wiki-maintain 2026-11-17: regenerate skills-index (299 skills/67 cats), freshness report (7 stale); verified live & updated current-profile (ops profile deleted, config v52, 6 MCP servers, 15 plugins, 21/26 pins), memory-backend (2901 memories, coverage working/episodic 100%), model-providers (pin split, pulse_memory aux slot on dead nous credential)

This commit is contained in:
Tony0410
2026-11-17 03:11:54 +08:00
parent e5612c8a8d
commit 4574c4ebc3
5 changed files with 180 additions and 28 deletions
+4 -4
View File
@@ -3,8 +3,8 @@ title: Memory Backend
type: system
status: active
created: 2026-07-22
updated: 2026-10-26
verified_on: 2026-10-26
updated: 2026-11-17
verified_on: 2026-11-17
confidence: high
tags: [memory, mnemosyne, agent]
sources: [raw/configs/hermes-config-sanitized.txt]
@@ -34,7 +34,7 @@ Persist and recall agent memory across sessions.
- **Sleep threshold:** 50 turns
- **Nudge interval:** 10
- **Ignore patterns:** none set
- **Live totals (2026-10-26):** 2,350 memories — 842 working / 1,090 episodic / 418 `memoria_facts`
- **Live totals (2026-11-17, sqlite):** 2,901 memories — 1,130 working / 1,249 episodic / 522 `memoria_facts`
## Storage Location
- Mnemosyne home: `~/.hermes/mnemosyne/`
@@ -48,7 +48,7 @@ Persist and recall agent memory across sessions.
- Health monitoring currently treats capture as stale only after 24 hours; an activity-aware comparison is preferable to catch dry-intake failures sooner without false alarms during idle periods.
- The 2026-08-01 automatic-capture failure and repair are documented in [[incidents/2026-08-01-mnemosyne-capture-failure]].
- **Vector recall degraded silently until 2026-10-03** — `fastembed` was missing from every uv-managed runtime, so `mnemosyne_recall` fell back to FTS5 keyword scoring with no error raised. Repaired 2026-10-03; the 15-min dependency guard now asserts `fastembed` and the model cache path. See [[incidents/2026-10-03-mnemosyne-vector-recall-fts-only]].
- **Embedding coverage re-verified 2026-10-26** — `check_embedding_coverage.py` reports 1,899 / 2,350 (80.8 %) embedded; **working 815 / 842 (96.8 %)** and **episodic 1,084 / 1,090 (99.4 %)**. Both tiers are **above the 95 % reindex threshold**, so no reindex was triggered. `fastembed` 0.8.1 is present in **both** managed runtimes and 0.8.0 in the legacy venv; `FASTEMBED_CACHE_PATH` is pinned. The 2026-10-03 follow-up (assert `fastembed` in *all* managed runtimes) is now satisfied.
- **Embedding coverage re-verified 2026-11-17** — `check_embedding_coverage.py` reports 2,379 / 2,901 (82.0 %) embedded; **working 1,130 / 1,130 (100 %)** and **episodic 1,249 / 1,249 (100 %)** (the gap is `memoria_facts`, unvectorized by design). Both tiers are **above the 95 % reindex threshold**, so no reindex is needed. Earlier (2026-10-26) check: 1,899 / 2,350 (80.8 %), working 96.8 %, episodic 99.4 %.
**Correction (2026-07-22, per Anthony):** this page previously referenced Hindsight as a possibly-used backend on this box. That was a mix-up with a different Hermes instance — see [[comparisons/memory-backends]].