rogee/gochat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(automation): deliver retryable external actions

Browse Source
This commit is contained in:
rogee
2026-06-05 09:54:50 +08:00
parent 72e980def9
commit 382933e601
7 changed files with 874 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/CHATWOOT_PARITY_DEVELOPMENT_PLAN.md
+35 -16
View File
@@ -16,9 +16,9 @@ Build GoChat as a Go backend that can directly reuse the frontend from `referenc
## Current Baseline
- Latest implementation checkpoint: `4e28559 feat(automation): record rule execution outcomes`.
- Latest documentation checkpoint: B9.1d delivery landing plan recorded in this tracker.
- Worktree status at this documentation checkpoint: B9.1c execution outcome logging is implemented and committed; next active slice is B9.1d retryable webhook/email transcript action delivery, followed by B9.2 macro execution parity.
- Latest implementation checkpoint: `feat(automation): deliver retryable external actions`.
- Latest documentation checkpoint: this checkpoint, recorded with the B9.1d implementation.
- Worktree status at this implementation checkpoint: B9.1d retryable webhook/email transcript delivery is implemented; next active slice is B9.2 macro CRUD and execution parity.
- `go test ./...` passes.
- Route dump succeeds with `TOTAL: 830` after adding the Chatwoot-compatible applied-SLA index route.
- Route parity artifacts now exist under `docs/parity/` and are generated by `cmd/route_parity`.
@@ -44,11 +44,10 @@ Next ordered checkpoints:
| Order | Slice | Required outcome | Primary verification |
| --- | --- | --- | --- |
| 1 | B9.1d | Webhook and email transcript actions are timeout-bound, retryable, and observable through the worker/action boundary. | Fake HTTP/mailer tests and log/retry metadata assertions. |
| 2 | B9.2 | Macro CRUD and macro execute side effects match Chatwoot frontend expectations. | Macro handler/service tests that reload conversations/messages after execution. |
| 3 | B10 | Audit, CustomRole, and remaining InboxLimit surfaces enforce Chatwoot admin behavior. | Permission matrix tests, audit writer/list tests, inbox/account limit tests. |
| 4 | B11 | Captain/Copilot enterprise screens have real persistence and safe LLM feature gates. | Captain/Copilot handler/service fixtures and disabled-state tests. |
| 5 | B12 | Reused Chatwoot frontend smoke runs repeatably against GoChat. | Checked smoke command plus gap report under `docs/parity/`. |
| 1 | B9.2 | Macro CRUD and macro execute side effects match Chatwoot frontend expectations. | Macro handler/service tests that reload conversations/messages after execution. |
| 2 | B10 | Audit, CustomRole, and remaining InboxLimit surfaces enforce Chatwoot admin behavior. | Permission matrix tests, audit writer/list tests, inbox/account limit tests. |
| 3 | B11 | Captain/Copilot enterprise screens have real persistence and safe LLM feature gates. | Captain/Copilot handler/service fixtures and disabled-state tests. |
| 4 | B12 | Reused Chatwoot frontend smoke runs repeatably against GoChat. | Checked smoke command plus gap report under `docs/parity/`. |
## Execution Snapshot
@@ -140,12 +139,13 @@ This ledger records the committed parity checkpoints that future slices should b
| `feat(automation): align rule trigger coverage` | Completed B9.1b listener/event coverage: automation rules now honor Chatwoot skip rules for automation-origin events, auto-reply conversations, activity messages, and auto-reply emails; listener extraction works from event conversation ID, `conversation_id`, conversation payloads, and message payloads; conversation update/status/priority paths dispatch `changed_attributes`; initial conversation messages and provider webhook messages carry the message/conversation data needed by automation. | `go test ./internal/automation -run 'AutomationRuleListener\|MatchAndExecute\|Condition' -count=1`; `go test ./internal/service -run 'Conversation.*Automation\|Message.*Automation\|ConversationService' -count=1`; `go test ./internal/handler/webhook -run 'Incoming\|Webhook' -count=1`; `go test ./internal/automation -count=1`; `go test ./internal/service -count=1`; `go test ./internal/handler/webhook -count=1`; `go test ./...`; `git diff --check`. No route changes; route dump remains `TOTAL: 830`. | Continue B9.1c execution logs and stop-on-match behavior, then B9.1d retryable webhook/email transcript actions and B9.2 macro parity. |
| `feat(automation): record rule execution outcomes` | Completed B9.1c observability/parity guard: rule evaluations now record skipped condition outcomes, condition failures, event names, aggregate action counts, and per-action success/failure results. Matching rules continue executing in ID order without stop-on-match, matching the current Chatwoot listener loop, and failed actions do not prevent later actions from running. | `go test ./internal/automation -run 'AutomationRuleService_MatchAndExecute\|ExecutionLogService' -count=1`; `go test ./internal/automation -count=1`; `go test ./internal/handler/webhook -run 'Incoming\|Webhook' -count=1`; `go test ./internal/service -run 'Conversation.*Automation\|Message.*Automation\|ConversationService' -count=1`; `go test ./...`; `git diff --check`. No route changes; route dump remains `TOTAL: 830`. | Continue B9.1d retryable webhook/email transcript actions, then B9.2 macro payload/execute parity. |
| `docs: land automation delivery plan` | Recorded the complete B9.1d retryable external-action landing contract: webhook payload/delivery, transcript mailer boundary, retry/timeout behavior, action-result observability fields, B9.3 durable-worker deferral, and exit commands. It also refreshed stale B9 tracker entries that still pointed at B9.1b. | Documentation-only checkpoint; `git diff --check` before commit. | Start B9.1d implementation, then continue B9.2 macro CRUD/execute parity. |
| `feat(automation): deliver retryable external actions` | Completed B9.1d external action delivery: automation webhook actions now build Chatwoot-style `automation_event.<event>` payloads and deliver through a timeout-bound 3-attempt HTTP boundary; email transcript actions now split comma recipients, build transcript subject/body, deliver through a fakeable SMTP/mailer boundary, and both action types write delivery metadata into `action_results`. Chatwoot array params now round-trip `send_webhook_event` URLs. | `go test ./internal/automation -run 'ActionService\|AutomationRuleService_MatchAndExecute\|ExecutionLogService' -count=1`; `go test ./internal/automation -count=1`; `go test ./...`; `git diff --check`. No route changes; route dump remains `TOTAL: 830`. | Continue B9.2 macro CRUD and macro execute side-effect parity; durable queued worker scheduling remains B9.3/Phase 5. |
## Next Slice Contract
Completed implementation slice: B5.1-B5.5 now cover inbox serializer shape, Chatwoot frontend create/update binding, working-hours persistence, out-of-office behavior, inbox member assignment payload/mutation semantics, channel-specific config depth, AgentCapacityPolicy/InboxCapacityLimit API data contracts, and dedicated Email/Twilio/LINE channel route response shapes. B6 is now in Review after `a16c23c` and `f08c743`: Chatwoot search response envelopes, frontend query params, Meilisearch sender filters, mocked hit serialization, env-gated live Meilisearch validation, release-mode DB fallback rejection, and reindex Meilisearch-only guard are covered. B7 is now in Review after `a98dc2c`, `95224fa`, `a11bb96`, `213bf2b`, `4f85ef1`, and `d23f3f0`, enforcing per-inbox capacity limits, aligning SLA policy CRUD payloads, persisting applied SLA records, making core FRT/NRT/RT breach events idempotent, aligning the applied-SLA reports API, and creating SLA miss notifications. B3 and B4 remain in review for deeper side effects and browser validation.
Next implementation slice: start B9.1d retryable webhook/email transcript action delivery now that B9.1a CRUD payloads, B9.1b listener coverage, and B9.1c execution outcome logs are committed.
Next implementation slice: start B9.2 macro CRUD and macro execution parity now that B9.1a CRUD payloads, B9.1b listener coverage, B9.1c execution outcome logs, and B9.1d retryable external actions are implemented.
