AI JSON Workflow

JSON-based workflow for AI-driven config generation and external tooling.

Why JSON?

• AI Design - Describe targeting in plain English, AI generates config
• Self-Documenting - Every JSON includes DesignNotes explaining settings
• Zero C++ Required - AI generates, you import. No compilation
• Version Control - JSON diffs show exactly what changed
• Bulk Editing - Edit multiple presets/styles in one file

Bidirectional Workflow

Import (AI to Unreal):

Export (Unreal to AI):

Data Assets are always the source of truth at runtime. JSON and DataTables sync with external tools.

Import: JSON to Data Asset

When AI generates new targeting configs:

1. AI Creates/Updates JSON

AI generates or updates DT_TargetingPresets.json or DT_TargetingStyles.json by referencing task definitions in DT_TargetingTasks.json.

AI tweaks sort weights and falloff in DT_TargetingPresets.json.

2. Reimport DataTables

1. Right-click DT_TargetingTags -> Reimport
2. Right-click DT_TargetingPresets or DT_TargetingStyles -> Reimport

3. Create Data Asset

1. Content Browser -> Right-click -> Miscellaneous -> Data Asset
2. Pick Agentic Targeting Preset or Agentic Targeting Style

4. Set the Tag

1. In the Details panel, set the Preset Tag (or Style Tag)
2. This must match a row in the DataTable (e.g., Agentic.Targeting.Preset.Freeflow)

5. Import from DataTable

1. Right-click on the new Data Asset
2. Select Import from DataTable in context menu

All properties are prepopulated from JSON. The Data Asset is ready for use.

Export: Data Asset to JSON

When designer makes changes in Editor:

1. Update DataTable

1. Right-click on the modified Data Asset
2. Select Update DataTable
3. The corresponding row in DataTable updates

2. Export JSON

1. Right-click on DataTable -> Asset Actions -> Export as JSON
2. Overwrite the existing JSON file
3. Now synced for external/AI workflows

JSON Structure

Preset JSON

{
    "Name": "MyPreset",
    "PresetTag": { "TagName": "Agentic.Targeting.Preset.MyPreset" },
    "DisplayName": "My Preset",
    "Description": "What this preset does",
    "Tasks": [
        {
            "TaskTag": { "TagName": "Agentic.Targeting.Task.AOE" },
            "Params": { "Radius": "1500" }
        },
        {
            "TaskTag": { "TagName": "Agentic.Targeting.Task.Filter.Targetable" },
            "Params": {}
        }
    ],
    "DesignNotes": "AI: Created for close-range melee with fast switching",
    "CreatedBy": "Claude"
}

Style JSON

{
    "Name": "MyStyle",
    "StyleTag": { "TagName": "Agentic.Targeting.Style.MyStyle" },
    "DisplayName": "My Style",
    "Description": "Custom lock behavior",
    "LockCapacity": "Single",
    "SoftLockUpdatePolicy": "None",
    "ValidationPolicy": "PoolOnly",
    "ValidationTickInterval": 1.0,
    "GracePeriod": 0.5,
    "RelockAction": "TryDifferent",
    "RelockFallback": "Release",
    "DefaultLockPresetTag": { "TagName": "Agentic.Targeting.Preset.Melee" },
    "DesignNotes": "AI: Souls-like with faster grace period",
    "CreatedBy": "Claude"
}

AI Design Loop

The complete AI-driven targeting design workflow:

AI Design Examples

Natural Language to Config

Prompt: "Make a Sekiro-style targeting - faster than Souls, auto-switch on death, 10m range"

AI generates:

{
    "Name": "SekiroStyle",
    "StyleTag": { "TagName": "Agentic.Targeting.Style.Sekiro" },
    "DisplayName": "Sekiro Quick Lock",
    "GracePeriod": 0.3,
    "ValidationFailAction": "TrySwitch",
    "DesignNotes": "Sekiro-style: Faster grace (0.3s vs 0.5s Souls), auto-switch on death"
}

Iterate with Feedback

User: "Lock breaks too easily during combat"

AI updates:

{
    "GracePeriod": 0.5,
    "DesignNotes": "Increased grace from 0.3s to 0.5s per user feedback"
}

File Locations

Best Practices

1. Always include DesignNotes - Document why, not just what
2. Use descriptive Names - "SoulsStyleMelee" not "Style1"
3. Keep JSON in sync - Export after Data Asset changes
4. Reference existing tasks - Check DT_TargetingTasks.json for valid TaskTags