* Increase max connections
* Classify hard and transient hatchet errors
* Fan out partial success
* Force reprocessing of error transcripts
* Stop retrying on 402 payment required
* Avoid httpx/hatchet timeout race
* Add retry wrapper to get_response for for transient errors
* Add retry backoff
* Return falsy results so get_response won't retry on empty string
* Skip error status in on_workflow_failure when transcript already ended
* Fix precommit issues
* Fail step on first fan-out failure instead of skipping
* fix: switch structured output to tool-call with reflection retry
Replace the two-pass StructuredOutputWorkflow (TreeSummarize → acomplete)
with astructured_predict + reflection retry loop for structured LLM output.
- Enable function-calling mode (is_function_calling_model=True)
- Use astructured_predict with PromptTemplate for first attempt
- On ValidationError/parse failure, retry with reflection feedback
- Add min_length=10 to TopicResponse title/summary fields
- Remove dead StructuredOutputWorkflow class and its event types
- Rewrite tests to match new astructured_predict approach
* fix: include texts parameter in astructured_predict prompt
The switch to astructured_predict dropped the texts parameter entirely,
causing summary prompts (participants, subjects, action items) to be
sent without the transcript content. Combine texts with the prompt
before calling astructured_predict, mirroring what TreeSummarize did.
* fix: reduce TopicResponse min_length from 10 to 8 for title and summary
* ci: try fixing spawning job in github
* ci: fix for new arm64 builder