| Step | Required result | Reference source | Verification |
| --- | --- | --- | --- |
@@ -162,7 +162,7 @@ Next implementation slice: start B9.1d retryable webhook/email transcript action
| N11 | Keep B9.1a automation-rule CRUD payload parity as current automation API baseline. | Chatwoot automation controller/Jbuilder views, dashboard automation API/store/helper files, current Go handler/service/validator. | Done by `feat(automation): align automation rule payloads`; frontend envelopes, `attribute_key`, operator aliases, array action params, account-scoped update/delete/clone, and empty delete are covered. |
| N12 | Keep B9.1b automation listener trigger coverage as current event baseline. | `AutomationRuleListener`, `BaseListener`, conversation/message service dispatch, provider webhook persister. | Done by `feat(automation): align rule trigger coverage`; reference listener events, skip rules, changed attributes, and provider message reachability are covered. |
| N13 | Keep B9.1c execution outcome logging as current automation observability baseline. | Chatwoot automation listener/action service rescue behavior and Go execution log service. | Done by `feat(automation): record rule execution outcomes`; skipped/failed/success action results, event names, and no-stop-on-match behavior are covered. |
| N14 | Implement B9.1d external action delivery boundary before macro expansion. | Chatwoot `AutomationRules::ActionService`, `WebhookJob`, transcript mailer path, current Go action service. | Webhook/email transcript actions must be timeout-bound, retryable, and observable through action result metadata with fake HTTP/mailer tests. |
| N14 | Keep B9.1d external action delivery boundary as current action-delivery baseline. | Chatwoot `AutomationRules::ActionService`, `WebhookJob`, transcript mailer path, current Go action service. | Done by `feat(automation): deliver retryable external actions`; webhook/email transcript actions are timeout-bound, retryable, observable in action results, and covered by fake HTTP/mailer tests. |
| N15 | Update this tracker after every implementation checkpoint. | This document. | `git diff --check`; `go test ./...` for Go changes. |
Current B2 profile checkpoint:
@@ -384,7 +384,7 @@ Upcoming enterprise task boards:
| B8 | B8.2 | Align public CSAT submit/update flow, lock window, message linkage, and one-response-per-message behavior. | `reference/chatwoot/app/controllers/public/api/v1/csat_survey_controller.rb`, CSAT response model/services. | Public handler tests for submit, update within window, update after lock, duplicate response rejection. | Done by `ef3a909` |
| B8 | B8.3 | Wire resolve-triggered CSAT survey send with idempotency and channel-aware message creation. | Chatwoot CSAT listener/jobs and inbox CSAT settings. | Listener tests prove one CSAT message per resolved conversation and no send when disabled. | Done by `ef3a909`; WhatsApp/Twilio external template dispatch remains integration follow-up |
| B8 | B8.4 | Align CSAT downloads/export payloads used by reports screens. | CSAT report/download controller paths and frontend report API. | CSV/download tests cover filters and frontend column names. | Done by `b36cf07` |
| B9 | B9.1 | Align automation rule CRUD payloads, validation, condition groups, event names, listener skip rules, event data, execution outcomes, and external action delivery. | Chatwoot automation rule controllers/models, `AutomationRuleListener`, action services/jobs, and dashboard automation builder. | Handler/service/listener tests for CRUD, validation, event mapping, skip rules, changed attributes, provider-dispatched events, execution logs, and retry metadata. | Doing; B9.1a done by `3403770`, B9.1b done by `feat(automation): align rule trigger coverage`, B9.1c done by `4e28559`, B9.1d is active next |
| B9 | B9.1 | Align automation rule CRUD payloads, validation, condition groups, event names, listener skip rules, event data, execution outcomes, and external action delivery. | Chatwoot automation rule controllers/models, `AutomationRuleListener`, action services/jobs, and dashboard automation builder. | Handler/service/listener tests for CRUD, validation, event mapping, skip rules, changed attributes, provider-dispatched events, execution logs, and retry metadata. | Review; B9.1a done by `3403770`, B9.1b done by `feat(automation): align rule trigger coverage`, B9.1c done by `4e28559`, B9.1d done by `feat(automation): deliver retryable external actions` |
| B9 | B9.2 | Implement automation action side effects for labels, status, priority, assignee/team, custom attributes, private notes, and snooze. | Chatwoot automation action services/listeners. | Rule listener tests assert real conversation mutations and stop-on-match behavior. | Todo |
| B9 | B9.3 | Add delayed action scheduling and durable worker parity after the B9.1d synchronous retry boundary lands. | Chatwoot automation jobs, delayed action handling, and Phase 5 worker plan. | Worker tests cover queued delayed actions, retries, and observable failures. | Todo |
