Problem
OutputContext.fail() returns codes like UPLOAD_FAILED, INIT_METHOD_REQUIRED with a retryable flag — excellent foundation. But there's no error catalog documenting which errors are transient vs permanent.
Details
upload.ts catches errors generically — agents get UPLOAD_FAILED with a message string but no breakdown (gas issue? provider rejection? PDP failure?)
- Without classification, agents can't make intelligent retry/abort decisions
- The
retryable flag exists but isn't consistently documented
Ask
- Document common error codes with descriptions
- Classify each as transient (retry) vs permanent (abort/escalate)
- Add recommended retry strategies (backoff, max attempts)
- This enables agents to handle failures programmatically instead of just surfacing error strings
Problem
OutputContext.fail()returns codes likeUPLOAD_FAILED,INIT_METHOD_REQUIREDwith aretryableflag — excellent foundation. But there's no error catalog documenting which errors are transient vs permanent.Details
upload.tscatches errors generically — agents getUPLOAD_FAILEDwith a message string but no breakdown (gas issue? provider rejection? PDP failure?)retryableflag exists but isn't consistently documentedAsk