Files
archie/internal/infrastructure/postgres/queries/steps.sql
T
sam b8cb7145a6 feat(workflow): agent-call children, call steps, and the waited-on callee sweep
Plan item 4 of docs/prds/execution-tree-state-machine.md (archie-core-1216),
plus the callee half of CancelExecution that plan item 5 deferred until the
tree could decide it.

Every agent call a stage makes records itself: AgentStage's bridge, the
baseline-fix auto-fix run, the adversarial reviewer and a workflow.call each
start a child StepExecution of kind agent or call under the stage's own step
-- TaskContext.StepID, which the engine sets before the stage body starts and
clears when it returns -- and finish it with the call's own outcome and the
tokens it accounted. A child's recording write failure parks the execution
the same way a stage's does: an agent call that ran unrecorded is the
invisibility the step table exists to end. A stage invoked outside the
engine's recorded run -- the stage unit tests -- records no child and runs
unchanged.

The call step is what makes the callee half of the PRD's Cancellation section
decidable, and this lands it: a call step closes when its callee ends, so an
open call step is a callee the caller waits on (wait:true) and a closed one
is a wait:false callee that runs on. CancelExecution reads the open call
steps before its sweep closes them and cancels each named callee through its
own CancelExecution in the caller's write's aftermath -- the callee's own
steps, its own events, its own move to the same destination, and the callee's
own wait:true callees beyond it, the depth the enqueue enforces bounding the
recursion. A callee that cannot be cancelled is reported and the caller's own
record stands.

StartStep refuses a call step whose callee is not this run's own (the
callee's call_parent_task_id must name the recording execution): a
compromised container could otherwise record a step naming a foreign
execution and cancel it. StartStepRequest gains called_execution_id (field 6;
non-breaking, the column and its named consumer already exist) and
StepStart carries it.

PRD decisions recorded here:
- The reviewer records as one agent child named "review": the Reviewer
  contract runs one subagent per review, the PRD model's review:1/review:2
  are illustrative of depth-1 children. A completed report closes it
  succeeded, a not-run review failed with the skip reason; the contract
  carries no usage, so tokens are 0.
- A failed-callee call step closes as failed with the callee's detail when
  the wait ends otherwise; the single open case is the caller itself being
  cancelled, whose CancelExecution already swept this step -- a late write
  must not pretend the wait is still the caller's business.
- A call step's callee is cancelled to the same destination the caller's
  action named (parked stays parked, closed_wont_do ends the whole chain),
  and a callee that cannot be cancelled is reported while the caller's own
  record stands either way.
- The agent child's detail is the result summary on success and the error's
  text on failure, with the call's own TokensUsed -- the PRD's only usage
  field, untracked per call for the reviewer contract.
2026-09-27 07:14:15 +10:00

82 lines
4.2 KiB
SQL

-- Step-execution queries (docs/prds/execution-tree-state-machine.md). The
-- store method owns the transaction that pairs each step transition with its
-- event row; these are only the guarded row writes and reads.
-- name: LockExecutionForStep :one
-- FOR UPDATE locks the execution's row, so the status check, the step insert
-- and the event append are one serialized operation per execution -- the way
-- guardTransition locks a task's row for its transition. The row also carries
-- every field the transition's event needs (org, workspace, the task
-- identity) and the attempt the step must belong to.
SELECT id, status, org_id, workspace_id, owner, repo, issue_number, workflow, attempt
FROM tasks WHERE id = $1 FOR UPDATE;
-- name: InsertStepExecution :one
INSERT INTO step_executions (org_id, workspace_id, execution_id, attempt, parent_id, depth, kind, name, called_execution_id, status)
VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, 'pending')
RETURNING id;
-- name: StepCalleeParent :one
-- The caller linkage a call step's callee must carry: the callee's
-- call_parent_task_id names the execution whose tree records the call. A
-- call step naming a task that is not this execution's callee is refused,
-- because a cancel of that execution would otherwise cancel work the
-- caller never started.
SELECT call_parent_task_id FROM tasks WHERE id = $1;
-- name: StartStepExecution :execrows
-- The pending -> running half of StartStep, guarded by the row's own status:
-- the step transition table routes the pair, so a row that is not pending is
-- stale rather than silently restarted.
UPDATE step_executions SET status = 'running', started_at = now()
WHERE id = $1 AND status = 'pending';
-- name: LockStepExecution :one
-- FOR UPDATE, shared by a start's parent check and a finish's guarded write:
-- the row is held for the caller's transaction, so the state a check reads is
-- the state the guarded write sees.
SELECT id, org_id, workspace_id, execution_id, attempt, parent_id, depth, kind, name, called_execution_id, status, detail, tokens_used, started_at, finished_at
FROM step_executions WHERE id = $1 FOR UPDATE;
-- name: FinishStepExecution :execrows
UPDATE step_executions
SET status = $1, detail = $2, tokens_used = $3, finished_at = now()
WHERE id = $4 AND status = $5;
-- name: CancelAttemptSteps :many
-- The half of CancelExecution that cancels the current attempt's
-- non-terminal steps, returning them so the store writes one event row per
-- transition in the same transaction. Earlier attempts' rows never match:
-- history is never rewritten.
UPDATE step_executions
SET status = 'cancelled', finished_at = now()
WHERE execution_id = $1 AND attempt = $2 AND status IN ('pending', 'running')
RETURNING id, name;
-- name: LockRunningExecutions :many
-- The executions a crashed or replaced daemon left running, locked for the
-- recovery transaction so the step sweep and the requeue cannot race a
-- concurrent write. The row carries the identity the steps' events need.
SELECT id, org_id, workspace_id, owner, repo, issue_number, workflow, attempt
FROM tasks WHERE status = 'running' FOR UPDATE;
-- name: InterruptAttemptSteps :many
-- The crash-recovery step edge (docs/prds/execution-tree-state-machine.md):
-- the steps an interrupted execution left running move to the table's
-- interrupted outcome, guarded by the row's own status so only the pair the
-- step transition table routes is written. Returning them lets the store
-- write one event row per transition in the same transaction.
UPDATE step_executions
SET status = 'interrupted', finished_at = now()
WHERE execution_id = $1 AND attempt = $2 AND status = 'running'
RETURNING id, name;
-- name: WaitingCallSteps :many
-- The call steps of the execution's current attempt that are still open and
-- name a callee: a call step closes when the callee ends, so an open one is
-- a callee the caller waits on (wait:true). The callee sweep of CancelExecution
-- walks these, because a callee whose call step already closed -- wait:false,
-- or one that ended before the cancel arrived -- runs on.
SELECT id, called_execution_id FROM step_executions
WHERE execution_id = $1 AND attempt = $2
AND kind = 'call' AND status IN ('pending', 'running') AND called_execution_id <> 0;