Skip to content

v6.2.0-edge.2

Pre-release
Pre-release

Choose a tag to compare

@MJ-GH-bot MJ-GH-bot released this 03 Oct 21:35
· 8298 commits to next since this release

What's Changed

  • fix(react-test-harness): bound what the browser boundary carries by @jordanfanapour in #4878
  • docs(releases): v6.2.0-edge.1 release notes by @SDesai-BC in #4888
  • fix(memory-leaks): Round 16 audit — bound realtime session/room caches, fix process-kill escalations by @AN-BC in #4783
  • fix: core defects found building Collaboration by @AN-BC in #4891
  • feat(ai): MJ agents can call Jev: typed decisions end to end by @cmbrock-BC in #4814
  • feat(ai-reranker): AIRerankerRunner, so reranking gets failover and run rows by @cmbrock-BC in #4842
  • feat(ai-prompts): AIImageGenerationRunner, so image generation gets failover and run rows by @cmbrock-BC in #4844
  • chore(metadata): remove four duplicate AI Model Cost entries by @SDesai-BC in #4862
  • fix(dashboards): keyboard-operable query splitter and integration entity-map rows by @CaelebB-BC in #3766
  • fix(codegen): apply VirtualEntities keys and finish config-declared v… by @CaelebB-BC in #4828
  • feat(metadata): current models for CodeGen prompts; rank-lineage guard (#4912) by @rkihm-BC in #4920
  • fix(actions-bizapps-accounting): BC verbs use MJ Integrations connections, honor CompanyIntegrationID, never take the first match by @MS-BC in #4904
  • ci(test): run the Redis-gated suites against a real Redis, and fail when they skip by @jordanfanapour in #4908
  • fix(server): save hidden realtime tool turns for magic-link sessions (#4791) by @SDesai-BC in #4797
  • fix(open-app-engine): create SQL Server app schemas owned by the core schema owner by @SDesai-BC in #4760
  • chore(ai-prompts): pin the image carrier prompt's models; deprecate ResolveImageGenerationAPIKey (#4844 follow-ups) by @cmbrock-BC in #4926
  • fix(ai-vectors): wire colocated vector DBs when EntityVectorSyncer has no provider by @cmbrock-BC in #4924
  • fix(ai-prompts): a decision answered by LLM Decision carries its chat model's cost by @cmbrock-BC in #4880
  • feat(ai-prompts): text-to-speech, speech-to-text and video runners (plan Task 0.6) by @cmbrock-BC in #4909
  • feat(testing-engine): decision-judge oracle, criteria as typed Likelihoods by @cmbrock-BC in #4847
  • feat(ai-agents): opt-in payload change check, asked as typed decisions by @cmbrock-BC in #4848
  • feat(ai-agents): narrow the catalog an agent is shown, once per run (off by default) by @cmbrock-BC in #4849
  • feat(ai-prompts): AIEmbeddingRunner, so embeddings get failover and run rows everywhere by @cmbrock-BC in #4843
  • feat(server): RunDecision GraphQL mutation and GraphQLAIClient.RunDecision by @cmbrock-BC in #4854
  • feat(feature-pipelines): pipeline types and the driver seam (LLM today, Decision next) by @cmbrock-BC in #4841
  • fix(metadata): retire the Job Function and Seniority Derivation prompt (#4918) by @bg-BC in #4933
  • feat(ai-core-plus,ng-conversations,messaging-adapters): add a compose:email actionable command by @cmbrock-BC in #4568
  • feat(ng-conversations): route each unmentioned message to the right agent with one decision (opt-in) by @cmbrock-BC in #4860
  • feat(ai-agents): collapse Sage's two-turn agent discovery into one decision (opt-in) by @cmbrock-BC in #4873
  • feat(ai-vector-dupe): Decision and DecisionThenPrompt reasoning modes for duplicate detection by @cmbrock-BC in #4855
  • feat(ng-flow-editor): author a Flow agent's Decision step, and route on its answers by @cmbrock-BC in #4874
  • feat(flow-agents): a Decision step, routed through the decisions root in both execution modes by @cmbrock-BC in #4868
  • feat(ai-agent-manager): author Flow agents with Decision steps by @cmbrock-BC in #4889
  • feat(task-graph): Decision nodes, a decisions root, exhaustive Choice forks, and a real hold by @cmbrock-BC in #4857
  • feat(graphql-dataprovider): connect-only boot for anonymous embeds by @SDesai-BC in #4934
  • test(integration): measure LLM vs Decision Feature Pipelines (plan 5.5) by @cmbrock-BC in #4890
  • feat(record-set-processor): the Decision Feature Pipeline type by @cmbrock-BC in #4856
  • feat(feature-pipelines): the builder chooses a pipeline type, with one shared rule set for what each type can produce by @cmbrock-BC in #4871
