feat(cognitive): expose bounded local canary
CI / test (push) Successful in 13s

This commit is contained in:
m
2026-08-17 10:29:03 +03:00
parent 2a0aee0073
commit bcb106e0a3
8 changed files with 78 additions and 10 deletions
+2 -1
View File
@@ -1,6 +1,6 @@
# Assistant project status
Audit date: 2026-08-17. Active implementation: DEV-000058 revision 1.
Audit date: 2026-08-17. Active implementation: DEV-000059 revision 1.
## Current milestone
@@ -12,6 +12,7 @@ Audit date: 2026-08-17. Active implementation: DEV-000058 revision 1.
- Discovery never activates a provider, model or binding. Until a later benchmark and explicit promotion task, all discovered bindings remain `INACTIVE` and production remains fully LLM-free.
- DEV-000057 executed a bounded synthetic qualification corpus against all 11 current local model revisions. Four exact binding revisions demonstrated the minimum advisory threshold; seven remain unsupported. Qualification evidence does not change lifecycle or activation.
- DEV-000058 promoted the exact CPU `Qwen2.5-3B-Instruct-Q4_K_M` lineage to `VALIDATED/CANARY` through an authenticated, evidence-gated, append-only CAS transition. Ordinary routing still excludes it; only an explicit `AllowCanary` request can select it. All other current bindings remain `INACTIVE`.
- DEV-000059 enables that exact canary behind a fail-closed cognitive-worker flag and a 4 KiB output clamp. Five authenticated synthetic Creator requests completed through the full production path with typed usage observations, zero failures and zero API cost. The binding remains `CANARY`, not `ACTIVE`.
## Delivered
@@ -0,0 +1,16 @@
{
"artifact_id":"local-cpu-canary-observation-2026-08-17-v1",
"binding_id":"binding-9e5e9779cd10b907c3695374",
"binding_revision":2865230063392044815,
"provider_model_id":"Qwen2.5-3B-Instruct-Q4_K_M",
"path":"authenticated Creator Web to Cognitive Gateway to local CPU adapter to Capability usage owner",
"synthetic_requests":5,
"completed":5,
"failed":0,
"latency_millis":{"minimum":2347,"maximum":4603,"average":3646},
"usage":{"input_units":225,"output_units":271,"actual_api_cost_micros":0},
"canary_max_output_bytes":4096,
"automatic_retry":false,
"openai_requests":0,
"activation":"CANARY"
}