Skip to content

v0.3.0

Choose a tag to compare

@github-actions github-actions released this 11 Aug 04:48
· 257 commits to main since this release
f84b6c9

What's Changed

  • feat(locomo): sharded full-run kit + merged summarize (#184) by @fazpu in #185
  • docs: move OSS site to docs.remember.dev by @fazpu in #188
  • fix(locomo): preserve temporal source anchors by @fazpu in #189
  • Adopt the RememberStack visual identity in OSS docs by @fazpu in #190
  • docs: LoCoMo benchmark corpus (runbook, findings, next steps) by @fazpu in #192
  • Add hybrid claim and live-source retrieval by @fazpu in #193
  • docs(benchmarks): PR #193 post-merge review risks + measurement plan by @fazpu in #194
  • fix(locomo): v8 — twenty-word answers + first-step invalid-completion retry by @fazpu in #195
  • design: agent retrieval surface — coverage analysis + binding tool catalog by @fazpu in #196
  • feat(retrieval): agent surface Batch B — entity/time recipes + protocol v9 by @fazpu in #197
  • feat(retrieval): agent surface Batch C — current_context with evidence-backed facts by @fazpu in #198
  • feat(retrieval): agent surface Batch D — multi_hop_context connection recipe by @fazpu in #200
  • feat(retrieval): agent surface Batch E — hybrid recall mechanics (tail-refill + dedup grouping) by @fazpu in #202
  • design(e1): D80 embedding-input policy (conventional, conditional location) by @fazpu in #199
  • feat(e1): implement D80 deterministic embedding-input policy by @fazpu in #201
  • design: open analytical query surface — sandboxed SQL over invariant views + saved-query registry by @fazpu in #204
  • design: full Cypher read surface over the P2 graph joins v1 by @fazpu in #205
  • design: complete the Lance surface, question_context v4 recall, facts-layer headline by @fazpu in #206
  • design: physical tenancy without RLS; plain views without security_barrier by @fazpu in #207
  • fix(e1): D80 follow-up nits — packaging, embedder sync, poison path by @fazpu in #203
  • feat(queryspace): Batch A — memory_v1 schema contract, manifest, and invariant test matrices by @fazpu in #208
  • feat(queryspace): Batch B — the SQL sandbox, tenancy roles, and discovery by @fazpu in #209
  • feat(queryspace): Batch C — public retrieval functions and PG confirmation by @fazpu in #210
  • fix(queryspace): bind every Batch A authorization coordinate by @fazpu in #213
  • fix(queryspace): default-deny the SQL grammar over node classes by @fazpu in #212
  • fix(queryspace): align Batch B contract authorities by @fazpu in #214
  • feat(queryspace): Batch D — graph helpers and the full Cypher read surface by @fazpu in #211
  • feat(queryspace): Batch E saved-query customer space by @fazpu in #215
  • feat(queryspace): Batch F dual-surface integration by @fazpu in #219
  • design: obs embed cache, checkpointing, relation-label analysis by @fazpu in #220
  • feat: obs rank cache, P1 checkpoints, deterministic fact labels by @fazpu in #221
  • design: observation adjudication efficiency proposals by @fazpu in #216
  • feat(openrouter): embedding provider order with fallbacks by @fazpu in #217
  • design: shared provider health routing proposal by @fazpu in #218
  • ci: PR fast gate (~10 min); nightly full suite by @fazpu in #222
  • refactor(queryspace): make clean pre-release cut by @fazpu in #223
  • test(queryspace): seed authoritative spike lineage by @fazpu in #224
  • docs(queryspace): surface current API contract by @fazpu in #225
  • feat(benchmarks): BEAM official nugget LLM-judge scorer by @fazpu in #226
  • fix: BEAM scorer + P2 export dual-review follow-ups by @fazpu in #228
  • docs: progressive-disclosure public site for agents and humans by @fazpu in #229
  • docs: Plane K page and mounts polish by @fazpu in #230
  • benchmark(locomo): score current system with Luna by @fazpu in #227
  • docs: configuration, troubleshooting, evaluation by @fazpu in #232
  • docs: strip internal phase/decision jargon from public site by @fazpu in #233
  • fix(selfhost): persist query role credentials by @fazpu in #234
  • fix(locomo): attest durable visible versions by @fazpu in #235
  • docs: sell-first README; remove pre-release banners by @fazpu in #236
  • feat(benchmarks): BEAM full-plane retrieval answer agent by @fazpu in #231
  • feat(workers): chunk-level extract (D84) by @fazpu in #238
  • Fix D84 advisory-lock signature by @fazpu in #239
  • feat(benchmark): expose full retrieval plane to LoCoMo by @fazpu in #240
  • fix(query): bound current-context authority plan by @fazpu in #241
  • fix(query): bound contradiction enrichment by @fazpu in #242
  • fix(e3): D86 unknown entity type gate — retry then drop by @fazpu in #243
  • docs: analyze LoCoMo v11 and question-turn claims by @fazpu in #244
  • design(retrieval): align context operations with authorities by @fazpu in #246
  • feat(mcp): Layer 1 memory verbs — ingest + pipeline_readiness by @fazpu in #245
  • feat(retrieval): implement D87 context operations by @fazpu in #247
  • design(e3): claim-level normalize fan-out (D88) by @fazpu in #248
  • fix(tests): seed the document lineage the memory_v1 views require by @fazpu in #253
  • fix(p2): project retracted edges, honour evidence lineage, and redirect merged endpoints by @fazpu in #249
  • release: RememberStack 0.3.0 by @fazpu in #256
  • ci(release): give verify a budget larger than the suite it runs by @fazpu in #257

Full Changelog: v0.2.0...v0.3.0