source_change or scheduled_research_trigger -> Analysis -> Signals -> Routing -> Targets -> Tasks -> Execution Decision -> AI/Human branch -> Result Event -> Feedback -> Evaluation
The debugger should teach the chain first, then let you inspect one trace and one stage at a time.
This page now emphasizes the runtime chain and the selected trace rather than leading with a wide technical table.
The main path is fixed and explicit: source change or scheduled research enters one shared chain, then AI and human execution branch and merge back into result, feedback, and evaluation.
Use explicit filters and pager controls; mobile no longer depends on hover or table scan to debug a trace.
Trace rows are now card-based so you can scan anchor, state, completeness, and reason quickly.
Select one stage to inspect evidence, provenance, lineage, and semantic audit.