Auditable Compensation & Hiring Event API

TREE-001 is a deterministic Apify Actor that monitors public Greenhouse, Lever, and Ashby job boards and emits auditable, machine-readable material hiring events.

Commercial contract

The Actor does not claim to know whether a position was filled, whether a company is growing, or why a posting changed. It reports observed changes from public sources, with source URLs, before/after facts, reason codes, source-completeness status, and deterministic event fingerprints.

V1 event types

• JOB_ADDED
• JOB_REMOVED_CONFIRMED only after two consecutive complete checks
• JOB_REOPENED
• COMPENSATION_CHANGED
• LOCATION_CHANGED
• WORKPLACE_TYPE_CHANGED
• JOB_MATERIAL_FIELDS_CHANGED
• DEPARTMENT_BUILDOUT

Source failures generate free source_error records and never trigger removal events.

State model

Use the same monitorNamespace for recurring checks. Each provider/board receives a separate persistent record in a named Apify key-value store. Named stores are required because state must survive individual Actor runs.

Input example

{
  "monitorNamespace": "portfolio-watchlist",
  "targets": [
    {"provider": "greenhouse", "board": "company-slug", "company": "Company A"},
    {"provider": "lever", "board": "company-slug", "company": "Company B", "region": "global"},
    {"provider": "ashby", "board": "company-slug", "company": "Company C"}
  ],
  "maxTargets": 25,
  "maxConcurrency": 5,
  "timeoutSeconds": 20,
  "thresholds": {"departmentAbsoluteChange": 5, "departmentPercentChange": 0.5}
}

Output guarantees

• First successful run establishes a baseline and reports no historical events.
• A failed or incomplete source check does not update state.
• A missing job is not declared removed after one check.
• Material field changes include explicit before/after values.
• Event fingerprints are deterministic for the compared facts.
• Unsupported targets result in structured errors.

Resource controls

• V1 supports only Greenhouse, Lever, and Ashby.
• No browser automation, proxies, paid APIs, or LLM calls.
• maxTargets defaults to 25 and is capped at 100 by the input schema.
• maxConcurrency is capped at 8.
• HTTP timeout is capped at 30 seconds.
• Failed sources are not billed as successful checks.

Monetization configuration

Configure these pay-per-event events in Apify Console:

• successful-source-check: proposed validation price $0.01
• material-event: proposed validation price $0.05

These are launch hypotheses. Do not publish final prices until benchmark runs measure compute and storage costs. The code checks Apify's charge limit before pushing billable records.

Cost benchmark procedure

1. Run three baselines containing 1, 10, and 25 boards.
2. Record compute units, wall-clock duration, dataset/storage volume, and source-success rate.
3. Repeat each set without changes.
4. Run controlled fixture-based changes to test event volume.
5. Calculate direct cost per successful board check and per material event.
6. Require a commercially sustainable contribution margin before publication.
7. Do not reduce price merely to attain the $1 Genesis milestone.

Local tests

python -m pip install pytest pytest-asyncio httpx
pytest -q

Deployment package

The repository includes:

• Apify Actor manifest
• input and dataset schemas
• Dockerfile
• source connectors
• stateful comparison engine
• pay-per-event hooks
• automated unit/connector tests
• one-time owner deployment checklist

Legal and data boundaries

• Public job-posting data only.
• No applicant, candidate, employee, or private ATS data.
• No login bypass, cookies, credential collection, or application submission.
• No inference of protected characteristics.
• No claim that an observed event proves corporate intent or performance.
• Users are responsible for their lawful use of output and compliance with source/platform terms.