Skip to content

Handover: family-memory retrieval upgrade + rig KPI guide - #94

Merged
arthware-dev merged 1 commit into
mainfrom
feature/improve-brain-retrieval-v2
Sep 15, 2026
Merged

arthware-dev merged 1 commit into
mainfrom
feature/improve-brain-retrieval-v2

Conversation

@arthware-dev

Copy link
Copy Markdown
Contributor

Handover doc for a dev-rig session to upgrade family-memory retrieval, plus
a KPI/interpretation guide for the agent-lab rig it should measure with.

  • docs/design/handover/memory-retrieval-upgrade.md: FTS5 + BM25 first
    (stdlib, incremental, German diacritics/compounds), hybrid via sqlite-vec +
    LFM2.5-Embedding-350M as a stretch, Hindsight as the OSS memory-layer trial
    option. The agent contract may change when it clearly improves retrieval,
    measured in the rig; two invariants stay (resolvable source links, cite
    vault paths not index ids). Supersedes the mem0 POC's engine choice.
  • tools/agent-lab/rig/README.md: KPIs (wall time, LLM calls, cached tokens,
    TTFT), the 4096-token cache-block rule, how to read a turn, A/B method,
    correctness checks.

Docs only. Follow-on to #93.

FTS5-first plan (hybrid stretch, Hindsight trial option) for a dev rig;
rig README gains KPI and A/B interpretation.
@arthware-dev
arthware-dev merged commit 0e1b3bb into main Sep 15, 2026
1 check passed
@arthware-dev
arthware-dev deleted the feature/improve-brain-retrieval-v2 branch September 22, 2026 20:13
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant