Matthew Meszaros 6b7e254e56 feat: add native Salesforce sync with Lead and Contact matching and links, a leased activity outbox that logs sends, replies, bounces, opt-outs and meetings as Tasks, Lead Status and Email Opt Out writeback, a pull loop with CRM pause rules, list view and Campaign imports, sandbox and My Domain OAuth that refreshes expired sessions, a Salesforce settings page with contact and inbox cards in web, and docs

2026-10-04 08:54:27 +02:00

..

migrations

feat: add native Salesforce sync with Lead and Contact matching and links, a leased activity outbox that logs sends, replies, bounces, opt-outs and meetings as Tasks, Lead Status and Email Opt Out writeback, a pull loop with CRM pause rules, list view and Campaign imports, sandbox and My Domain OAuth that refreshes expired sessions, a Salesforce settings page with contact and inbox cards in web, and docs

2026-10-04 08:54:27 +02:00

acquire_trace.go

feat: refresh the open conversation for inbox events that name only a message id, give unibox refresh keys added just before a max-wait flush a trailing flush, cancel a shared unibox overview computation once its last caller leaves and bound it and GET /unibox/overview at 30s, include scheduled sends in the coalesced inbox refresh, and word the slow pgxpool acquire log as acquire time rather than saturation

2026-10-01 22:41:00 -07:00

client.go

feat: coalesce and dedupe unibox realtime refreshes in useRealtimeEvents, forward TanStack Query abort signals through every unibox client and the shared Request helper with cancellations kept out of error handling, share one organization-scoped unibox overview computation per second with in-flight dedup and write-time invalidation, bound interactive unibox reads at 30s including pool wait, and log slow pgxpool connection acquisitions

2026-10-01 22:25:53 -07:00

error.go

feat: stop the password reset flow reporting success while sending nothing, by answering 200 only for an address with no account rather than for every cache and database fault, folding addresses to one case on every account lookup and write so a typed capital cannot miss the row or split an SSO account in two, refunding the two-per-four-hours budget when the failure was ours, retrying one transient send and quoting the link's real lifetime; and clear the rest of error tracking by grouping the engagement breakdown in a subquery so ORDER BY stops resolving opens against email_opens, dropping unibox_mailboxes and email_sync_state writes whose mailbox was deleted mid-sync instead of redelivering them forever, answering a corrupt argon2 hash with ErrCredentials rather than a 500, never filing a cancelled caller as a database incident, and reporting only the first websocket init failure of a streak

2026-09-18 11:42:49 +02:00

migrate_test.go

feat: answer a signup for an existing address with 409 conflict and let a concurrent SSO first sign-in resolve against the account it raced, adopt the stored organization DEK when a parallel first use stored one first, classify IMAP SERVERBUG, LIMIT, bare read-command failures and provider ALERT/EXPIRED responses as server-unreachable, throttle or credentials errors and keep the ALERT text, wait up to two minutes for a free connection slot before the boot migration gives up, guard the dashboard's DOM mutations against browser page translation, treat an aborted passkey sign-in as a cancellation, and drop posthog-js request timeouts from error tracking

2026-09-27 08:26:55 +02:00

migrate.go

feat: answer a signup for an existing address with 409 conflict and let a concurrent SSO first sign-in resolve against the account it raced, adopt the stored organization DEK when a parallel first use stored one first, classify IMAP SERVERBUG, LIMIT, bare read-command failures and provider ALERT/EXPIRED responses as server-unreachable, throttle or credentials errors and keep the ALERT text, wait up to two minutes for a free connection slot before the boot migration gives up, guard the dashboard's DOM mutations against browser page translation, treat an aborted passkey sign-in as a cancellation, and drop posthog-js request timeouts from error tracking

2026-09-27 08:26:55 +02:00

pool_share_test.go

feat: make the documented 25-connection pool default real, because pgxpool fills MaxConns with the host's core count before we see it so the guard meant to apply our default never once fired and production quietly ran 48 connections per process against a 76-connection server

2026-09-19 20:34:35 +02:00

types.go

New Repository: Add Backend Code

2026-01-17 14:11:14 +00:00