| B9 | B9.4 | Align macro CRUD/availability and macro execution side effects. | Chatwoot macros controller/model/action execution. | Macro handler/service tests cover frontend payloads and conversation mutations. | Todo |
@@ -415,7 +415,7 @@ B9 automation and macro execution breakdown:
| B9.1a | Align automation-rule CRUD request/response envelopes, condition/action validation, and rule status toggles. | Chatwoot automation rules controller/model and dashboard automation client. | Handler tests for index/show/create/update/delete/toggle payloads and validation failures. | Done by `feat(automation): align automation rule payloads` |
| B9.1b | Implement real rule trigger coverage for Chatwoot automation events: conversation created/updated/opened/resolved and message created. Provider webhook persistence must dispatch into those same conversation/message events. `contact_updated` is not a current Chatwoot automation trigger and stays out unless the local reference/frontend proves otherwise. | `reference/chatwoot/app/listeners/automation_rule_listener.rb`, `reference/chatwoot/app/listeners/base_listener.rb`, existing `channel.Dispatcher` events. | Listener tests prove matching rules execute once per event with account/inbox filters, changed attributes, provider-event reachability, and Chatwoot skip rules. | Done by `feat(automation): align rule trigger coverage` |
| B9.1c | Persist execution logs and stop-on-match behavior, including failed action records. The current Chatwoot listener runs every matching rule, so Go explicitly preserves no-stop-on-match behavior and records it in tests. | `reference/chatwoot/app/listeners/automation_rule_listener.rb`, `reference/chatwoot/app/services/automation_rules/action_service.rb`, Go execution log service. | Service tests cover success, skipped, failed/partial, per-action result records, and all-matching-rules execution. | Done by `feat(automation): record rule execution outcomes` |
| B9.1d | Make webhook and email transcript actions timeout-bound, retryable, and observable. | Chatwoot action services/jobs and Phase 5 worker plan. | Tests use fake HTTP/mailer boundaries and assert retry/log metadata. | Active next |
| B9.1d | Make webhook and email transcript actions timeout-bound, retryable, and observable. | Chatwoot action services/jobs and Phase 5 worker plan. | Tests use fake HTTP/mailer boundaries and assert retry/log metadata. | Done by `feat(automation): deliver retryable external actions` |
| B9.2a | Align macro CRUD payloads, visibility rules, and validation. | Chatwoot macros controller/model and dashboard macro client. | Handler tests for raw macro payloads, ownership, account scoping, and delete status. | Todo |
| B9.2b | Execute macro actions against real conversations: labels, status, priority, assignee, team, custom attributes, private notes, and attachments where supported. | Chatwoot macro execution service and conversation/message controllers. | Macro execute tests reload conversation/message rows and assert Chatwoot-shaped response payloads. | Todo |
@@ -500,7 +500,7 @@ B9.1b current checkpoint:
- Listener context extraction now supports `event.ConversationID`, `event.Data["conversation_id"]`, `event.Data["conversation"]`, and `event.Data["message"]`, with account fallback from conversation/message payloads.
- Conversation update/status/priority paths now dispatch Chatwoot-style `changed_attributes` so `attribute_changed` automation conditions can match status, priority, and SLA policy changes.
- Conversation creation with an initial message now dispatches `message.created` with `ConversationID`, `ContactID`, message payload, and conversation payload; provider webhook persistence is covered by an automation regression proving incoming provider messages trigger `message_created` rules once.
- B9 remains `Doing` because B9.1c execution log/stop-on-match semantics, B9.1d retryable webhook/email transcript actions, and B9.2 macro parity are still active.
- B9 remains `Doing` because B9.2 macro parity and B9.3 delayed/durable worker parity are still active.
B9.1b verification:
@@ -522,7 +522,7 @@ B9.1c current checkpoint:
- Condition matcher errors are logged as failed rule evaluations instead of disappearing into process logs only.
- Failed actions record per-action errors while later actions still run, matching Chatwoot `AutomationRules::ActionService#perform` rescue-and-continue behavior.
- The current Chatwoot listener loops through every matching rule; Go preserves this no-stop-on-match behavior and has regression coverage proving two matching rules both execute.
- B9 remains `Doing` because B9.1d retryable webhook/email transcript actions and B9.2 macro parity are still active.
- B9 remains `Doing` because B9.2 macro parity and B9.3 delayed/durable worker parity are still active.
B9.1c verification:
@@ -562,6 +562,24 @@ env GOCACHE=/tmp/gochat-gocache GOMODCACHE=/tmp/gochat-gomodcache go test ./...
git diff --check
```
B9.1d current checkpoint:
- `send_webhook_event` is no longer a no-op. It accepts Chatwoot array params as `url`, builds a conversation webhook payload with `event: automation_event.<rule event>`, latest message data, conversation IDs/status/priority/attributes, and posts through a timeout-bound HTTP deliverer.
