Skip to content

Infinity Context v0.2.0

Latest

Choose a tag to compare

@github-actions github-actions released this 26 Oct 23:45
· 1 commit to main since this release
Immutable release. Only release title and notes can be modified.
2fa20ae

What's Changed

  • fix: bind release CLI to repository by @777genius in #16
  • feat: establish cognitive architecture foundation by @777genius in #18
  • docs: clarify public quickstart by @777genius in #17
  • feat(benchmark): add resumable evaluation journal by @777genius in #21
  • feat(benchmark): seal publishable runtime v2 contract by @777genius in #22
  • feat(benchmark): seal LoCoMo ingestion manifest by @777genius in #23
  • feat(benchmark): add sealed canary runner by @777genius in #24
  • feat(benchmark): define resumable publishable keyless profile by @777genius in #20
  • fix(smoke): select full Compose profile by @777genius in #25
  • fix(smoke): preserve legacy file ceiling by @777genius in #26
  • fix(ci): lower smoke runner size baseline by @777genius in #28
  • feat(benchmark): add hosted Mem0 OSS canary authority by @777genius in #19
  • fix(memory): make episode replay conflict-safe by @777genius in #30
  • feat(benchmark): bind canonical episode projection inventory by @777genius in #31
  • feat(benchmark): add resumable operation journal by @777genius in #29
  • feat(benchmark): verify subscription runtime receipts by @777genius in #32
  • feat(benchmark): add Mem0 full-run v5 authority by @777genius in #33
  • feat(benchmark): add paired resumable ingestion E2E by @777genius in #34
  • feat: add attested Mem0 v5 benchmark adapter by @777genius in #35
  • feat(benchmark): add managed Mem0 v5 lane contract by @777genius in #36
  • feat(benchmark): bridge managed Mem0 v5 retrieval by @777genius in #37
  • feat(benchmark): authenticate Mem0 v5 transport evidence by @777genius in #38
  • feat(benchmark): complete Mem0 v5 lifecycle evidence by @777genius in #39
  • refactor(benchmark): introduce provider-neutral runner seams by @777genius in #40
  • feat(benchmark): add managed Mem0 v5 runner foundation by @777genius in #41
  • refactor(benchmark): add neutral execution evidence variants by @777genius in #42
  • feat(benchmark): complete managed v5 evidence foundations by @777genius in #43
  • feat(benchmark): authorize observed multi-operation receipts by @777genius in #44
  • feat(benchmark): seal Mem0 v5 execution evidence by @777genius in #45
  • feat(benchmark): route retrieval by bound target by @777genius in #46
  • feat: add temporal memory fact architecture by @777genius in #27
  • feat: add Mem0 v5 production lifecycle foundation by @777genius in #47
  • feat(benchmark): complete managed v5 production cutover by @777genius in #48
  • feat(benchmark): add managed v5 live cutover by @777genius in #49
  • test(benchmark): keep phase c preload portable by @777genius in #50
  • fix: reject inconsistent memory benchmark judge scores by @777genius in #51
  • feat(serving): attest embedding runtime identity by @777genius in #53
  • fix(retrieval): support advisory project anchors by @777genius in #54
  • feat(benchmark): add strict-v4 publishable comparison pipeline by @777genius in #52
  • fix(documents): keep scoped listing active-only by @777genius in #55
  • feat(retrieval): add production retrieval v2 by @777genius in #57
  • fix(sdk): align release canary query provenance by @777genius in #58
  • fix(sdk): complete trusted release and timeout semantics by @777genius in #59
  • fix(sdk): accept GitHub draft release URLs by @777genius in #60
  • fix: preserve bounded projected embedding text by @777genius in #61
  • fix(retrieval): preserve leases for unchanged lane evidence by @777genius in #62
  • fix(context): enforce canonical rendered character budget by @777genius in #64
  • test(architecture): guard canonical feature ownership boundaries by @777genius in #63
  • feat(facts): add canonical relation transaction ports by @777genius in #65
  • feat(retrieval): add explicit v3 thread selection by @777genius in #66
  • fix(release): keep SDK qualification releases in draft by @777genius in #67
  • fix(sdk): validate release version consistency for 0.3.1 by @777genius in #68
  • fix(retrieval): match locator keyword queries to the trigram index by @777genius in #69
  • fix(retrieval): suppress weak lexical fusion hits by @777genius in #70
  • fix(retrieval): start reconciliation lease after attestation by @777genius in #71
  • fix(retrieval): widen outbox aggregate IDs by @777genius in #72
  • feat(retrieval): expose bounded keyword and dense fusion weights by @777genius in #73
  • feat(memory): preserve relevant sources when context budgets fill by @777genius in #74
  • perf(memory): reuse computed distinctive term frequencies by @777genius in #75
  • fix(retrieval): map citation char ranges to source coordinates by @777genius in #76
  • fix(retrieval): quote clean chunk body in duplicate-member source refs by @777genius in #77
  • fix(context): new-source displacement under char cap (frozen-148: 90->93 complete, 201->206 sources) by @777genius in #78
  • docs(eval): end-to-end answer-quality paired measurement (Main-100 36/100 vs 0/100, held-48 18/48 vs 0/48) by @777genius in #79
  • fix(snippets): extend query window past a trailing hit cluster by @777genius in #80
  • tools(eval): recall probe for frozen stores + context repack helper by @777genius in #82
  • docs(quality): answer-failure vein map r1 (post snippet-edge fix) by @777genius in #81
  • docs(readme): refresh quality snapshot and drop latency claim by @777genius in #83
  • perf: request-local memo for lexical term frequencies (/v1/context latency) by @777genius in #84
  • perf: batch keyword neighbor and sibling-backfill chunk fetches by @777genius in #85
  • perf: replay cached length buckets in approximate term frequency by @777genius in #86
  • perf: keep approximate term frequency buckets memo-only by @777genius in #87
  • docs: measure competitor levers - lexical re-query and temporal filter by @777genius in #88
  • fix: preserve memory evidence and fence outbox ownership by @777genius in #89
  • chore(ci): add reviewrouter-review.yml dispatch target by @777genius in #90
  • refactor: isolate asset extraction lifecycle services by @777genius in #94
  • refactor: separate core context evidence policies by @777genius in #92
  • refactor: extract benchmark evidence policies and source identity by @777genius in #93
  • refactor(benchmarks): move retrieval intent out of server by @777genius in #97
  • refactor(postgres): separate schema bootstrap from unit of work by @777genius in #96
  • refactor(postgres): give canonical repository capabilities explicit owners by @777genius in #98
  • refactor(core): consolidate context ranking policy ownership by @777genius in #95
  • refactor(benchmarks): isolate complete candidate scoring policy by @777genius in #99
  • refactor(core): resolve compatibility exports lazily by @777genius in #100
  • fix: preserve extraction cancellation across concurrent heartbeats by @777genius in #102
  • test: restore strict authority oracles and exact PostgreSQL ACL cleanup by @777genius in #103
  • test: complete real multimodal acceptance across fair worker claims by @777genius in #101
  • fix(ingestion): verify complete authenticated managed document replays by @777genius in #104
  • chore: release 0.2.0 and SDK 0.3.2 with a simpler README by @777genius in #105

Full Changelog: v0.1.0...v0.2.0