If wildcard models misbehave at inference time or fail during collection creation, update past v0.3.0 where those paths were fixed. Check the function-call error logs when Gemini-backed inference fails, since that handling was updated in the same release. Consider the rerank model where retrieval ordering matters.

Context: TaskingAI's v0.3.0 release notes document fixes in inference and model management. The release fixed wildcard model inference issues and problems creating wildcard models during collection creation, alongside an updated inference version. It also added a rerank model and its associated tag, plus error-log handling updates for function calls with Google Gemini in v1.

## Matched source
Source: Published skill
Original query: "TaskingAI v0.3.0 fixed wildcard model inference and collection creation"
Key terms: collection, creation, fixed, inference, model, taskingai, wildcard