From 8aa28415466d5840511e832ab7e514f3cdc34bd7 Mon Sep 17 00:00:00 2001
From: Daniel Roth <daniel_roth@hotmail.co.uk>
Date: Tue, 7 Jul 2026 15:46:00 +0000
Subject: [PATCH] =?UTF-8?q?Dispatch=20the=20abandon=20flow=20last=20for=20?=
 =?UTF-8?q?a=20fired=20deal-change=20=F0=9F=9F=A9?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 applications/abri/dispatch.py | 35 ++++++++++++++++++++++++++++++-----
 1 file changed, 30 insertions(+), 5 deletions(-)

diff --git a/applications/abri/dispatch.py b/applications/abri/dispatch.py
index f69981162..00f9223cc 100644
--- a/applications/abri/dispatch.py
+++ b/applications/abri/dispatch.py
@@ -1,11 +1,13 @@
 """Dispatches the flows named in a trigger message through the orchestrator.
 
 The scraper decides which flows fire; this is a dumb dispatcher. Flows run
-in a fixed, retry-safe order — amend, then log, then tenant sync — so that
-redelivery re-runs idempotent flows before non-idempotent ones: amend is
-naturally idempotent, log is guarded by the job-number check, and tenant
-contact creation (not idempotent) runs last so its failure can never cause
-a duplicate job log, and a log failure can never duplicate contacts.
+in a fixed, retry-safe order — amend, then log, then tenant sync, then
+abandon — so that redelivery re-runs idempotent flows before non-idempotent
+ones: amend is naturally idempotent, log is guarded by the job-number check,
+and tenant contact creation (not idempotent) runs before abandon so its
+failure can never cause a duplicate job log, and a log failure can never
+duplicate contacts. Abandon runs last so a cancellation can never interfere
+with a log, amend or tenant sync in the same batch.
 """
 
 from typing import Dict, Protocol
@@ -13,9 +15,11 @@ from typing import Dict, Protocol
 from applications.abri.abri_trigger_request import AbriTriggerRequest
 from domain.abri.models import AbriRequestRejected, PlaceRef
 from orchestration.abri_orchestrator import (
+    AbandonJobOrchestrationResult,
     AmendJobOrchestrationResult,
     AppointmentChange,
     ConfirmedSurveyBooking,
+    DealAbandonment,
     DealDatabaseGateway,
     LogJobOrchestrationResult,
     LogJobWriteBackError,
@@ -35,6 +39,10 @@ class AbriFlows(Protocol):
         self, place_ref: PlaceRef, deal_id: str
     ) -> TenantDataSyncResult: ...
 
+    def abandon_job(
+        self, abandonment: DealAbandonment
+    ) -> AbandonJobOrchestrationResult: ...
+
 
 class AbriFlowRejectedError(Exception):
     """An OpenHousing rejection, surfaced as a failure so the task fails.
@@ -66,6 +74,8 @@ def dispatch_abri_flows(
         summary["log_job"] = _run_log(request, flows, deal_database)
     if "sync_tenant_data" in request.flows:
         summary["sync_tenant_data"] = _run_tenant_sync(request, flows)
+    if "abandon_job" in request.flows:
+        summary["abandon_job"] = _run_abandon(request, flows)
 
     return summary
 
@@ -122,6 +132,21 @@ def _run_log(
     return f"job {outcome.job_no} logged"
 
 
+def _run_abandon(request: AbriTriggerRequest, flows: AbriFlows) -> str:
+    # Runs last so a cancellation can never interfere with a log, amend or
+    # tenant sync in the same batch. A missing job_no raises the retriable
+    # not-yet-recorded error, which propagates so the task fails and redelivers.
+    outcome = flows.abandon_job(
+        DealAbandonment(
+            deal_id=request.hubspot_deal_id,
+            outcome=request.outcome,
+        )
+    )
+    if isinstance(outcome, AbriRequestRejected):
+        raise AbriFlowRejectedError(flow="abandon_job", rejection=outcome)
+    return f"job {outcome.job_no} abandoned"
+
+
 def _run_tenant_sync(request: AbriTriggerRequest, flows: AbriFlows) -> str:
     if request.place_ref is None:
         raise ValueError("sync_tenant_data fired without a place_ref")