Prompt Injection in Translation

This app uses the local Gemma4 endpoint from gemma-main.py:

LM-Studio quantized models: https://lmstudio.ai/models/gemma-4

http://0.0.0.0:8000/chat/local_llm/

It accepts Spanish .txt, .md, .docx, and .pdf files, extracts their text, sends it to the model in chunks, and saves an English .txt translation under translations/.

Run

conda activate <your_env>
python gemma-main.py
python app.py

Running python app.py starts Streamlit on port 8550 and opens Edge browser at:

http://127.0.0.1:8550

Keep the Gemma4 service running before you translate. You can also edit the endpoint in the app if the service is exposed at a different host or port.