Story-1000: Align terminal run statuses

Story 1000 Epic 749🔗 Chain of CustodyAgent Platform3 files · +222/−6

PR #22 · pramaan-io/pramaan-agents · merged 2026-06-03

How it flows

sequenceDiagram
  autonumber
  participant guard as Guardrail
  participant reducer as Reducer
  participant state as agent_run.status
  guard->>reducer: terminal failure event
  reducer->>reducer: _run_failed_status classifies
  reducer->>state: aborted_cost / circuit_open
  state-->>reducer: replay
  reducer->>reducer: terminal flag → same final state

src/pramaan_agents/db/reducer.py

Method	in → out	what it does
`_run_failed_status`	`(payload) → str`	classifies a failure payload as aborted_cost or circuit_open

What changed, file by file

src/pramaan_agents/db/reducer.py+42 −6

Reducer that turns failure events into agent run state — adds the _run_failed_status helper plus aborted_cost and circuit_open terminal constants so a terminal flag sticks on replay.

new method	signature
`_run_failed_status`	`(payload) → str`

What does NOT happen

replay reopens a terminal run
double charge on circuit_open
cost-cap retried as fresh
audit vs runtime drift
silent re-classification

Replay is idempotent

Terminal stays terminal

Audit log matches runtime

aborted_cost is first-class

circuit_open is first-class

No silent re-classification

Files changed (3)

File	±
`docs/walkthroughs/story-1000-walkthrough.html`	+123
`src/pramaan_agents/db/reducer.py` ★	+42 −6
`tests/test_reducer_events.py`	+57