feat: Add generic preset metadata field to Automation model

[malhotra5]

Summary

Follows #51. Once repository support lands for preset endpoints, we should record the repository configuration (and other preset-specific data) in a generic metadata field that the UI can consume.

Background

Preset automations (/v1/preset/prompt and /v1/preset/plugin) are built from user-provided configuration that gets packaged into a tarball. Currently, only the prompt field is stored on the Automation model (see models.py:56). The generated tarball contains additional configuration that would be useful for the UI to display and for users to understand what their automation does:

1. Repositories (from feat: add repository support to preset automation endpoints #51) — cloned into the sandbox at runtime
2. Plugins — the PluginSource list used by /v1/preset/plugin (stored in plugins_config.json inside tarball)
3. Prompt — already stored as a top-level field, but could be included in metadata for consistency

Proposal

Add a preset_metadata JSONB column to the Automation model:

# Preset-specific metadata (populated by preset endpoints, NULL for custom SDK automations)
preset_metadata: Mapped[dict | None] = mapped_column(JSONB, nullable=True)

Metadata Schema

The field is intentionally generic to accommodate future presets:

{
  "preset_type": "prompt" | "plugin",
  "prompt": "...",
  "plugins": [{"source": "github:owner/repo", "ref": "v1.0.0"}],
  "repos": [{"url": "OpenHands/repo", "ref": "main"}]
}

For prompt presets: preset_type, prompt, repos (once #51 lands)
For plugin presets: preset_type, prompt, plugins, repos

Changes Required

1. models.py — Add preset_metadata JSONB column
2. schemas.py — Add preset_metadata: dict | None to AutomationResponse
3. preset_router.py — Populate preset_metadata when creating automations:
   • Prompt preset: {"preset_type": "prompt", "prompt": body.prompt, "repos": [...]}
   • Plugin preset: {"preset_type": "plugin", "prompt": body.prompt, "plugins": [...], "repos": [...]}
4. Migration — Add the new column

Note

This does not apply to custom SDK automations (created via /v1 with a user-uploaded tarball), which will have preset_metadata = NULL.

---

This issue was created by an AI assistant (OpenHands).

[linear]

APP-1382

[malhotra5]

@OpenHands please do this

[openhands-ai]

I'm on it! malhotra5 can track my progress at all-hands.dev

[openhands-ai]

Final Summary

I have successfully implemented issue #56: "feat: Add generic preset metadata field to Automation model".

Checklist - All Requirements Addressed

✅ Model change - Added preset_metadata JSONB column to Automation model (nullable)

✅ Schema change - Added preset_metadata: dict | None to AutomationResponse

✅ Prompt preset endpoint - Populates preset_metadata with:

• preset_type: "prompt"
• prompt: user's prompt
• repos: (optional) when repos specified

✅ Plugin preset endpoint - Populates preset_metadata with:

• preset_type: "plugin"
• prompt: user's prompt
• plugins: list of plugin sources
• repos: (optional) when repos specified

✅ Migration created - 005_add_preset_metadata_column.py

✅ Tests added - 5 comprehensive tests in TestPresetMetadata class

✅ Custom SDK automations - Will have preset_metadata = NULL (no changes needed)

✅ Linting passed - All ruff checks pass

✅ Tests passed - 356 unit tests pass

✅ Branch created - openhands/add-preset-metadata-field

✅ PR opened - #89 (closes #56)

Changes Are Concise

All changes are directly related to the feature request:

• 5 files modified/created
• No extraneous changes
• Clean, minimal implementation following existing patterns

The implementation matches the exact schema proposed in the issue and handles both preset types (prompt and plugin) with optional repos field.