  • feat(record-set-processor): a Decision pipeline escalates its borderline records to an LLM pipeline by @cmbrock-BC in #4872
  • feat(feature-pipelines): escalation settings in the Feature Pipeline builder by @cmbrock-BC in #4881
  • feat(ai-reranker): DecisionReranker, scoring agent memory with typed decisions by @cmbrock-BC in #4851
  • feat(record-cloning): entity record graph cloning engine, remote operations and clone wizard by @AN-BC in #4585
  • Simpler record toolbar with per-user pinned actions by @ylee-BC in #4779
  • fix(ng-artifacts): stop ArtifactsModule from preloading React/Babel from unpkg on construction by @SDesai-BC in #4943
  • feat(ai-agents): Loop decision follow-ups from the #4814 review (#4900, #4901, #4902) by @cmbrock-BC in #4906
  • feat(ai, ai-agents, computer-use): realtime + browser-agent model calls run on the run's API key by @cmbrock-BC in #4642
  • fix(content-autotagging): a FieldPathResolver cache miss is "unknown", not "absent" by @cmbrock-BC in #4746
  • fix(search-engine): query Simple Vector Service indexes by Entity Document ID by @cmbrock-BC in #4925
  • feat(dupe): flag likely duplicates while a person enters a new record (Decision-mode entities) by @cmbrock-BC in #4869
  • fix(conversations): an agent completion survives a WebSocket that dies without closing (#4222) by @CaelebB-BC in #4540
  • feat(ai-core-plus): conversation routing acts on calibrated probabilities (plan Task 2.4) by @cmbrock-BC in #4876
  • feat(testing): a Decision Eval harness, which measures typed decisions against labels (plan Tasks 2.1 and 2.2) by @cmbrock-BC in #4875
  • feat(ai-agents): calibrated typed-decision gate for Memory Manager notes (plan 3.4) by @cmbrock-BC in #4897
  • test(decision-eval): measure Sage's agent-discovery decision (plan 3.1) by @cmbrock-BC in #4892
  • feat(ai-agents): calibrated agent-discovery threshold from the eval (plan 3.1) by @cmbrock-BC in #4893
  • test(integration): measure the duplicate check at entry (plan 3.8) by @cmbrock-BC in #4894
  • feat(ai-vector-dupe): calibrated duplicate bands from the measurement (plan 3.8) by @cmbrock-BC in #4896
  • test(integration): replay the finishIf gate on recorded rounds (plan 4.6) by @cmbrock-BC in #4895
  • fix(templates): DocumentText stores each record's rendered text; Templates form saves its editor content by @ylee-BC in #4696
  • fix(core): a failed IS-A chain save or delete leaves the chain as it was by @AN-BC in #4916
  • feat(ai-agents): one master switch for agents' own decision-model use, off by default by @cmbrock-BC in #4958
  • fix(cli): pin citizen-builder workspaces to the CLI's release and surface provisioning failures by @AN-BC in #4586
  • fix: restore the comma #4586 dropped from the root package.json by @rkihm-BC in #4968
  • fix(installer): find mj_generatedentities under apps/MJAPI/node_modules on pnpm installs by @ylee-BC in #4960
  • fix: date-only columns read as their stored day in the query viewer, record changes and timeline; export as YYYY-MM-DD by @rkihm-BC in #4967
  • fix(vectors): address vector indexes by ExternalID, not display Name by @ylee-BC in #4962
  • docs(agents): metadata ships only as release migrations — no install-time metadata phase by @ylee-BC in #4800
  • Rubrics: core evaluation primitive for records, tests, and agents by @AN-BC in #4937
  • docs(rubrics): task-first Rubrics Guide with diagrams by @AN-BC in #4979
  • fix(rubrics): ship Rubric hierarchy CodeGen output and regenerate stale generated types (fixes next Integration Tier) by @AN-BC in #4987
  • fix(rubrics): annotate the rubric form's single-key row load for the PK compliance gate by @bg-BC in #4994
  • AI Model Research Report — 2026-10-02 by @bg-BC in #4996
  • chore(release): v6.2.x Metadata_Sync migration + mj-sync writeback by @bg-BC in #4997
  • fix(integration-test-suite): IT67 refreshes the AIEngineBase vector-index cache by @bg-BC in #5003
  • fix(integration-test-suite): IT56 PG4 uses a seeded empty-grant child by @bg-BC in #5010
  • fix(pg-conversion): six defects in PostgreSQL migration conversion and baking by @bg-BC in #5017
  • chore(release): PostgreSQL counterparts for the v6.2.0-edge.2 migrations by @bg-BC in #5018
  • fix(rubrics): update rubric metadata tests for the release metadata-sync writeback by @bg-BC in #5021
  • v6.2.0-edge.2 by @bg-BC in #5022

Full Changelog: v6.2.0-edge.1...v6.2.0-edge.2