Aaron Hill
Aaron Hill
This is tricky, as we currently initialize the Rust-side logger when the Python native module is loaded
The gemini model thinking behavior is pretty poorly documented, but I don't think they actually support this. The only thoughts that you see as output are "thought summaries" - the...
It looks like thought signatures are only supported on AI Studio, not on GCP Vertex
Done in: * https://github.com/tensorzero/tensorzero/pull/2812 * https://github.com/tensorzero/tensorzero/pull/2808 * https://github.com/tensorzero/tensorzero/pull/2803
We need to pass a loader to Python: https://github.com/mitsuhiko/minijinja/pull/736
For now, let's just warn and continue in notebooks where we encounter an unknown content block type
Done in https://github.com/tensorzero/tensorzero/pull/4923
The model inference cache fixtures need to be regenerated
/regen-fixtures
Using the same concurrency key for both workflows prevents general-checks and merge-queue checks from running in parallel.