fix(journeys): assign-bot effect re-read parses wrong CRM envelope key → false negative (EVO-1930)

[daniloleonecarneiro]

Summary

Follow-up to the EVO-1919 effect-verification hardening. After EVO-1900/D11 made the assign-bot binding actually persist (agent_bot_inboxes gets a row), the EVO-1919 re-read started reporting a false negative on the assign-bot node: ... bot binding does not reflect ... after re-read, failing the node even though the binding was created.

Root cause: the re-read already targeted the correct inbox resource (getInboxBot → GET /inboxes/:id/agent_bot), but the confirm predicate parsed the wrong level of the CRM envelope. The CRM (AgentBotSerializer) returns, when a binding exists:

{ "success": true, "data": { "agent_bot": { "id": ... }, "configuration": { ... } } }

and data: null when there is no binding. The predicate read .id off the { agent_bot, configuration } wrapper (which has no id), so even a present binding yielded boundBotId === null → reported as "not persisted".

The EVO-1919 unit test masked this because it mocked a divergent shape ({ data: { data: bot } }) that does not match production.

Fix

• assign-bot.node.ts: unwrap the bot under data.agent_bot (tolerating a flattened shape) before reading id; compare string-coerced ids. Re-read still hits the inbox binding, never the conversation.
• assign-bot.node.spec.ts: mocks now mirror the real CRM envelope ({ success, data: { agent_bot, configuration } | null }). Added EVO-1930 regressions: bound bot confirms success (no false negative), numeric-id coercion, and a different-bound-bot legitimately fails.

(This branch also carries the EVO-1919 verifyEffect helper / journeyVerifyEffect config because those are not yet on develop; they are byte-identical to EVO-1919 and reconcile cleanly on the integration line.)

Test plan

• npx tsc -b — clean.
• npx jest assign-bot — 8/8 pass (incl. new EVO-1930 cases).
• npx jest crm-client src/modules/temporal/activities/nodes — 135 pass; 3 failures are the pre-existing EVOAI_CRM_API_TOKEN ctor failures (reproduce on develop), none new.

Notas

• AC met: binding present → node SUCCESS; binding absent / different bot → legitimate failure; re-read confirms agent_bot_inboxes of the correct inbox.

[sourcery-ai]

Sorry @daniloleonecarneiro, you have reached your weekly rate limit of 500000 diff characters.

Please try again later or upgrade to continue using Sourcery

[dpaes]

Approved — deep adversarial review (card + authoritative diff + real code): acceptance criteria met, no blocking findings. Merge intentionally held to land in the coordinated dependency-order pass. Branch is clean vs develop.