AI Content Pipeline Skill

This skill helps you work with the AI Content Pipeline - a unified Python package for multi-modal AI content generation.

Quick Reference Commands

Pipeline Execution

# Activate virtual environment first
source venv/bin/activate  # Linux/Mac
venv\Scripts\activate     # Windows

# Run a pipeline from YAML config
ai-content-pipeline run-chain --config input/pipelines/config.yaml

# Run with parallel execution (2-3x faster)
PIPELINE_PARALLEL_ENABLED=true ai-content-pipeline run-chain --config config.yaml

# Use short alias
aicp run-chain --config config.yaml

Single Operations

# Generate a single image
ai-content-pipeline generate-image --text "A beautiful sunset" --model flux_dev

# Create video from text (text -> image -> video)
ai-content-pipeline create-video --text "A beautiful sunset"

# List all available models
ai-content-pipeline list-models

Testing

# Quick smoke tests (30 seconds)
python tests/test_core.py

# Full integration tests (2-3 minutes)
python tests/test_integration.py

# Run all tests
python tests/run_all_tests.py

# Quick test mode
python tests/run_all_tests.py --quick

Available AI Models (28 Total)

Text-to-Image (4 models)

Image-to-Video (4 models)

Image-to-Image (6 models)

• Photon Flash, Photon Base, FLUX variants, Clarity Upscaler

Image Understanding (7 models)

• Gemini variants for description, classification, OCR, Q&A

Prompt Generation (5 models)

• OpenRouter models for video prompt optimization

YAML Pipeline Configuration

Create a pipeline config file in input/pipelines/:

name: "My Content Pipeline"
description: "Generate image and convert to video"

steps:
  - name: "generate_image"
    type: "text-to-image"
    model: "flux_dev"
    params:
      prompt: "A majestic mountain landscape at sunset"
      width: 1920
      height: 1080

  - name: "create_video"
    type: "image-to-video"
    model: "veo_3"
    params:
      image: "{{step_1.output}}"
      prompt: "Camera slowly pans across the landscape"
      duration: 5

Parameter Templating

Use {{step_N.output}} to reference outputs from previous steps.

Cost Estimation

Typical costs per operation:

• Text-to-Image: $0.001-0.004 per image
• Image-to-Image: $0.01-0.05 per modification
• Image-to-Video: $0.08-6.00 per video (model dependent)

Always estimate before large pipelines:

ai-content-pipeline estimate --config config.yaml

Environment Setup

Required environment variables in .env:

FAL_KEY=your_fal_api_key
PROJECT_ID=your-gcp-project-id
OUTPUT_BUCKET_PATH=gs://your-bucket/output/
ELEVENLABS_API_KEY=your_elevenlabs_key
OPENROUTER_API_KEY=your_openrouter_key
GEMINI_API_KEY=your_gemini_key

Project Structure

ai-content-pipeline/
├── packages/
│   ├── core/ai_content_pipeline/    # Main pipeline
│   ├── providers/                   # Google Veo, FAL AI
│   └── services/                    # TTS, video tools
├── input/                           # Pipeline configs
├── output/                          # Generated content
└── tests/                           # Test suites

Common Tasks

Creating a New Pipeline

1. Create YAML config in input/pipelines/
2. Define steps with model and parameters
3. Use {{step_N.output}} for chaining
4. Run with aicp run-chain --config your_config.yaml

Adding a New Model

1. Check packages/providers/ for the provider
2. Implement model interface
3. Register in model registry
4. Add to list-models output

Debugging Pipeline Issues

1. Run with verbose logging
2. Check output/ for intermediate files
3. Verify API keys in .env
4. Test individual steps manually