Files
jbergner a6bc71fb3a
release-tag / Resolve release metadata (push) Successful in 30s
release-tag / Build knowledge (push) Failing after 4m51s
release-tag / Build control (push) Failing after 5m0s
release-tag / Build agent (push) Failing after 5m0s
release-tag / Build agent-data-init (push) Failing after 5m5s
release-tag / Build neuroforge-worker (push) Failing after 5m7s
release-tag / Build neuroforge (push) Failing after 5m9s
Init
2026-08-26 18:34:41 +02:00

32 lines
1.8 KiB
Plaintext

NeuroForge v0.7.0 validation
Date: 2026-08-17
PASS go test ./...
PASS go vet ./...
PASS go test -race ./internal/store
PASS go test -race ./internal/brain
PASS go test -race ./internal/httpapi
PASS go test -race ./internal/vector
PASS Admin JavaScript syntax (node --check)
PASS OpenAPI 3.1 YAML parse; version 0.7.0; 60 paths
PASS Linux amd64 static builds: server, worker, benchmark
PASS Runtime release-binary smoke: /readyz, /admin security headers, learning-policy API, /metrics auth/export, graceful SIGTERM shutdown
PASS Admin token absent from runtime log
PASS Knowledge Explorer integration: /learn -> provenance summary -> explainable knowledge search
PASS Learning Policy: blocks explicit learn before provider call
PASS Learning Policy: duplicate suppression
PASS Learning Policy: strongly negative assistant response archival
PASS Secrets masked by default
PASS Knowledge Event persistence across restart
PASS Explainable SearchHit score decomposition
PASS Sequential PQ migration scanner: latest live vectors only; archived/deleted records excluded
Operational caveats / deliberate boundaries:
- TLS and per-client/IP rate limiting belong at a reverse proxy/ingress.
- /readyz validates configured routes and cluster readiness; it intentionally does not perform network calls to Ollama/OpenAI on each probe.
- Knowledge Graph is browser-bounded (max 250 nodes), not a renderer for the entire database.
- Opening Knowledge Explorer summary may perform a metadata scan over the store; normal /metrics scrapes do not.
- Existing pre-v0.7 memories do not gain invented provenance; they appear as legacy/unknown.
- Cluster remains Raft-like rather than a claim of complete Raft consensus/log matching.
- Disk-PQ remains an approximate candidate generator; returned candidates are exact-reranked against original vectors when available.