atlas-iac/docs/evidence/hermes_suite_acceptance_20260929.json

302 lines
8.0 KiB
JSON

{
"configuration_revision": "suite-v6-20260929",
"deployed_commit": "299b9ed16d79aeefb3dc5590a19dc07dc54167e7",
"date": "2026-09-29",
"client_host": "titan-jh",
"client_lan_ip": "192.168.22.8",
"destination": "192.168.22.50:443",
"tls_hostname": "worker.bstein.dev",
"proxy_bypass": true,
"laptop_test_performed": false,
"synthetic_only": true,
"results": [
{
"size": 14,
"http_status": 200,
"request_bytes": 9847,
"client_wall_seconds": 16.947,
"exact_alias_coverage": true,
"server_wall_seconds": 16.706,
"provider": "claude",
"model": "claude-opus-4-8",
"context_limit": 1000000,
"output_limit": 64000,
"input_tokens": 4185,
"output_tokens": 1250,
"cli_turns": 2,
"cost_usd_estimate": 0.052175,
"singletons": 4,
"quality": {
"coverage": true,
"families": 9,
"pair_precision": 1.0,
"pair_recall": 1.0,
"false_merge_pairs": 0,
"missed_merge_pairs": 0
},
"description_review": "No observed cross-family objective attribution; manual review against synthetic source patterns.",
"compaction": false,
"truncation": false,
"attempted_destinations": [
"switchyard:atlas/planning/claude",
"claude:claude-opus-4-8"
],
"job_id": "efa5151c8dfa4f35a52dc83244d64762",
"provider_job_attempts": 1,
"automatic_retries_observed": 0,
"rate_limit_events_observed": 0
},
{
"size": 75,
"http_status": 200,
"request_bytes": 55984,
"client_wall_seconds": 25.755,
"exact_alias_coverage": true,
"server_wall_seconds": 24.922,
"provider": "claude",
"model": "claude-opus-4-8",
"context_limit": 1000000,
"output_limit": 64000,
"input_tokens": 19145,
"output_tokens": 2336,
"cli_turns": 2,
"cost_usd_estimate": 0.15412499999999998,
"singletons": 3,
"quality": {
"coverage": true,
"families": 9,
"pair_precision": 1.0,
"pair_recall": 1.0,
"false_merge_pairs": 0,
"missed_merge_pairs": 0
},
"description_review": "No observed cross-family objective attribution; manual review against synthetic source patterns.",
"compaction": false,
"truncation": false,
"attempted_destinations": [
"switchyard:atlas/planning/claude",
"claude:claude-opus-4-8"
],
"job_id": "f0e076ffe90643779593d778920cffc7",
"provider_job_attempts": 1,
"automatic_retries_observed": 0,
"rate_limit_events_observed": 0
},
{
"size": 363,
"http_status": 200,
"request_bytes": 273761,
"client_wall_seconds": 80.071,
"exact_alias_coverage": true,
"server_wall_seconds": 78.828,
"provider": "claude",
"model": "claude-opus-4-8",
"context_limit": 1000000,
"output_limit": 64000,
"input_tokens": 89753,
"output_tokens": 8568,
"cli_turns": 2,
"cost_usd_estimate": 0.6629649999999999,
"singletons": 3,
"quality": {
"coverage": true,
"families": 9,
"pair_precision": 1.0,
"pair_recall": 1.0,
"false_merge_pairs": 0,
"missed_merge_pairs": 0
},
"description_review": "No observed cross-family objective attribution; manual review against synthetic source patterns.",
"compaction": false,
"truncation": false,
"attempted_destinations": [
"switchyard:atlas/planning/claude",
"claude:claude-opus-4-8"
],
"job_id": "682e87d11aa344c3b4bd33bcab7ea431",
"provider_job_attempts": 1,
"automatic_retries_observed": 0,
"rate_limit_events_observed": 0
}
],
"field_lengths": {
"description": {
"min_bytes": 48,
"median_bytes": 234,
"max_bytes": 245
},
"preconditions": {
"min_bytes": 67,
"median_bytes": 136,
"max_bytes": 143
},
"case_type": {
"min_bytes": 7,
"median_bytes": 8,
"max_bytes": 15
},
"success_criteria": {
"min_bytes": 73,
"median_bytes": 132,
"max_bytes": 135
}
},
"whole_input_mock_transport": [
{
"cases": 14,
"requested": "claude-opus-4-8[1m]",
"captured": [
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 12832,
"complete_source": true
},
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 12818,
"complete_source": true
}
],
"returncode": 0
},
{
"cases": 75,
"requested": "claude-opus-4-8[1m]",
"captured": [
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 60921,
"complete_source": true
},
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 60907,
"complete_source": true
}
],
"returncode": 0
},
{
"cases": 363,
"requested": "claude-opus-4-8[1m]",
"captured": [
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 287914,
"complete_source": true
},
{
"path": "/v1/messages?beta=true",
"model": "claude-opus-4-8",
"max_tokens": 64000,
"bytes": 287900,
"complete_source": true
}
],
"returncode": 0
}
],
"policy_checks": {
"synthetic_scope": {
"configuration_revision": "suite-v6-20260929",
"allowed_external_providers": [
"claude",
"codex"
],
"generalized_external_providers": [],
"external_scope": "exact_synthetic_fixtures"
},
"local_scope": {
"configuration_revision": "suite-v6-20260929",
"allowed_external_providers": [],
"generalized_external_providers": [],
"external_scope": "exact_synthetic_fixtures"
},
"operational_scope": {
"configuration_revision": "suite-v6-20260929",
"allowed_external_providers": [
"claude"
],
"generalized_external_providers": [
"claude"
],
"external_scope": "generalized_claude_and_exact_synthetic_fixtures"
},
"synthetic_modified_synthetic_preflight": {
"status": 403,
"error": "external_data_not_approved",
"provider": null
},
"local_modified_synthetic_preflight": {
"status": 403,
"error": "provider_forbidden",
"provider": null
},
"operational_modified_synthetic_preflight": {
"status": 200,
"error": null,
"provider": "claude"
},
"operational_codex": {
"status": 403,
"error": "provider_forbidden"
},
"operational_claude_codex": {
"status": 403,
"error": "provider_forbidden"
},
"operational_unknown": {
"status": 400,
"error": "unknown_provider"
},
"local_only_True": {
"status": 422,
"error": {
"code": "capacity_or_unsupported_backend",
"details": {
"candidates": {
"local": "capacity"
},
"input_bytes": 11069,
"output_estimate": 2718
}
}
},
"local_only_False": {
"status": 422,
"error": {
"code": "capacity_or_unsupported_backend",
"details": {
"candidates": {
"local": "capacity"
},
"input_bytes": 11069,
"output_estimate": 2718
}
}
}
},
"oversize_curl": {
"http_status": 413,
"expect_continue": true
},
"unknowns": [
"Provider training/data-sharing/retention settings",
"Physical provider serving hardware and region",
"Remaining account quota",
"Accurate tokenizer and full 1M-context benchmark",
"Real-case grouping quality",
"Model internal attention to every input field"
]
}