- The HTTP webhook deliverer retries retryable failures up to three attempts by default, treats `5xx` and `429` as retryable, records response code/body, and is injectable for tests.
- `send_email_transcript` is no longer a no-op. It parses comma-separated recipients, builds a Chatwoot-style `[#display_id] Conversation Transcript` subject plus transcript body, and delivers per recipient through a fakeable SMTP transcript boundary.
- `ActionService.ExecuteWithResult` preserves the existing `Execute` API while returning per-action metadata for automation execution logs.
- `AutomationExecution.action_results` now includes optional `delivery_type`, `target`, `attempts`, `response_code`, `response_body`, and `retryable` fields. Failed external actions still allow later actions/rules to continue under the B9.1c behavior.
- B9.1 moves to `Review`. Durable queued scheduling and delayed action parity remain B9.3/Phase 5; the next active implementation slice is B9.2 macro CRUD and execution parity.
B9.1d verification:
```bash
env GOCACHE=/tmp/gochat-gocache GOMODCACHE=/tmp/gochat-gomodcache go test ./internal/automation -run 'ActionService|AutomationRuleService_MatchAndExecute|ExecutionLogService' -count=1
env GOCACHE=/tmp/gochat-gocache GOMODCACHE=/tmp/gochat-gomodcache go test ./internal/automation -count=1
env GOCACHE=/tmp/gochat-gocache GOMODCACHE=/tmp/gochat-gomodcache go test ./...
git diff --check
```
B10 admin and limits breakdown:
| Step | Implementation target | Reference source | Required tests | Status |
@@ -687,7 +705,7 @@ Remaining slice landing plan:
| B6 | Compare current Meilisearch document fields and global/entity search payloads against Chatwoot frontend consumers. | Add optional live Meilisearch integration gate, tighten account-scoped filters, and document DB fallback as development-only. | Done only after mocked tests and live-shape review prove search payloads work without DB-only assumptions. |
| B7 | Done: assignment capacity enforcement, SLA policy CRUD payload parity, applied-SLA conversation persistence, idempotent SLA breach lifecycle, applied-SLA report/list payload parity, and SLA miss notification fan-out. | Capacity-aware assignment policy binding review and durable SLA job scheduling remain broader Phase 5/P4.9 follow-ups, not hidden SLA gaps. | Review after `d23f3f0`; move to Done only after assignment policy binding review is either implemented or split out of B7. |
| B8 | Done: account-side CSAT list/metrics/review-note payload parity, public submit/update idempotency, resolve-triggered generic survey message send, and CSV download filters/columns. | Channel-specific WhatsApp/Twilio template send hooks and durable queueing are tracked under Phase 5. | Review after `b36cf07`; move to Done only after channel-specific send hooks are either implemented or formally split as external integration scope. |
| B9 | B9.1d retryable webhook/email transcript delivery, then B9.2 macro CRUD and macro execution side effects. | Automation rule condition/action parity, execution logs, webhook delivery retry, email transcript delivery, delayed actions, and macro action mutation parity. | Done only after Chatwoot listener events are covered, actions mutate real conversations, external actions are retryable/observable, and macro flows pass frontend payload tests. |
| B9 | B9.2 macro CRUD and macro execution side effects. | Macro payload parity, action mutation parity, delayed actions, and durable queued worker follow-ups. | Done only after Chatwoot listener events are covered, actions mutate real conversations, external actions are retryable/observable, macro flows pass frontend payload tests, and durable worker gaps are either implemented or explicitly split. |
| B10 | Audit list payload and audit writer boundary for representative mutating core resources. | CustomRole permission-key parity, AccountUser permission resolution, InboxLimit enforcement in inbox/channel creation paths. | Done only after authorization tests prove admin-only surfaces and non-admin denial shapes match Chatwoot. |
| B11 | Captain Assistant CRUD, inbox binding, responses, documents, and custom tools payload fixtures. | Copilot threads/messages/tasks, playground/tool-call behavior, document sync/embedding feature gates, streaming fallback. | Done only after LLM-dependent behavior is either implemented behind config or safely stubbed with frontend-compatible disabled states. |
| B12 | Boot reused Chatwoot frontend against GoChat auth/profile/inbox/conversation/contact flows. | Add smoke paths for widget init/message, public CSAT, reports, and enterprise screens as B7-B11 land. | Done only after the smoke command is repeatable and writes a checked gap report. |
@@ -979,7 +997,7 @@ Enterprise work package breakdown:
