bdsqqq/dots/emitting-context-rich-events

CONTEXT-RICH EVENTS \\

emit context-rich events at the granularity of significant operations.

source: igorbedesqui.com pattern language.

consider first: WORKFLOW = DATA

forces: autonomous agents act on your behalf — you need to reconstruct what happened after the fact. scattered log lines optimized for writing (log("payment failed")) can't answer the questions you actually ask during debugging: "which workflow, for which issue, at what step, with what inputs?"

low-context logs force you to grep across files, correlate by timestamp, and hope the right fields were included. they answer "something happened" but not "what happened to THIS request." the more autonomous the system, the wider the gap between what was logged and what you need to know.

therefore: emit context-rich events at the granularity of significant operations. each event carries the full context needed to understand that operation in isolation — identifiers, business state, timing, outcome. one event per operation, not many log lines scattered across the call stack. transport (console, structured platform, both) is an orthogonal decision.

intended consequences:

• ✓ debugging becomes querying — filter by workflow, step, outcome
• ✓ events are self-contained — no cross-referencing needed
• ✓ high-cardinality fields (workflow ID, issue ID) enable precise queries
• △ requires discipline — every call site must attach context, not just

a message string

• △ event size grows with context; cost scales with dimensionality

consider next: HUMAN-FRIENDLY IDENTIFIERS