v0.3.17

Fixed

• Per-issue sparklines failed to load on Postgres (INT4/INT8 type mismatch)

v0.3.16

Added

• Startup check refuses open-mode ingest on a public bind with no rate limit, unless open_ingest_unlimited_acknowledged is set; stackpit init now generates a default rate limit
• Admin browser logins get a random per-session cookie with 24h expiry; logout revokes it
• Create GitHub, Forgejo, or GitLab issues from a Stackpit issue

Changed

• Rate limits now enforce their scope: a per-project limit is shared across all of the project's DSN keys, and the global limit across all ingest
• Traces list queries are bounded to the newest 50,000 spans

Fixed

• Ingest auth no longer reads a database error as "project missing", which in open mode could let a new key register into an existing project
• Web sessions fail closed instead of falling back to the default organization when a database error occurs during org resolution

v0.3.15

Added

• Move a project to another organization from its settings

v0.3.14

Added

• Issue detail renders breadcrumbs, user, and extra context
• Span operation aggregation on the spans page
• Transaction duration distribution
• Cross-links between replays and traces
• Error-free sessions in Release Health

Changed

• Styled 404 page

v0.3.13

Added

• Per-release detail page: session health plus the release's issues
• Expandable spans with metadata in the trace waterfall
• Minified-stacktrace hint linking to source-map settings

Changed

• Release Health errored/crashed counts link through to the release
• Trace waterfall compresses long idle gaps
• Client reports show per-report dropped counts and reasons
• Release versions rendered as short hashes
• Issue breadcrumbs expanded by default

Fixed

• Long issue titles no longer overflow the issues table

v0.3.12

Security

• crossbeam-epoch bumped for RUSTSEC-2026-0204

Fixed

• Environment filter on an integration no longer drops all its notifications
• Global email "Test" button no longer a no-op with SMTP
• Org deletion no longer undercounts alert rules and digest schedules

v0.3.11

Added

• Commercial license infrastructure (offline Ed25519 verification, admin activation page)
• Prometheus /metrics endpoint, token-authed and license-gated (Observability)
• [email] enabled switch to configure a mailer but skip sending

Changed

• Email config now uses polymail's provider schema (config feature, polymail 0.1.6). Breaking: the [email.smtp] block collapses into [email], and its fields are renamed: username/password -> user/pass, starttls -> start_tls. Omit the whole [email] section to leave mail unconfigured.

Fixed

• Graceful shutdown no longer discards events that were already acknowledged
• Failed write batches retry with backoff instead of being dropped
• Sourcemap bundle assembly caps total size at 512 MiB
• Open mode: a failed key auto-registration returns 500 instead of accepting events

v0.3.10

Added

• SMTP email provider via a new [email.smtp] block
• Test button on per-project integrations sends a real notification
• Test button on digest schedules sends a real or sample digest
• Notification emails link back to the issue and project pages

Changed

• polymail updated to 0.1.3 (SMTP backend)

Fixed

• Per-project "To address" field no longer shows the raw config JSON
• Recipient is validated as an email address on submit
• Missing [filter] config no longer rejects all ingest

v0.3.9

Security

• Source maps are now isolated per project. They were keyed globally by debug id, so one tenant could read or overwrite another tenant's source maps and original source
• Ingest authentication caches denials and the open-mode project count, and caps per-IP authentication-failure floods, so a stream of unknown-key requests can no longer hammer the database

Added

• Concurrent ingest writers, configurable via ingest_writers (PostgreSQL)
• Configurable ingest write batch size via ingest_batch_size
• stackpit-bench: open-loop ingestion benchmark (ramp to knee, soak, CSV + SVG chart)

Fixed

• Event batches that hit a transient write failure and retry no longer double-count issue, tag, session, and transaction metrics, lose counts, or store unreadable double-compressed payloads
• Retried SDK envelopes no longer inflate counts; the ingest endpoint now accepts a whole envelope or none
• Retention deletion reconciles issue counts against exactly the rows it deletes
• Writer shutdown is no longer lost under a full queue, and a flush failure during drain no longer drops events without accounting

Changed

• Payloads are compressed on the accept path instead of on the writer, raising sustained ingest throughput
• Per-project navigation badge counts are cached briefly instead of recomputed on every page render
• The writer bounds queued memory by bytes, not just message count, to avoid running out of memory under bursts of large payloads
• Admin bulk deletes run in chunks so they no longer stall ingestion
• Organization, auth, and background writes go through the write pool, and the read pool is enforced read-only

Full Changelog: v0.3.8...v0.3.9

v0.3.8

Fixed

• PostgreSQL: release-health daily sessions chart crashed on a column decode error
• Docker images failed to build after the library-crate split

Full Changelog: v0.3.7...v0.3.8