Releases: mlimarenko/IronRAG.Confluence
Releases · mlimarenko/IronRAG.Confluence
Release list
v0.1.9
0.1.9 — 2026-06-18
- Build against framework
v0.0.11:routing.yamland per-kind policy
overrides are reloaded before each sweep and webhook-handled item, so routing
rule changes take effect without a connector restart. - Invalid or temporarily unreadable routing configuration keeps the last valid
routing state active, avoiding accidental document movement or deletion during
configuration edits.
v0.1.8
0.1.8 — 2026-06-18
- Build against framework
v0.0.10: IronRAG upload, replace, and delete
admission now fail fast with configurable mutation timeouts and are recorded
as deferred instead of holding a sweep indefinitely. - Inherits deferred cursor handling for primary and dependent items, plus a
bounded post-sweep reaper list call, so a slow backend can be retried by a
later sweep without advancing cursors for work that was not accepted.
v0.1.7
0.1.7 — 2026-06-18
- Build against framework
v0.0.9: one source ref's fetch/push work,
including dependents, is bounded bySYNC_ITEM_TIMEOUT_SECONDS. A stuck item
is cancelled and counted as an item error so the sweep can continue to
sync.doneinstead of holding the single-flight sync lock indefinitely. - Inherits
sync.item.startandsync.item_timeoutlogs for identifying the
source ref that is currently in flight or timed out.
v0.1.6
0.1.6 — 2026-06-18
- Build against framework
v0.0.8: full sweeps are now single-flight inside
the connector process, so manual, startup, and periodic triggers cannot run
overlapping mutation passes against the same cursor database. Concurrent
manual/sync/runrequests return HTTP 409 while an existing sweep is active. - Inherits framework cancellation cleanup for manual sync requests that are
disconnected before the sweep finishes.
v0.1.5
0.1.5 — 2026-06-18
- Build against framework
v0.0.7: large legacy cursor databases now cap
best-effort document-detail backfill before source enumeration, and the
same-route path backfills cursor ownership via the target library's
external-key lookup before trying slower document-detail calls.
v0.1.4
0.1.4 — 2026-06-18
- Build against framework
v0.0.6: legacy cursor library backfill is now
bounded and non-fatal, so a slow document-detail lookup no longer makes
manual/sync/runreturn 500 before Confluence enumeration starts.
Unresolved legacy cursor ownership still blocks possible duplicate uploads;
destructive cleanup for unknown historical libraries is deferred to a later
sweep.
v0.1.3
0.1.3 — 2026-06-18
- Build against framework
v0.0.5: cursor rows now retain the target
library, orphan scans follow cursor pagination, route moves avoid deleting
newly-routed documents, legacy cursor rows are backfilled from IronRAG
document details, and transient conflicting mutations are retried on a later
sweep without advancing the cursor. - Tightened connector type annotations so strict mypy passes against the typed
framework package.
v0.1.2
0.1.2 — 2026-06-10
- Page attachments and inline images now ingest as attached context of
their source page instead of as peer documents. Each dependent is
auto-linked to its page via the framework'sparent_external_key; no
config change is required. Inherited from the connector SDK — IronRAG
derives the child's role from the declared parent and media class.
v0.1.1
0.1.1 — 2026-05-29
- Pages whose own body is empty but which host attachments are no longer
dropped. The page is emitted with a synthesized index body listing its
attachments, and each attachment is ingested as its own document.
Previously an empty page body caused the whole item — including its
attachments — to be skipped. - Build against framework
v0.0.3: content-addressed idempotency keys
(fixes409 idempotency_conflicton re-rendered page updates),
single-request external-key lookup via the listsearchfilter, and
cursor document-id persistence on the unchanged path (steady-state
sweeps no longer re-scan the IronRAG list endpoint).