Every run,
with its evaluation.

14 lineages, one gallery each.

v2 · 5 runs · judge 32.8

claude-code · Opus 5.5

anthropic/claude-opus-5.5

v2 · 5 runs · judge 25.3

claude-code · Sonnet 5

anthropic/claude-sonnet-5

v2 · 5 runs · judge 28

opencode · GPT-5.6 Terra

openai/gpt-5.6-terra

v2 · 1 run · judge 32.8

opencode · Gemini 3.8 Flash

google/gemini-3.8-flash

v2 · 5 runs · judge 31.7

hermes · Jev Router

typesafe/jev-router

v1 · 10 runs · judge 29.8

gemini-cli · Gemini 3.5 Flash

gemini-3.5-flash

v1 · 10 runs · judge 29.3

opencode · Gemini 3.5 Flash

google/gemini-3.5-flash

v1 · 3 runs · judge 28.8

hermes · Fusion router

openrouter/fusion

v1 · 10 runs · judge 28.7

claude-code · Opus 4.7

claude-opus-4-7

v1 · 3 runs · judge 28.3

opencode · Fusion router

openrouter/fusion

v1 · 10 runs · judge 28.2

hermes · Gemini 3.5 Flash

openrouter/google/gemini-3.5-flash

v1 · 10 runs · judge 27.3

hermes · GLM 5.1

openrouter/z-ai/glm-5.1

v1 · 11 runs · judge 26.2

claude-code · Sonnet 4.6

claude-sonnet-4-6

v1 · 10 runs · judge 24

hermes · Sonnet 4.6

openrouter/anthropic/claude-sonnet-4.6

Scores, verdicts, cost and trajectory.

About the verdict: these runs used runbook v1, which wrote no validation report, so Jetty computed no verdict for them. The verdict here is derived the way the v2 runbook would score it, with the external judge in place of the self-score: pass when every replayable code check passes on the stored SVG (svg-parses, svg-size-under-50kb, no-image-tags, png-rendered) and the judge total is at least 28/40. New runs of the v2 runbook get Jetty's own verdict. Cost is Jetty's metered estimate; opencode and hermes runs reported no token usage, so they are marked not metered rather than zero.

Launch runs from the API.

Once the task exists in your collection (see Author · run it on your Jetty), a run is one multipart POST. Top-level keys you pass replace the task defaults; keys you omit (snapshot, file_paths, runbook_evals_version) keep the task's values.

1. Run it

   Copy

   # Launch a run. Anything in init_params overrides the task defaults (top-level keys are replaced)
   curl -s -X POST "https://flows-api.jetty.io/api/v1/run/$COLLECTION/pelican-bicycle-svg" \
     -H "Authorization: Bearer $JETTY_TOKEN" \
     -F 'init_params={"agent": "claude-code", "model": "anthropic/claude-sonnet-5", "model_provider": "openrouter",
                      "vars": {"prompt": "Execute the runbook end-to-end. You are headless: never stop to ask questions.",
                               "results_dir": "/app/results", "max_rounds": 3, "pass_threshold": 28}}'
   # => {"trajectory_id": "…", "workflow_id": "…"}
   
   # Other agent/model pairs we ran (all model_provider=openrouter):
   #   claude-code  anthropic/claude-opus-5.5
   #   opencode     openrouter/google/gemini-3.8-flash
   #   opencode     openrouter/openai/gpt-5.6-terra
   #   hermes       openrouter/typesafe/jev-router

2. Read the verdict and collect the drawing

   Copy

   # Poll until status is completed, then read Jetty's evaluation and download the outputs
   curl -s "https://flows-api.jetty.io/api/v1/db/trajectory/$COLLECTION/pelican-bicycle-svg/$TRAJ" \
     -H "Authorization: Bearer $JETTY_TOKEN" | jq '.status, .attributes.evaluation.verdict, .steps.run.outputs.usage'
   curl -s -o pelican.zip "https://flows-api.jetty.io/api/v1/trajectory/$COLLECTION/pelican-bicycle-svg/$TRAJ/download" \
     -H "Authorization: Bearer $JETTY_TOKEN"
   # final.svg, final.png, scores.json, report.md, validation_report.json, rounds/
   # or open https://jetty.io/$COLLECTION/pelican-bicycle-svg/$TRAJ