instappt 0.0.2

Add your description here

InstaPPT

InstaPPT is an AI-powered PowerPoint translation tool designed to preserve the original visual fidelity of your slides while providing high-quality translations.

Demo

Here is a visual comparison of the original and translated slides:

View full comparison PDF

Features

• AI Translation: Uses advanced LLMs (like GPT-4, DeepSeek) for accurate, context-aware translation.
• Visual Fidelity: Preserves fonts, colors, sizes, and layout.
• Visual Comparison Report: Generates a side-by-side PDF report showing the original and translated slides as images, ensuring "what you see is what you get".
• Smart Caching: Caches translation results to save costs and time on repeated runs.
• Cross-Platform: Works on macOS (optimized) and Windows.

Architecture

graph TD
    User([User]) -->|Input| Interface
    
    subgraph Interface ["User Interface"]
        CLI["Command Line"]
        UI["Web UI (Gradio)"]
    end
    
    Interface --> Controller["Core Controller"]
    
    subgraph Pipeline ["Processing Pipeline"]
        Parser["PPTX Parser"] -->|Extract Text| Segments["Text Segments"]
        
        subgraph Translation ["LLM Translation Loop"]
            Step1[Translation] --> Step2[Evaluation]
            Step2 -- "Low Score" --> Step3[Optimization]
            Step2 -- "High Score" --> Ready[Ready]
            Step3 --> Ready
        end
        
        Segments -->|Parallel Processing| Translation
        Translation <-->|API Calls| LLM[("LLM Provider")]
        
        Ready --> Assembler["PPTX Assembler"]
        Assembler -->|Generate| Artifacts
    end
    
    Controller --> Pipeline
    
    subgraph Artifacts ["Output Files"]
        File1["Translated PPTX"]
        File2["Comparison PDF"]
        File3["Evaluation Report"]
    end

Prerequisites

1. Python 3.10+
2. LibreOffice: Required for converting PPTX to PDF.
   • macOS: brew install --cask libreoffice
   • Windows: Install from official website.
3. Poppler: Required for the image-based visual comparison.
   • macOS: brew install poppler
   • Windows: Download binary and add to PATH.

Installation

Install via Pip

uv pip install instappt

Development Setup

# Create virtual environment
uv venv

# Activate virtual environment
source .venv/bin/activate

# Install dependencies
uv sync

Usage

CLI Usage (After Installation)

If you installed the package via pip, you can use the instappt command directly:

instappt --input input.pptx --output output_dir --lang English --config model_info.json

Development Usage

If you are running from source:

python main.py --input input.pptx --output output_dir --lang English

Using a Configuration File

Create a model_info.json file to configure your LLM settings:

{
    "translator": {
        "model": "gpt-4o",
        "api_key": "sk-...",
        "base_url": "https://api.openai.com/v1"
    }
}

Run with config:

python main.py --input input.pptx --output output_dir --lang English --config model_info.json

Force Re-translation (Disable Cache)

If you want to ignore the cache and force a fresh translation:

python main.py --input input.pptx --output output_dir --lang English --config model_info.json --no-cache

SDK Usage

You can also use InstaPPT as a Python library in your own projects.

from instappt.core import PPTTranslator
from instappt.models import SDKConfig, ModelConfig

# 1. Configure Models
config = SDKConfig(
    translator_config=ModelConfig(
        model="gpt-4o",
        api_key="sk-...",
        base_url="https://api.openai.com/v1"
    ),
    optimizer_config=ModelConfig(
        model="gpt-4o",
        api_key="sk-...",
        base_url="https://api.openai.com/v1"
    ),
    evaluator_config=ModelConfig(
        model="gpt-4o",
        api_key="sk-...",
        base_url="https://api.openai.com/v1"
    ),
    enable_cache=True # Set to False to disable caching
)

# 2. Initialize Translator
translator = PPTTranslator(config, concurrency=32)

# 3. Process PPTX
input_file = "presentation.pptx"
output_file = "presentation_translated.pptx"
target_language = "English"

translator.process_ppt(input_file, output_file, target_language)

# 4. Generate Reports (Optional)
translator.generate_reports(
    output_dir="output_folder",
    report_prefix="my_report_",
    original_pptx=input_file,
    translated_pptx=output_file
)

Output

The tool generates:

1. Translated PPTX: [filename]_translated.pptx
2. Visual Comparison: [filename]_comparison.pdf (Side-by-side images)
3. Assessment Report: [filename]_assessment_report.pdf (Quality metrics)

License

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). See the LICENSE file for details.

All files are saved in the specified output directory.