this year (the ValueError raised while parsing streamed tool calls in the evals judge), which was merged in July. I would like to do more here, but on something you actually want rather than something I picked on my own.
CONTRIBUTING asks for an issue before starting on anything sizeable, so this is that issue.
I can commit real time over the next couple of months to one substantial piece of work, with tests and docs, and I will stay on it through review rather than dropping it after the first round. Areas I am comfortable in: the voice pipeline and its timing and interruption behaviour, tool calling and streamed LLM output, plugin implementations for STT/TTS providers, and test coverage for any of the above.
Is there something on your list that an outside contributor could take? A flaky area, a plugin you want brought up to the current interface, a subsystem that needs tests. If nothing fits right now, that is a fine answer and I will not keep pinging.