Workflow arms A/B/D — plumber (rescue test)

Profile

A-luna-baseline

ready · QA pass

arm baselineexemplar —source creativeregen 0

7,956 in9,905 out$0.1020 cost91.8s workflow wall66.3s LLM call time

Step	Provider:model	Effort	Prompt variant	Tokens	Cost	Step wall time
`plan`	openrouter:anthropic/claude-opus-4.8	—	default	1,690 / 1,514	$0.0463	20.9s
`art_direction`	openai:gpt-5.6-luna	medium	default	—	—	—
`theme_emit`	openai:gpt-5.6-luna	medium	landing	2,606 / 7,449	$0.0473	39.2s
`page_generate`	openai:gpt-5.6-luna	medium	landing	—	—	—
`honesty_repair`	openai:gpt-5.6-luna	medium	landing	—	—	—
`honesty_judge`	openrouter:anthropic/claude-haiku-4.5	—	default	3,660 / 942	$0.0084	6.3s
`restyle`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—

Submitted profile JSON

{
  "name": "A-luna-baseline",
  "workflowMode": "baseline",
  "steps": {
    "plan": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "restyle": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "honesty_judge": {
      "provider": "openrouter",
      "model": "anthropic/claude-haiku-4.5"
    },
    "theme_emit": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    },
    "page_generate": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    },
    "honesty_repair": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    }
  }
}

Resolved run profile

[
  {
    "step": "plan",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  },
  {
    "step": "art_direction",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 1600,
    "temperature": 0.4,
    "timeoutMs": 90000
  },
  {
    "step": "theme_emit",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "page_generate",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_repair",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_judge",
    "role": "claims_judge",
    "provider": "openrouter",
    "model": "anthropic/claude-haiku-4.5",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 4096,
    "temperature": 0,
    "timeoutMs": 60000
  },
  {
    "step": "restyle",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  }
]

result 3734dff1-3e2b-4a1d-9d70-886ed896c010 job 8884787d-da01-478a-9559-c8c2f73e948a site bo-site-3734dff1-3e2b-4a1d-9d70-886ed896c010

Profile

A-opus-baseline

ready · QA pass

arm baselineexemplar —source creativeregen 0

A-opus-baseline ↗$0.2946

9,803 in11,374 out$0.2946 cost139.8s workflow wall116.7s LLM call time

Step	Provider:model	Effort	Prompt variant	Tokens	Cost	Step wall time
`plan`	openrouter:anthropic/claude-opus-4.8	—	default	1,690 / 1,365	$0.0426	19.4s
`art_direction`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—
`theme_emit`	openrouter:anthropic/claude-opus-4.8	—	landing	4,256 / 8,844	$0.2424	89.6s
`page_generate`	openrouter:anthropic/claude-opus-4.8	—	landing	—	—	—
`honesty_repair`	openrouter:anthropic/claude-opus-4.8	—	landing	—	—	—
`honesty_judge`	openrouter:anthropic/claude-haiku-4.5	—	default	3,857 / 1,165	$0.0097	7.7s
`restyle`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—

Submitted profile JSON

{
  "name": "A-opus-baseline",
  "workflowMode": "baseline",
  "steps": {
    "plan": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "restyle": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "honesty_judge": {
      "provider": "openrouter",
      "model": "anthropic/claude-haiku-4.5"
    },
    "theme_emit": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8",
      "promptVariant": "landing"
    },
    "page_generate": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8",
      "promptVariant": "landing"
    },
    "honesty_repair": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8",
      "promptVariant": "landing"
    }
  }
}

Resolved run profile

[
  {
    "step": "plan",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  },
  {
    "step": "art_direction",
    "role": "generation",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 1600,
    "temperature": 0.4,
    "timeoutMs": 90000
  },
  {
    "step": "theme_emit",
    "role": "generation",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 300000
  },
  {
    "step": "page_generate",
    "role": "generation",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 300000
  },
  {
    "step": "honesty_repair",
    "role": "generation",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 300000
  },
  {
    "step": "honesty_judge",
    "role": "claims_judge",
    "provider": "openrouter",
    "model": "anthropic/claude-haiku-4.5",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 4096,
    "temperature": 0,
    "timeoutMs": 60000
  },
  {
    "step": "restyle",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  }
]

result 8a0b29c7-83b4-4a63-a9c3-6ecb5cded9ab job bef0e02a-d01e-461e-b820-13cebc9e9899 site bo-site-8a0b29c7-83b4-4a63-a9c3-6ecb5cded9ab

