Trace ↔ Graph Alignment¶
This page explains how Step Evidence Records (SER) align with the GraphV1 pipeline graph. The canonical mapping allows tools to reconstruct execution topology solely from SER files.
Node Identity¶
Every pipeline node is assigned a deterministic
node_uuid
by the GraphV1 builder.The same
node_uuid
appears: * In the canonical spec (canonical_spec.nodes[*].node_uuid
). * In each SER record asids.node_id
.Rule: Consumers must always join SER data to the graph using
node_uuid
.
SER Output¶
SER records contain minimal information. Optional fields may be supplied by the runtime and consumers should ignore unknown fields gracefully.
Pipeline & Run Identity¶
pipeline_id
: deterministic hash of the canonical spec.run_id
: unique per execution.Both appear in
pipeline_start
records and in theids
of each SER.
Trace → Graph Usage¶
To recreate the execution graph from SER files:
Read
pipeline_start
to obtain the canonical graph.For each SER record, use
ids.node_id
andtopology.upstream
to build edges.Combine with per-step information from
action
,io_delta
andchecks
.