LLM PDF Translator

LLM PDF Translator summarizes and translates PDF documents using local or hosted large language models.

Features

• Summarize a document to provide translation context.
• Translate PDFs while preserving the original layout.
• Configure different LLM endpoints via config.yaml.

Configuration

Edit config.yaml to choose the model and base URL for the desired endpoint.

Ollama

base_url: http://ollama-gpu:11434
model: "gemma3:1b"

The base_url must use the container name of your Ollama server (e.g., ollama-gpu) when running on the ollama-network Docker network.

OpenRouter

base_url: "https://openrouter.ai/api/v1"
model: "google/gemma-3-27b-it:free"

OpenAI

base_url: "https://api.openai.com/v1"
model: "gpt-4o-mini"

When using OpenRouter or OpenAI, update core/translate.py to initialize OpenAIClient instead of OllamaClient so that the correct API client is used.

Environment Variables

Create a .env file with the necessary API keys:

OPENAI_API_KEY=your-openai-key
OPENROUTER_API_KEY=your-openrouter-key

Only the key for the chosen endpoint is required.

Build and Run

Build the Docker image:

docker build -t llm-pdf-translator .

Run the container, mapping the Streamlit port and connecting to the Ollama network:

docker run --rm -it --network=ollama-network -p 8765:8501 llm-pdf-translator

The app will be available at http://localhost:8765.