Ollama Batch Processor 📚

A powerful PySide6-based GUI application for batch processing text files using Ollama LLM models. Supports translation, audiobook formatting, and intelligent text paraphrasing with customizable pipeline operations.

Features

• 🌐 Translation: Professional translation between languages with context preservation
• 🎧 Audiobook Formatting: Optimize text for text-to-speech systems
• ✍️ Paraphrasing: Improve flow, simplify language, remove idioms, adjust tone
• 📊 Pipeline Processing: Chain multiple operations in custom order
• ✂️ Smart Chunking: Intelligent text splitting with overlap and boundary detection
• 💾 Progressive Saving: Each pipeline step saved to separate files
• 🔄 Batch Processing: Process multiple files sequentially
• 🎯 Model Selection: Use different Ollama models per operation

Requirements

• Python 3.8+
• Ollama installed and running
• At least one Ollama model installed

Installation

1. Install Ollama

Download and install from ollama.ai

# Install a model
ollama pull mistral
# or
ollama pull llama3.2
ollama pull aya-expanse:32b

2. Install Python Dependencies

pip install PySide6 aiohttp qasync ollama

3. Run the Application

# Start Ollama server (in separate terminal)
ollama serve

# Run the application
python main.py

Configuration

Edit config.json to customize:

• Ollama host: Default http://localhost:11434
• Chunking presets: Adjust chunk sizes and overlap
• Operation settings: Modify prompts, icons, defaults
• UI settings: Window size, titles

Usage

Basic Workflow

1. Start Ollama: Run ollama serve in a terminal
2. Launch App: Run python main.py
3. Add Files: Click "📁 Add File" and select .txt files
4. Configure Pipeline:
   • Check operations to enable (Translation, Audiobook, Paraphrase)
   • Drag operations to reorder
   • Configure each operation's settings
5. Select Models: Choose Ollama model for each operation
6. Set Chunking: Select preset or enable "Process entire file"
7. Start Processing: Click "🚀 START"
8. Monitor Progress: Watch Activity Log and progress bar
9. Access Outputs: Find processed files and step files in output directory

Translation

• Set source and target languages
• Handles idioms intelligently
• Maintains consistency across chunks
• Temperature: 0.2-0.4 recommended

Audiobook Formatting

Enable options:

• Expand Contractions: "don't" → "do not"
• Spell Out Numbers: "123" → "one hundred twenty-three"
• Remove Special Characters: Clean non-standard symbols
• Add Reading Markers: Insert TTS-friendly markers

Paraphrasing

Enable sub-operations:

• Improve Flow: Better sentence structure and transitions
• Simplify Language: Make complex text accessible
• Remove Idioms: Convert figurative to literal language
• Adjust Tone: Formal, casual, professional, or conversational

Chunking Settings

Presets:

• Fast (2000/150): Quick processing
• Balanced (2500/200): Default, good quality
• High Context (3000/250): Better continuity
• Large (4000/300): Fewer API calls
• Extra Large (6000/400): Maximum context

Process Entire File: Disable chunking for small files (< 2500 chars)

Output Files

The application creates multiple output files:

input.txt                          # Original file
input_step_01_translated.txt       # After translation
input_step_02_audiobook.txt        # After audiobook formatting
input_processed.txt                # Final output

Troubleshooting

"Nothing Happens" When Clicking Start

Check:

1. Is Ollama running? → ollama serve
2. Is a model installed? → ollama list
3. Did you add input files?
4. Is at least one operation checked?
5. Is a model selected in dropdown?

Connection Errors

# Test Ollama
ollama list

# If empty, install a model
ollama pull mistral

# Test inference
ollama run mistral "hello"

Model Not Loading

Large models (30B+) take 1-2 minutes to load on first use. Watch console output for [DEBUG] messages.

Performance Tips

• Smaller chunks: Faster processing, less context
• Larger chunks: Slower but better quality
• Combined operations: More efficient than separate runs
• Fast models: Use smaller quantized models for speed
• GPU: Ensure Ollama uses GPU for better performance

Advanced Configuration

Custom Prompts

Edit config.json operation prompts:

{
  "operations": {
    "translation": {
      "prompts": {
        "system_first": "Your custom system prompt...",
        "user_first": "Your custom user prompt..."
      }
    }
  }
}

Temperature Settings

• Translation: 0.2-0.4 (deterministic)
• Paraphrasing: 0.4-0.6 (creative)
• Creative Writing: 0.7-1.0 (very creative)

Pipeline Order

Operations execute in order from top to bottom. Typical workflows:

1. Translation → Audiobook: Translate then optimize for TTS
2. Paraphrase → Simplify → Remove Idioms: Multi-step text cleanup
3. Translation → Paraphrase (tone): Translate and adjust formality

Keyboard Shortcuts

• Ctrl+O: Add files
• Ctrl+S: Select output directory
• Ctrl+R: Start processing
• Esc: Stop processing

Diagnostic Tools

Test Ollama Connection

python test_ollama.py

Shows detailed connection diagnostics and model availability.

Debug Mode

Console output shows [DEBUG] messages for:

• Model calls with parameters
• Response types and content length
• Error tracebacks
• Pipeline execution flow

Known Limitations

• Only .txt files supported
• No real-time progress within chunks (model inference time varies)
• Large models require significant RAM/VRAM
• Async operations prevent UI responsiveness during heavy processing

Tips for Best Results

1. Pre-process text: Remove excessive whitespace, fix encoding
2. Use appropriate models: Match model size to task complexity
3. Test with small files first: Verify settings before batch processing
4. Monitor first chunk: Shows model loading time and quality
5. Enable deduplication: Removes duplicate paragraphs at chunk boundaries
6. Save intermediate steps: Useful for debugging and iterative refinement

Architecture

• main.py: PySide6 GUI and application logic
• Translator.py: Ollama API interface and text processing
• config.json: Configuration and prompts
• qasync: Async event loop integration with Qt

License

MIT License - feel free to modify and distribute.

Contributing

Improvements welcome! Focus areas:

• Additional operation types
• Better error recovery
• Real-time streaming output
• Support for other file formats
• Memory optimization for large files

Credits

Built with:

• Ollama - Local LLM inference
• PySide6 - Qt GUI framework
• qasync - Qt async support

---

Version: 1.0
Last Updated: October 2025