{
  "schema_version": 1,
  "verdict": "FAIL",
  "run_id": "gateway-4fe838aba2",
  "victor_commit": "e25fdf7513a00845c12509808239c15b3962689d",
  "harness_sha256": "fbe7b835d8546ae4c7330e9bcf948db2c6c3733dde9fd5eb1f79f9e38e236f4c",
  "harness_source_unchanged": true,
  "fresh_victor_state": true,
  "external_observer": "Private temporary metadata-only capture; no request/response bodies exported",
  "source_tree_clean_at_launch": true,
  "sandhi": {
    "source_commit": "eb38ff4b4c90121ac3f2e2e033ac5d5b2d057e2d",
    "binary_sha256": "5933e8c702d482d075d9b88a579a8bb0d6feaa50b8e8e2ed69ccd5ad23c23826",
    "pid": 47910,
    "gateway_url": "http://127.0.0.1:18788",
    "upstream": "http://127.0.0.1:18081/v1",
    "rollback_upstream": "http://127.0.0.1:18080/v1"
  },
  "inferflux": {
    "source_commit": "c5d4eb89f71dcfb3064b974d63b88c4ce88f42a1",
    "binary_sha256": "312f49e63dae7609dd804e3b74aa4a4b2dd06b72aebf8aa6a322ed619e974de6",
    "pid": 2254917,
    "endpoint": "aiserver1 loopback8081",
    "model": "qwen3-coder-30b",
    "backend": "llama_cpp_cuda",
    "scope": "Accepted partial CUDA offload; not full Qwen GPU or R9700 acceptance",
    "gate_url": "https://github.com/anvai-labs/inferflux/actions/runs/35431794400"
  },
  "timeouts": {
    "gateway_buffered_seconds": 120,
    "gateway_transparent_retries": 0,
    "gateway_stream_setup_seconds": 30,
    "gateway_stream_idle_seconds": 90,
    "gateway_reservation_ttl_seconds": 900,
    "victor_client_seconds": 120,
    "victor_inferflux_client_retries": 2,
    "victor_zai_client_retries": 3,
    "subagent_attempts": 3,
    "subagent_backoff_initial_seconds": 1,
    "subagent_backoff_cap_seconds": 60,
    "harness_forwarder_seconds": 300,
    "observer_seconds": 900,
    "observer_whole_run_bound_seconds": 2400,
    "observed_timeouts": 0,
    "no_limits_changed": true
  },
  "wire_sqlite_c4_dashboard": {
    "calls": 27,
    "distinct_sessions": 7,
    "http_200": 27,
    "timeouts": 0,
    "max_seconds": 62.38,
    "explicit_reporting_coverage": 27,
    "c4_pages": 41,
    "by_provider": {
      "inferflux": {
        "calls": 17,
        "tokens_in": 32281,
        "tokens_out": 2469,
        "cache_read_tokens": 0,
        "cache_creation_tokens": 0
      },
      "zai": {
        "calls": 10,
        "tokens_in": 8892,
        "tokens_out": 15807,
        "cache_read_tokens": 41664,
        "cache_creation_tokens": 0
      }
    }
  },
  "inclusive_input_tokens": 82837,
  "output_tokens": 18276,
  "review_verdict": "needs_work",
  "review_counterexamples_independently_verified": true,
  "missing_deliverables": [
    "fallbacka.py",
    "test_fallbacka.py"
  ],
  "pytest": {
    "exit_code": 0,
    "passed": 6,
    "expected_member_test_files": 7
  },
  "original_harness_exit_code": 1,
  "original_harness_failed_assertion": "review[\"verdict\"] == \"approved\"",
  "victor_member_usage_reconciliation": "NOT REACHED: unchanged harness asserts the review verdict before member-usage reconciliation and evidence.json persistence; gateway wire/ledger evidence was retained externally",
  "startup_failure_preserved": {
    "run_id": "gateway-422aab45c2",
    "model_calls": 0,
    "cause": "Fresh-state ZAI constructor credential validation; corrected by privately reusing the existing gateway virtual key",
    "replayed_inference": false
  },
  "limitations": [
    "No inference retry or timeout is erased from this verdict.",
    "Existing harness buffers responses; no streaming, origin cancellation or TTFT acceptance claimed.",
    "Cache-field reporting is not proof of backend executed reuse; completed-cache diagnostics not captured here.",
    "Accepted runtime is partial CUDA offload, not R9700 or full GPU Qwen acceptance.",
    "Enabled-session lease behavior and exact sampled completion-token units remain outside this run.",
    "C4 completeness/basis/outcome/physical attempts remain unavailable.",
    "The full mixed-team acceptance verdict fails; keep C5 and InferFlux #184 open.",
    "The earlier passing five-call stream replay and original failed evidence were not modified or repeated."
  ],
  "started_at": "2026-09-19T18:06:28.802632+00:00",
  "finished_at": "2026-09-19T18:19:30.791192+00:00",
  "preflight": {
    "victor_tests_passed": 94,
    "sandhi_terminal_accounting_tests_passed": 4
  }
}