docs/features/JSON_OUTPUT_FORMAT.md

LoRAInfoExtractor JSON Output Format

Overview

The LoRAInfoExtractor node now outputs structured JSON data instead of concatenated strings, providing better data handling and parsing capabilities for downstream processing.

Implementation Details

Return Types and Names

• RETURN_TYPES: ("STRING", "STRING", "WANVIDLORA")
• RETURN_NAMES: ("lora_json", "lora_info", "lora_passthrough")

Output Structure

The first output (lora_json) contains a JSON string with the following structure:

{
    "loras": [
        {
            "index": 0,
            "name": "LoRA Name",
            "info": "Detailed info string",
            "path": "/path/to/lora.safetensors",
            "strength": 1.0,
            "has_civitai_data": true
        }
    ],
    "count": 1,
    "combined_display": "LoRA Name + Another LoRA"
}

JSON Fields Description

Root Level Fields

• loras: Array of LoRA objects containing detailed metadata for each processed LoRA
• count: Integer representing the total number of valid LoRAs processed
• combined_display: String with " + " separated LoRA names for easy display (maintains backward compatibility)

LoRA Object Fields

• index: Zero-based index of the LoRA in the processing order
• name: Clean LoRA name (from CivitAI or extracted from filename)
• info: Human-readable information string including source (CivitAI/Local) and strength
• path: Full file path to the LoRA file
• strength: Numeric strength value from the WANVIDLORA configuration
• has_civitai_data: Boolean indicating whether CivitAI metadata was successfully retrieved

Example Outputs

Multiple LoRAs with Mixed Sources

{
    "loras": [
        {
            "index": 0,
            "name": "Realistic Vision",
            "info": "CivitAI: Realistic Vision (v5.1) by SomeCreator",
            "path": "/ComfyUI/models/loras/realistic_vision.safetensors",
            "strength": 1.0,
            "has_civitai_data": true
        },
        {
            "index": 1,
            "name": "CustomStyle",
            "info": "Local: CustomStyle (strength: 0.8)",
            "path": "/ComfyUI/models/loras/custom_style.safetensors",
            "strength": 0.8,
            "has_civitai_data": false
        }
    ],
    "count": 2,
    "combined_display": "Realistic Vision + CustomStyle"
}

No Valid LoRAs (Error Case)

{
    "loras": [],
    "count": 0,
    "combined_display": "No Valid LoRAs",
    "error": "No valid LoRAs found"
}

Exception Handling

{
    "loras": [],
    "count": 0,
    "combined_display": "Error Extracting LoRA",
    "error": "Specific error message"
}

Processing Logic

Multiple LoRA Processing

1. Input Validation: Check if input is WANVIDLORA list structure
2. LoRA Filtering: Skip entries with path: 'none' or empty paths
3. CivitAI Lookup: Query CivitAI API for each valid LoRA (if enabled)
4. Name Extraction: Use CivitAI name or extract from filename
5. JSON Construction: Build structured JSON with all metadata
6. Return: JSON string, human-readable info, and original passthrough

Single LoRA Processing (Compatibility)

• Maintains backward compatibility with non-list inputs
• Returns same JSON structure with single LoRA in array
• Handles both CivitAI and local extraction methods

Usage Examples

Parsing JSON Output in Python

import json

# From ComfyUI workflow
lora_json, lora_info, lora_passthrough = lora_extractor_node_output

# Parse JSON
lora_data = json.loads(lora_json)

# Access data
total_loras = lora_data["count"]
display_name = lora_data["combined_display"]

# Iterate through LoRAs
for lora in lora_data["loras"]:
    print(f"LoRA: {lora['name']} (strength: {lora['strength']})")
    if lora["has_civitai_data"]:
        print("  Has CivitAI metadata")
    else:
        print("  Local metadata only")

JavaScript Processing

// Parse JSON output
const loraData = JSON.parse(loraJsonOutput);

// Access combined display for UI
document.getElementById('lora-display').textContent = loraData.combined_display;

// Process individual LoRAs
loraData.loras.forEach((lora, index) => {
    console.log(`LoRA ${index + 1}: ${lora.name} at ${lora.strength} strength`);
});

Benefits of JSON Format

Structured Data Access

• Programmatic Processing: Easy to parse and process in any language
• Type Safety: Clear data types for each field
• Extensibility: Easy to add new fields without breaking existing code

Backward Compatibility

• Combined Display: combined_display field maintains the " + " separated format
• Info String: Second output still provides human-readable information
• Passthrough: Third output preserves original WANVIDLORA data

Enhanced Metadata

• Strength Values: Individual strength values for each LoRA
• Source Information: Clear indication of CivitAI vs local metadata
• File Paths: Full paths available for further processing
• Processing Status: Error handling and success indicators

Migration Guide

From String Format

Old Usage (string concatenation):

# Old format returned: "LoRA1 + LoRA2"
lora_names = old_output[0]
names_list = lora_names.split(" + ")

New Usage (JSON parsing):

# New format returns structured JSON
lora_json = json.loads(new_output[0])
names_list = [lora["name"] for lora in lora_json["loras"]]
combined_display = lora_json["combined_display"]  # Same as old format

Accessing Additional Data

With JSON format, you now have access to:

• Individual LoRA strength values
• Full file paths
• CivitAI metadata indicators
• Processing error information
• Structured arrays for iteration

Implementation Status

• ✅ JSON Structure: Complete with all metadata fields
• ✅ Multiple LoRA Support: Handles WANVIDLORA list structure
• ✅ Single LoRA Compatibility: Backward compatibility maintained
• ✅ Error Handling: Proper JSON error responses
• ✅ CivitAI Integration: Preserves CivitAI metadata detection
• ✅ Testing: Validated with mock data and real WANVIDLORA structures

Technical Notes

JSON Library

Uses Python's built-in json module for serialization:

• Consistency: Standard JSON formatting with 2-space indentation
• Reliability: Built-in error handling and validation
• Performance: Efficient serialization and parsing

Memory Considerations

• JSON strings are larger than simple concatenated strings
• Rich metadata provides more value than storage overhead
• Structured data enables better caching and processing optimization

Future Enhancements

Potential additions to JSON structure:

• Model Hash: File hash for verification
• File Size: LoRA file size information
• Last Modified: File timestamp data
• Tags: CivitAI tag information
• Ratings: CivitAI popularity/rating data