Skip to content

Application

Persona-driven product scenarios under application/tasks/: a simulated user runs a survey, chatbot, website, or native app; a verifier scores the trial; batch reporting rolls up a cohort.

Task types

Type Reference task Notes
Survey example-survey_product-feedback Host auto; form / questionnaire
Chat example-chat-api_support_chatbot Host auto + optional sidecar
Web example-web-playwright_quote-choice Docker; see web interaction
OS-app example-computer-use-linux_note-to-csv Docker or use.computer (macOS/iOS)

Author a task (one path)

  1. Contractsapplication/task-spec/README.md (type diagrams, metrics, reporting). Deep dives: application/task-spec/docs/.
  2. Checklisttask-guide.md (task.toml, registration).
  3. Copy an example-* under application/tasks/, edit, smoke via Quickstart.

Ops notes also live in application/tasks/README.md.

Run

How Where
Hands-on path Quickstart
Playground UI Quickstart § Playground
HTTP API playground-api.md
Job recipes / generator Configuration · Quickstart § CLI
uv run python application/scripts/generate_application_job.py \
  --task application/tasks/example-survey_product-feedback \
  --execution-mode auto \
  --persona-ids 0042
uv run harbor run -c configs/jobs/application-task-job-recipe/<generated>.yaml
Doc Role
Handbook Docs home
Quickstart First runs
Task contracts Specs and metrics
Persona Pools, 1M, sampling
Environment Matraix Playground, agents, images