https://github.com/cztomsik/dan

Transform stdin using language models.
https://github.com/cztomsik/dan

Last synced: 5 months ago
JSON representation

Transform stdin using language models.

• Host: GitHub
• URL: https://github.com/cztomsik/dan
• Owner: cztomsik
• Created: 2023-07-04T14:22:49.000Z (almost 2 years ago)
• Default Branch: main
• Last Pushed: 2023-10-02T15:40:26.000Z (over 1 year ago)
• Last Synced: 2023-10-02T21:06:45.019Z (over 1 year ago)
• Language: Zig
• Homepage:
• Size: 22.5 KB
• Stars: 6
• Watchers: 1
• Forks: 0
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
> This project is temporarily **on hold**.\

> I am now working on **[Ava, Personal Language Server](https://www.avapls.com/)**, a GUI app for running LLMs.

# DAN (do anything)

`dan` is a command-line interface tool that uses language models to perform transformations on standard input data.

https://github.com/cztomsik/dan/assets/3526922/1a915c33-438d-48c7-8250-b61dd78f44a4

## Usage

```

Usage: dan [options] [instruction]

Transform stdin using language models.

Examples:

  echo "Tell a joke." | dan

  cat hello.c | dan "Explain this code. Reason step by step."

Options:

  -h, --help               Display this help and exit.

  -v, --version            Display version info and exit.

  -m, --model         Model name or path.

  -p, --prompt        Prompt template.

  --top-k             Top-k sampling. (default: 40)

  --top-p             Top-p sampling. (default: 0.9)

  --temperature       Temperature sampling. (default: 0.8)

Notes:

  - If no instruction is given, dan will read from stdin.

  - Different models may work better with different prompt templates.

  - The configuration file is located at ~/.danrc

    see https://github.com/cztomsik/dan#configuration

```

## Installation

Follow these steps to install DAN:

- Download the binary file from the project's GitHub actions page: https://github.com/cztomsik/dan/actions

- Extract the binary file from the downloaded .zip file.

- Change the permissions of the binary file to make it executable: chmod +x dan

- Add the binary to your system PATH.

## Download models

We only support latest GGML formats. You can find some models here:

  - https://huggingface.co/TheBloke

  - https://www.reddit.com/r/LocalLLaMA/

## Configuration

**Optional:** You can create a `~/.danrc` with your custom defaults & preconfigured models:

```json

{

  "defaults": {

    "model": "name-of-the-default-model-or-path",

    "prompt": "USER: {instruction}{input}\nASSISTANT:",

  },

  "models": [

    {

      "name": "llama",

      "path": "/path/to/open-llama-xxx-ggml.bin"

    }

  ]

}

```

And then you can use the model by name (or even leave it in this case because we have default).

```bash

echo "Tell a joke." | dan -m ollama

```

## Build from source

If you want to build DAN from its source code, use the following commands:

```bash

# Clone the repository and its submodules

git clone https://github.com/cztomsik/dan

cd dan

git submodule update --init --recursive

# Build the project

zig build -Doptimize=ReleaseFast

# Run the program

./zig-out/bin/dan -h

```