sendnrw/glpi-neuroforge-mega
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v1.6.0
release-tag / release-image (push) Successful in 6m50s
Details

Browse Source
This commit is contained in:
groot
2026-09-02 10:26:50 +02:00
parent 8c4ce2d6c2
commit 9a4370e4df
272 changed files with 31244 additions and 5980 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/ARCHITECTURE.md
+4
View File
@@ -200,3 +200,7 @@ Die beiden Evidenzklassen werden absichtlich nicht vermischt. Outcome-Memories l
Für source-begrenzte Exact-Fallbacks hält der Store einen rebuildbaren In-Memory-Index `Provenance.Source -> Memory IDs`. Damit wächst der Fallback mit der betreffenden Integration/Source statt mit dem gesamten Memory-Katalog. HNSW/Disk-PQ bleiben globale Kandidatenindizes.
Die Qualitätsmessung ist vom Schreibpfad getrennt: `/api/quality/replay` ist read-only und evaluiert live die aktuelle Knowledge-/Outcome-Retrieval-Konfiguration gegen einen bereitgestellten historischen Fallkorpus.
## v1.6.0 Master/Subagent consistency model
NeuroForge owns the authoritative durable state. Subagents are stateless executors: they register capabilities, claim only compatible jobs and return lease-fenced results. Jobs and second-phase master applies are persisted through the normal WAL/checkpoint path, so restarts do not lose queued/retrying/apply-wait work. The graph backfill uses ANN candidate selection plus bounded exact relink jobs; pairwise synapses form an n:m adjacency graph and retrieval can traverse several bounded hops. See `MASTER-SUBAGENT-ORCHESTRATOR.md`.