gen_kvgume8tfu

Decision timeline

1. admission+0 ms

   embeddings request by prin_app under profile @company/embeddings (requested "@company/embeddings")

2. cache+0 ms

   exact cache hit — served from cache, no provider dispatch

3. route+0 ms

   selected openai/text-embedding-4 via openai/acct_openai_pool_a_use1 on node_cf_use1_01 (A-tier, us-east-1) — objective=cost, 1 candidate scored, 0 in fallback chain, 0 hard-filter rejections [router router-v3.4.1, policy policy-2026-07-14]

   1 candidates considered1 survived hard filters0 rejectedobjective: costfallback chain: 0router-v3.4.1 · policy-2026-07-14

   Score breakdown — top 1

   #	Candidate	quality	cost	latencyPenalty	cacheSavings	contractValue	shadowPrice	retryRisk	affinityPenalty	Total
   1	openai/text-embedding-4 via openai · node_cf_use1_01 selected	+0.258	-0.002	-0.002	0.000	0.000	-0.010	-0.001	+0.300	0.543

4. settlement+38 ms

   settled $0.00 — 21255 in / 0 out tokens (21255 total)

   No ledger entries — this generation never reached the reservation step (cache hit or pre-route block).

   Cost layers (memo §15.3)

   Provider list

   $0.0017

   Contract-adjusted

   $0.0017

   Internal charge

   $0.0020

   Contract savings

   $0.00

   contract ct_openai_metered (metered)

5. complete+38 ms

   status=succeeded, model=openai/text-embedding-4, 38ms end-to-end

fabric_metadata

The route-explanation envelope an API caller receives with X-OpenRouter-Metadata: enabled or routing.explain, reconstructed from this generation record.

{
  "requested": "@company/embeddings",
  "resolved_model": "openai/text-embedding-4",
  "strategy": "exact-cache",
  "policy_version": "gp_baseline@?",
  "router_version": "cache",
  "provider": null,
  "region": null,
  "account_pool": null,
  "node": null,
  "cache": {
    "response": "HIT",
    "prompt_affinity": "MISS",
    "predicted_cached_tokens": 0
  },
  "guardrails": [],
  "attempts": [],
  "routing_reason": [
    "exact response cache hit — provider dispatch skipped"
  ]
}