AI Task System Quick Reference

This document provides a quick reference for common AI Task system commands and workflows.

Environment Setup

Always activate the conda environment before using AI Task:

conda activate lmu

Set your API key:

export GEMINI_API_KEY="your-api-key-here"

Common Commands

AI Partitur Command

Run a partitur workflow for a specific profile:

ai-partitur <partitur_name> <profile_id> [--overwrite]

Examples:

# Run the inqua_full partitur for profile 32101
ai-partitur inqua_full 32101

# Force regeneration of outputs
ai-partitur inqua_full 32101 --overwrite

# Use a specific partitur file path
ai-partitur --file /path/to/custom.yml 32101

# Check version
ai-partitur --version

YouTube Processing

Download and transcribe a YouTube video:

./transcribe_youtube.sh <production_id> <instruction_file>

Example:

./transcribe_youtube.sh ai_reporter_nepi_00002 ../ai-reporter-nepi/production/instructions/youtube_transcription.ai

Local Video Processing

Transcribe a local video file:

python transcribe_video.py --video <video_path> --output <output_path> --template <template_path>

Example:

python transcribe_video.py --video ./video/sample.mp4 --output ./transcription/sample.txt --template ./templates/transcription.j2

Directory Structure Shortcuts

Creating Required Directories

For a new profile with ID 32101:

mkdir -p profile/profile_32101/{audio,transcription,sequence}

Moving Files to Standard Locations

Copy audio file to profile directory:

cp /path/to/audio.m4a profile/profile_32101/audio/document_32101.m4a

Standard File Naming Patterns

Audio files: document_<id>.m4a
Transcriptions: <prefix>_<id>_transcription_<version>.txt
Analysis files: <prefix>_<id>_<analysis_type>_<version>.txt
Document files: <prefix>_<id>_<document_type>_<version>.docx

Where: - <prefix>: Project identifier (e.g., INQUA2) - <id>: Profile identifier (e.g., 32101) - <version>: Version number (typically 01)

Common Workflows

Complete Audio Processing Workflow

Create profile directories:

mkdir -p profile/profile_<id>/{audio,transcription,sequence}

Place audio file in profile directory:

cp /path/to/audio.m4a profile/profile_<id>/audio/document_<id>.m4a

Run partitur to process the audio:
```
ai-partitur inqua_full <id>
```

Creating a Custom Partitur

Create a YAML file in your partitur directory:
```
nano partitur/custom_workflow.yml
```
Define the partitur structure (see template below)
Run the partitur:
```
ai-partitur custom_workflow <id>
```

Partitur Template

Basic partitur template:

name: custom_workflow
description: "Custom workflow for audio processing"
template_dir: "instruction"

pipe:
  # Transcribe audio
  - name: transcribe_audio
    type: llm
    model: gemini-1.5-pro
    tmpl: "transcription_template.j2"
    source-file: "profile/profile_{{id}}/audio/document_{{id}}.m4a"
    result-file: "profile/profile_{{id}}/transcription/output_{{id}}_transcription_01.txt"
    overwrite: true

  # Additional steps as needed
  # ...

Troubleshooting Checklist

✓ Conda environment activated (conda activate lmu)
✓ API key set (export GEMINI_API_KEY="...")
✓ Directories exist and have correct permissions
✓ File naming follows standard patterns
✓ Partitur file is correctly formatted YAML
✓ Templates exist in the specified template directory
✓ Input files exist at the specified paths