| SLA | Policy CRUD parity, conversation SLA assignment, first-response/next-response/resolution timers, business-hours handling, breach events, notifications. | Policy payloads, applied-SLA attach, seconds-based thresholds, core state transitions, breach idempotency, report payloads, reference `only_during_business_hours` timing, and notification fan-out are covered. Durable scheduling remains Phase 5. | Review |
| Assignment and capacity | Assignment policy CRUD, inbox policy binding, round-robin/availability/capacity selection, manual assignment limits, fallback behavior. | Manual and automatic assignment respect policy, availability, team/inbox membership, and limits. | Review |
| CSAT account/public side | Survey send on resolve, response list, metrics, filters, downloads, review notes, resend/idempotency, public lock. | Account report list/metrics/review-note payload fixtures are covered by `f441680`; public object-shaped submit/update, one-response-per-message idempotency, and resolve-triggered generic survey message send are covered by `ef3a909`; download CSV is covered by `b36cf07`. | Review |
| Automation rules | CRUD payload parity, condition/action parity, event listener coverage, delayed actions, execution logs, no-stop-on-match behavior, webhook and transcript actions. | B9.1a CRUD, B9.1b listener coverage, and B9.1c execution outcome logging are covered; next gate is B9.1d retryable webhook/email transcript delivery, followed by macro parity. | Doing |
| Automation rules | CRUD payload parity, condition/action parity, event listener coverage, delayed actions, execution logs, no-stop-on-match behavior, webhook and transcript actions. | B9.1a CRUD, B9.1b listener coverage, B9.1c execution outcome logging, and B9.1d retryable webhook/email transcript delivery are covered. Durable delayed actions remain B9.3/Phase 5. | Review |
| Macros | Macro CRUD, availability by account/user, execute side effects, validation, audit/log output. | Execute changes conversation labels/status/assignee/team/notes/custom attributes exactly as frontend expects. | Todo |
| Audit | Audit model parity, mutating action coverage, request metadata, filters/pagination, admin endpoint payloads. | Representative mutations across core and enterprise features emit audit records. | Todo |
| Custom roles | Permission-key parity, account-user role resolution, policy middleware, create/update/delete behavior. | Permission matrix tests and frontend admin payload fixtures. | Todo |
@@ -1210,3 +1228,4 @@ Verification milestone gates:
- 2026-06-05: B9.1b automation trigger checkpoint committed as `feat(automation): align rule trigger coverage`; automation listener now covers all current Chatwoot reference events, skips automation-origin/auto-reply/activity auto-reply-email cases, extracts account/conversation context from message and conversation event payloads, conversation update/status/priority paths dispatch `changed_attributes`, initial conversation messages dispatch full message-created context, and provider webhook persistence has a regression proving message-created automation reachability. Focused automation/service/webhook tests, package tests, full `go test ./...`, and `git diff --check` passed. Next slice is B9.1c execution log/stop-on-match parity.
- 2026-06-05: B9.1c execution outcome checkpoint committed as `feat(automation): record rule execution outcomes`; automation evaluations now persist event names, skipped condition outcomes, condition errors, per-action success/failure JSON, and aggregate action counts. Failed actions continue to later actions, and every matching rule still executes in ID order to match the current Chatwoot listener's no-stop-on-match loop. Focused automation tests, automation package tests, relevant service/webhook regressions, full `go test ./...`, and `git diff --check` passed. Next slice is B9.1d retryable webhook/email transcript action delivery.
- 2026-06-05: B9.1d planning checkpoint prepared as `docs: land automation delivery plan`; the active tracker now points at B9.1d, records the Chatwoot webhook/transcript reference contracts, scopes the retryable/testable action delivery boundary, names B9.3 durable-worker follow-up work, and lists the required focused/full verification commands before macro work starts.
- 2026-06-05: B9.1d external action delivery checkpoint prepared as `feat(automation): deliver retryable external actions`; `send_webhook_event` now builds Chatwoot-style automation webhook payloads and delivers through a timeout-bound retryable HTTP boundary, `send_email_transcript` parses recipients and sends generated transcript emails through a fakeable SMTP boundary, and automation execution logs now record delivery type, target, attempts, response code/body, retryability, and failure errors. Focused automation tests, automation package tests, full `go test ./...`, and `git diff --check` passed. Next slice is B9.2 macro CRUD and execution parity.