neurobrix hub --category IMAGE
neurobrix hub --search whisper
Expected: a table with size and licence. Pick against your machine: sizes range widely, see the catalogue. Small upscalers and speech models run on CPU; large ones need a GPU.
neurobrix import <org>/<Model>
Expected: IMPORT COMPLETE, with the name to use next. The archive lands in ~/.neurobrix/store/ and is extracted to ~/.neurobrix/cache/<Model>. If the model is gated, see Licences. If you lose the connection, run the same command again: the download resumes.
Each kind has its own required input. When one is missing the engine says so rather than guessing.
# Language and code models: --prompt, optional --max-tokens, --temperature, --system
neurobrix run --model <Model> --prompt "Explain recursion in one sentence"
# Image: --prompt, optional --steps --cfg --seed --height --width; output .png
neurobrix run --model <Model> --prompt "A lighthouse at dusk" --output out.png
# Video: --prompt, optional --num-frames --fps; output .mp4
neurobrix run --model <Model> --prompt "Waves on a beach" --output out.mp4
# Text to speech: --prompt (the text), optional --reference-audio, --speaker; output .wav
neurobrix run --model <Model> --prompt "Hello" --output hello.wav
# Speech to text: --audio
neurobrix run --model <Model> --audio speech.wav
# Vision-language: --prompt and --input-image
neurobrix run --model <Model> --prompt "Describe this" --input-image photo.png
# Upscaler
neurobrix upscale --model <Model> --input in.png --output out.png
Audio-native models need --audio and --prompt. Multimodal models need --mode text|image|audio. The text output of speech-to-text without --output is not documented yet: pass --output explicitly if you need a file. Full flags: CLI reference.
A missing input gives, for example:
ZERO FALLBACK: family 'upscaler' requires --input-image. Provide it on the command line.
This is by design: see ZERO principles.
neurobrix serve --model <Model>
neurobrix run --prompt "..."
neurobrix stop
serve keeps the model loaded; run without --model uses it. Only one task runs at a time. See Serving.
neurobrix list
neurobrix info --models
neurobrix inspect <Model>
neurobrix validate <path-to-file.nbx>
neurobrix remove <Model>
Serving · Inference · Troubleshooting
© 2025–2026 Hocine Benkelaya. NeuroBrix engine released under the Apache License 2.0.
Developed by Hocine Benkelaya · Supported and funded by Neural Network Holding Ltd · Commercial distribution by WizWorks OÜ