Skip to content

workflow.md

Latest commit

 

History

History
39 lines (28 loc) · 2.06 KB

workflow.md

File metadata and controls

39 lines (28 loc) · 2.06 KB
Raw

OpenHound Collector Development Workflow

Use this workflow when developing a new OpenHound collector or making broad collector changes. This file describes the order of work only. Implementation details belong in the relevant skill files under .agents/skills/.

Development Flow

  1. Understand the target service.

    • Identify credentials, API base URL, pagination model, primary resources, stable IDs, and relationships.
    • Read .agents/standards/openhound.md before changing collector code.
    • Use .agents/skills/openhound/references/plan-collector.md.

  2. Define the collector graph shape.

    • Decide the source prefix, node kinds, edge kinds, common node properties, and node ID strategy.
    • Use .agents/skills/openhound/references/graph-schema.md.

  3. Register the extension and pipeline phases.

    • Wire collect, optional preproc, convert, extension metadata, and package entry points.
    • Use .agents/skills/openhound/references/register-extension.md.

  4. Implement source collection.

    • Add or update API clients, source context, DLT resources, transformers, auth, and secrets.
    • Use .agents/skills/openhound/references/source-collection.md.

  5. Add collected assets and relationships.

    • Add Pydantic asset models, graph property dataclasses, kind constants, exports, as_node, and edges.
    • Use .agents/skills/openhound/references/add-asset.md.

  6. Add preprocessing and lookup only when needed.

    • Use this when conversion needs cross-table joins, derived tables, or relationship resolution from collected data.
    • Use .agents/skills/openhound/references/preproc-lookup.md.

  7. Validate the collector.

    • Check structure, run available tests and static checks, and confirm collect/preproc/convert behavior where possible.
    • Use .agents/skills/openhound/references/validate-extension.md.

Working Rules

  • Keep AGENTS.md as the entrypoint and .agents/standards/openhound.md as the source of OpenHound rules.
  • Use a uv virtual environment outside the repository for validation commands to avoid modifying the user's local .venv.