GitHub user weiqingy deleted a comment on the discussion: [Discussion][Observability 1/2] Recording Agent Traces in the Event Log
Could we make the `input_run_id` derivation contract explicit before locking in the schema? The model looks directionally strong to me: separating business-event causality from execution hierarchy makes the trace much more useful than lineage alone, especially for cases like “LLM succeeded but structured parsing failed.” One detail I am still unsure about is run identity. The proposal says the runtime creates or restores a run context, but it does not define how `input_run_id` is derived. In #841, there was concern that using `inputEvent.id` is not stable enough: Java may create a fresh UUID when wrapping an input on replay, while Python content-derived IDs can collapse distinct inputs with identical payloads. Would it make sense to define `input_run_id` from stable runtime coordinates, for example job/agent identity + business key + per-key input sequence, or another explicitly replay-aware source? That would also make future OTel export cleaner, since this field naturally maps to trace identity and would need stable semantics across restart/replay. GitHub link: https://github.com/apache/flink-agents/discussions/900#discussioncomment-17692457 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