Profile

B-luna-exemplar

ready · QA pass

arm exemplarexemplar type-first-urgent-v1source creativeregen 1 · deterministic:entity_address_mismatch,unbacked_service_area

B-luna-exemplar ↗$0.1007

9,776 in15,374 out$0.1007 cost107.0s workflow wall85.8s LLM call time

Step	Provider:model	Effort	Prompt variant	Tokens	Cost	Step wall time
`plan`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—
`art_direction`	openai:gpt-5.6-luna	medium	default	—	—	—
`theme_emit`	openai:gpt-5.6-luna	medium	landing	2,852 / 6,896	$0.0442	38.9s
`page_generate`	openai:gpt-5.6-luna	medium	landing	—	—	—
`honesty_repair`	openai:gpt-5.6-luna	medium	landing	3,221 / 7,197	$0.0464	39.8s
`honesty_judge`	openrouter:anthropic/claude-haiku-4.5	—	default	3,703 / 1,281	$0.0101	7.1s
`restyle`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—

Submitted profile JSON

{
  "name": "B-luna-exemplar",
  "workflowMode": "exemplar",
  "steps": {
    "plan": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "restyle": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "honesty_judge": {
      "provider": "openrouter",
      "model": "anthropic/claude-haiku-4.5"
    },
    "theme_emit": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    },
    "page_generate": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    },
    "honesty_repair": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "landing"
    }
  }
}

Resolved run profile

[
  {
    "step": "plan",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  },
  {
    "step": "art_direction",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 1600,
    "temperature": 0.4,
    "timeoutMs": 90000
  },
  {
    "step": "theme_emit",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "page_generate",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_repair",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "landing",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_judge",
    "role": "claims_judge",
    "provider": "openrouter",
    "model": "anthropic/claude-haiku-4.5",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 4096,
    "temperature": 0,
    "timeoutMs": 60000
  },
  {
    "step": "restyle",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  }
]

result 82a727f6-c21f-4a33-8fb9-0f98562d57da job 975e65dd-b0d3-4a4b-a36d-da71fbffb1bc site bo-site-82a727f6-c21f-4a33-8fb9-0f98562d57da

Profile

D-luna-artdir

ready · QA pass

arm art_directionexemplar —direction gpt-5.6-luna → gpt-5.6-luna · A single restrained copper pipe line with one elbow joint, repeated only as dividers and terminating beside the tap-to-call control; no photos, icons, or extra ornament.source fallbackregen 0

D-luna-artdir ↗$0.0158

4,247 in2,140 out$0.0158 cost39.4s workflow wall17.0s LLM call time

Step	Provider:model	Effort	Prompt variant	Tokens	Cost	Step wall time
`plan`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—
`art_direction`	openai:gpt-5.6-luna	medium	focused	931 / 884	$0.0062	9.3s
`theme_emit`	openai:gpt-5.6-luna	medium	default	0 / 0	$0.0000	607ms
`page_generate`	openai:gpt-5.6-luna	medium	default	—	—	—
`honesty_repair`	openai:gpt-5.6-luna	medium	default	—	—	—
`honesty_judge`	openrouter:anthropic/claude-haiku-4.5	—	default	3,316 / 1,256	$0.0096	7.1s
`restyle`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—

Submitted profile JSON

{
  "name": "D-luna-artdir",
  "workflowMode": "art_direction",
  "steps": {
    "plan": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "restyle": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "honesty_judge": {
      "provider": "openrouter",
      "model": "anthropic/claude-haiku-4.5"
    },
    "art_direction": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "focused"
    },
    "theme_emit": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    },
    "page_generate": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    },
    "honesty_repair": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    }
  }
}

Resolved run profile

[
  {
    "step": "plan",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  },
  {
    "step": "art_direction",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "focused",
    "maxTokens": 1600,
    "temperature": 0.4,
    "timeoutMs": 240000
  },
  {
    "step": "theme_emit",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "page_generate",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_repair",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_judge",
    "role": "claims_judge",
    "provider": "openrouter",
    "model": "anthropic/claude-haiku-4.5",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 4096,
    "temperature": 0,
    "timeoutMs": 60000
  },
  {
    "step": "restyle",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  }
]

result 69462064-e7af-4d96-8a4e-24b7f74c3a96 job 5d7da474-17bf-4a27-a46c-b21e72e901a7 site bo-site-69462064-e7af-4d96-8a4e-24b7f74c3a96

