v1.5.0

1.5.0 (2026-08-31)

Features

• a --source for every subcommand, and the operations story (#188) (b69a0b4)
• a ceiling on total retry time (#162) (547df29)
• a keyed pseudonym transform (#164) (d118b38), closes #143
• a Kubernetes operator (#202) (bb7c06d), closes #147
• a lookup transform (#165) (60ab49e)
• a pgvector sink (#193) (8781038), closes #185 #186
• a Qdrant sink (#194) (1d22926), closes #187
• a rate limit for the initial load (#155) (5ff180d), closes #144
• add and remove a table on reload (#199) (293a7ac), closes #170
• aggregate children keep a count live (#189) (ba5ab79), closes #179
• fan out a delimited column, and let the element be the join parent (#190) (ee0c0ae), closes #180
• flatten a one-to-one child into the parent document (#191) (ad72c29)
• Helm readiness probe and PodDisruptionBudget (#160) (8172a82)
• load a directory of configs and validate them together (#168) (ccd8427)
• many-to-many children through a junction table (#156) (2c5b2c5)
• opt-in OpenTelemetry traces (#167) (afdd8ff), closes #152
• reload the configuration without a restart (#169) (ab2ad96), closes #145
• require_alias refuses writes past the alias (#158) (e6fe2a4)
• retry a source that is unreachable at startup (#197) (510b91f), closes #172
• run every configured source in one process (#174) (99570b3)

Bug Fixes

• a MySQL table removed on reload stops being decoded (#200) (8c037e7)
• writes through the alias satisfy require_alias on Elasticsearch (#196) (8dabf40), closes #178

What's Changed

• ci: the e2e suites take one machine-wide lock by @kennywillbe in #157
• docs: three scope decisions from the survey by @kennywillbe in #153
• ci: ci-local waits for another run instead of refusing by @kennywillbe in #159
• feat: a rate limit for the initial load by @kennywillbe in #155
• docs: secrets-manager recipes for the environment by @kennywillbe in #154
• ci: ci-local holds the e2e lock around seeding too by @kennywillbe in #161
• feat: many-to-many children through a junction table by @kennywillbe in #156
• feat: require_alias refuses writes past the alias by @kennywillbe in #158
• feat: a lookup transform by @kennywillbe in #165
• feat: Helm readiness probe and PodDisruptionBudget by @kennywillbe in #160
• feat: a ceiling on total retry time by @kennywillbe in #162
• ci: run the required checks on the merge queue by @kennywillbe in #166
• docs: how pull requests land without a merge queue by @kennywillbe in #171
• ci: the e2e lock owner claim ends with the lock by @kennywillbe in #175
• feat: a keyed pseudonym transform by @kennywillbe in #164
• ci: action majors that run on Node 24 by @kennywillbe in #176
• feat: load a directory of configs and validate them together by @kennywillbe in #168
• ci: cache the dependency build and replace fixed sleeps by @kennywillbe in #177
• feat: reload the configuration without a restart by @kennywillbe in #169
• feat: opt-in OpenTelemetry traces by @kennywillbe in #167
• chore: current contact links and verification steps in the templates by @kennywillbe in #182
• feat: run every configured source in one process by @kennywillbe in #174
• ci: isolated stacks so ci-local runs can go in parallel by @kennywillbe in #173
• feat: a --source for every subcommand, and the operations story by @kennywillbe in #188
• feat: aggregate children keep a count live by @kennywillbe in #189
• feat: fan out a delimited column, and let the element be the join parent by @kennywillbe in #190
• feat: flatten a one-to-one child into the parent document by @kennywillbe in #191
• ci: prove the PostgreSQL-derived services in the matrix by @kennywillbe in #192
• feat: a pgvector sink by @kennywillbe in #193
• feat: a Qdrant sink by @kennywillbe in #194
• ci: every sink runs the conformance kit by @kennywillbe in #195
• fix: writes through the alias satisfy require_alias on Elasticsearch by @kennywillbe in #196
• feat: retry a source that is unreachable at startup by @kennywillbe in #197
• feat: add and remove a table on reload by @kennywillbe in #199
• fix: a MySQL table removed on reload stops being decoded by @kennywillbe in #200
• ci: the init section talks to its own stack when isolated by @kennywillbe in #201
• feat: a Kubernetes operator by @kennywillbe in #202
• chore: release 1.5.0 by @kennywillbe in #163

Full Changelog: v1.4.0...v1.5.0