Getting started¶

1. Install¶

The quickest way is a ready-made build from the releases page (see download steps).

To build it yourself, follow the install guide for Windows or Linux. In short:

cargo build --release                    # CLI + web interface
cargo build --release --features trace   # also lets you run models (needs more RAM)

The program is target/release/neurondump (neurondump.exe on Windows). The examples below assume it is on your PATH, or that you created an alias:

alias neurondump=./target/release/neurondump

2. Find your models¶

neurondump list

NeuronDump looks in these places automatically:

Source	Where
Ollama	`OLLAMA_MODELS`, or `~/.ollama/models` (`%USERPROFILE%\.ollama\models` on Windows)
llama.cpp	`LLAMA_CACHE`, or `~/.cache/llama.cpp`
Hugging Face	`HF_HUB_CACHE`, `HF_HOME/hub`, or `~/.cache/huggingface/hub`
LM Studio	`~/.lmstudio/models`

For other folders, add --scan <dir> or save them in the config:

neurondump --scan ~/my-models list
neurondump config set scan_dirs "C:/models, D:/more-models"

3. Name a model¶

All of these work:

neurondump inspect llama3.2:1b                      # Ollama name (tag defaults to "latest")
neurondump inspect someuser/somemodel:q8            # namespaced Ollama model
neurondump inspect ./downloads/model.gguf           # any GGUF file
neurondump inspect 17d2889ccc34                     # the id shown by `list`

If you mistype a name, NeuronDump suggests the closest match.

4. First look inside¶

neurondump inspect llama3.2:1b       # summary: architecture, layers, tokenizer, templates
neurondump layers llama3.2:1b        # one line per layer
neurondump tensors llama3.2:1b --layer 0
neurondump vocab llama3.2:1b --class code --limit 20
neurondump template llama3.2:1b
neurondump stats llama3.2:1b         # reads all weights, takes a few seconds

5. Choose an interface¶

neurondump            # asks: CLI (interactive) or Web interface, and can remember your choice
neurondump cli        # interactive prompt with tab completion
neurondump web --open # local web page

Where files go¶

What	Where
Settings	`~/.config/neurondump/config.toml` (`%APPDATA%\neurondump\config.toml` on Windows)
Cached results	`~/neurondump/cache/`
Slimmed models	`~/neurondump/slim/`

Change the output folder with neurondump config set output_dir <dir>.

Next: How it works · CLI reference