Profile

D-terra-directs-luna

ready · QA pass

arm art_directionexemplar —direction gpt-5.6-terra → gpt-5.6-luna · Repeat one thin cobalt pipe-route line with right-angle turns; it connects section anchors, service links, map pin, and ends at each call action.source creativeregen 0

D-terra-directs-luna ↗$0.0668

8,820 in8,455 out$0.0668 cost73.4s workflow wall51.9s LLM call time

Step	Provider:model	Effort	Prompt variant	Tokens	Cost	Step wall time
`plan`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—
`art_direction`	openai:gpt-5.6-terra	medium	focused	2,157 / 578	$0.0141	7.6s
`theme_emit`	openai:gpt-5.6-luna	medium	default	2,934 / 6,653	$0.0429	37.1s
`page_generate`	openai:gpt-5.6-luna	medium	default	—	—	—
`honesty_repair`	openai:gpt-5.6-luna	medium	default	—	—	—
`honesty_judge`	openrouter:anthropic/claude-haiku-4.5	—	default	3,729 / 1,224	$0.0098	7.2s
`restyle`	openrouter:anthropic/claude-opus-4.8	—	default	—	—	—

Submitted profile JSON

{
  "name": "D-terra-directs-luna",
  "workflowMode": "art_direction",
  "steps": {
    "plan": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "restyle": {
      "provider": "openrouter",
      "model": "anthropic/claude-opus-4.8"
    },
    "honesty_judge": {
      "provider": "openrouter",
      "model": "anthropic/claude-haiku-4.5"
    },
    "art_direction": {
      "provider": "openai",
      "model": "gpt-5.6-terra",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "focused"
    },
    "theme_emit": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    },
    "page_generate": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    },
    "honesty_repair": {
      "provider": "openai",
      "model": "gpt-5.6-luna",
      "effort": "medium",
      "timeoutMs": 240000,
      "promptVariant": "default"
    }
  }
}

Resolved run profile

[
  {
    "step": "plan",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  },
  {
    "step": "art_direction",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-terra",
    "effort": "medium",
    "promptVariant": "focused",
    "maxTokens": 1600,
    "temperature": 0.4,
    "timeoutMs": 240000
  },
  {
    "step": "theme_emit",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "page_generate",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_repair",
    "role": "generation",
    "provider": "openai",
    "model": "gpt-5.6-luna",
    "effort": "medium",
    "promptVariant": "default",
    "maxTokens": 20000,
    "temperature": 0.85,
    "timeoutMs": 240000
  },
  {
    "step": "honesty_judge",
    "role": "claims_judge",
    "provider": "openrouter",
    "model": "anthropic/claude-haiku-4.5",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 4096,
    "temperature": 0,
    "timeoutMs": 60000
  },
  {
    "step": "restyle",
    "role": "plan",
    "provider": "openrouter",
    "model": "anthropic/claude-opus-4.8",
    "effort": null,
    "promptVariant": "default",
    "maxTokens": 8192,
    "temperature": 0.2,
    "timeoutMs": 120000
  }
]

result 5f235cb0-1613-47ba-a1a9-3b0fd237cb63 job 96cd4d0c-5ef4-4cff-a89b-db3a3e66a4e5 site bo-site-5f235cb0-1613-47ba-a1a9-3b0fd237cb63

Workflow provenance

Git SHA: 7755bdf
Captured: Jul 15, 2026, 5:25 PM UTC

startValidate business data; mark the job and bakeoff result running.
pipeline / planBuild the full SEO page plan through the selected plan profile.
pipeline / generateGenerate the homepage theme and scoped pages, repairing or falling back safely.
pipeline / assembleAssemble raw HTML/CSS/JS with shared SEO, navigation, business facts, and legal pages.
pipeline / auditRun deterministic QA plus the independent honesty judge.
pipeline / writeWrite the validated mutable draft and its source/provenance artifacts to tenant R2.
finishFinalize generation job, site metadata, and page rows in D1.
domainsBest-effort domain candidate discovery used by the real claim path.
bakeoff-finalizeBakeoff-only: publish an immutable preview snapshot and roll up LLM usage.