v0.5.0

What's Changed

• docs: add thin AGENTS.md pointing to CLAUDE.md by @vvooki-sys in #14
• perf: run local ONNX embedding on the blocking pool by @vvooki-sys in #13
• fix: redact PII at ingress before content reaches LLM/embedding sinks by @vvooki-sys in #15
• docs: date the security page and require bumping it on changes by @vvooki-sys in #16
• fix: redact PII in caller metadata before persistence by @vvooki-sys in #17
• fix: remove plaintext event: record and purge existing rows by @vvooki-sys in #18
• chore: add opt-in event: purge to container entrypoint by @vvooki-sys in #19
• chore: ignore loomem-server/memory runtime artifacts by @vvooki-sys in #21
• fix: anchor extracted event_date to conversation_date when no explicit cue by @vvooki-sys in #20
• feat: extract user dated personal experiences as event facts by @vvooki-sys in #22
• fix: sanitize full Tantivy operator set in memory_search queries by @vvooki-sys in #23
• fix: configure reqwest::Client connection pool for stable extraction under load by @vvooki-sys in #24
• chore(deps): bump anyhow 1.0.102 -> 1.0.103 (clear RUSTSEC-2026-0190 Audit) by @vvooki-sys in #28
• feat: opt-in max_context_tokens budget on memory_search (MEM1 read-path) by @vvooki-sys in #26
• feat(retrieval): boost user-state facts in composite score via attribution by @vvooki-sys in #25
• fix: source reqwest pool settings from config + separate empty-vs-failed extraction by @vvooki-sys in #29
• feat(extractor): reject generic assistant lists as first-class facts by @vvooki-sys in #30
• docs(landing): hero rewrite - memory as bridge word, editions ladder, try.loomem.ai link by @vvooki-sys in #31
• docs(landing): add client-logo memory strip under hero CTAs by @vvooki-sys in #33
• sec: harden defaults and wire rate limiting (2026-07-01 audit, items 1-7) by @vvooki-sys in #32

Full Changelog: v0.4.1...v0.5.0

v0.4.1

What's Changed

• fix(mcp): gate caller trust elevation behind config (follow-up to #9 Greptile) by @vvooki-sys in #12

Full Changelog: v0.4.0...v0.4.1

v0.4.0

What's Changed

• fix: embed consolidated chunks via configured embedder + dream enqueue by @vvooki-sys in #7
• fix: back-fill missing embeddings via the configured embedder by @vvooki-sys in #8
• refactor: fetch the OpenAI key inline in the embed-missing fallback by @vvooki-sys in #11
• feat: optional relations param on memory_store (deterministic graph edges) by @vvooki-sys in #10
• feat(mcp): derive trust tier from source + surface it in memory_search by @vvooki-sys in #9

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

• feat: auto-trigger dream consolidation after N new chunks per stream by @vvooki-sys in #3
• feat: expose chunk_id in search + per-stream embedding readiness by @vvooki-sys in #4
• docs: fix memory_status returns in mcp-tools (MCP text, not admin JSON) by @vvooki-sys in #5
• feat: MemIR trust-provenance scoring (cycles/001) by @vvooki-sys in #6

Full Changelog: v0.2.2...v0.3.0

v0.2.2

Full Changelog: v0.2.1...v0.2.2

v0.2.1

What's Changed

• Port engram /154 + /153 + /156 (Experience type, configurable topics, history-preserving rewrite) by @vvooki-sys in #1
• docs: reposition landing & README as the context layer for AI agents by @vvooki-sys in #2

New Contributors

• @vvooki-sys made their first contribution in #1

Full Changelog: v0.2.0...v0.2.1

v0.2.0

Full Changelog: https://github.com/vvooki-sys/loomem/commits/v0.2.0