This release has been tested on Windows 11 64-bit hardware. The software does work if you run it on a machine with only a CPU; try it out with a few images or audio files. GPU strongly recommended.
Notes for Developers Only:
**Pre-requisites for windows/macos/Linux **
-
python 3.13 , poetry checkout git repo src and run poetry install
-
install ollama and pull models ibm/granite4.1:3b, moondream:latest, gemma3:4b , gemma3:1b see "startup\ollama_check.sh"
-
install and run pgvector docker container on macos/Linux/windows
"startup\docker-compose.yml"
"startup\pgvector_start.sh" "startup\check_pgvector_db.sh" -
download and extract onnx models from : https://umass-my.sharepoint.com/:u:/g/personal/jaikumar_umass_edu/IQAiVKIZKA8QT6EEVDLqmc5NAWuYAy8wl9I3K2Ken5G7lLc?e=w8DjIl
extract rb_3.1_onnx_models.zip to rescuebox <ROOT> or top level directory ( where "src" folder exists) -
install ffmpeg.exe for audio transcribe https://www.osxexperts.net/ffmpeg81arm.zip
-
start backend server poetry run python -m rb.api.main , confirm by accessing http://localhost:8000
-
start frontend UI server poetry run python frontend/main.py, confirm by accessing http://localhost:8080
NOTES:
** for macos** poetry toml file are upto date.
Frontend (NiceGUI) developer docs: frontend/docs/README.md — workflow, database, tests, and related topics.
src-tauri folder to build windows rescuebox installer. src-tauri\nsis\README.txt for current build notes.
image-embeddings converted to onnx model , text embeddings onnx convert pending..
on windows : frontend.spec and backend.spec run with pyinstaller to create exe and package with src-tauri build steps.
on linux : run_backend_server and run_ui_server was used to demo nvidia/cuda gpu with dgx-spark server
See the LICENSE file for license details.