NotLlamaFS

NotLlamaFS is an AI self-organizing file manager. It automatically renames and organizes your files based on their contents and well-known conventions (e.g., time) using Chat AI. Currently the support is made for text files and image files (if used model supports it).

It's made out of fun and in my opinion is useless.

Inspiration

I just wanted llama_fs to actually work. So I rewrote it from scratch.

Open your ~/Downloads directory. Or your Desktop. It's probably a mess...

Proclamation

It's not doing anything yet, just demonstrating the concept. NotLlamaFS is not a fork, but a rewrite from scratch!

Installation

Prerequisites

Before installing, ensure you have the following requirements:

• Python 3.10 or higher
• pip (Python package installer)

If you want to use NotLlamaFS with local llama models, you need to install Ollama and pull the llama3 and llava models like that:

ollama pull llama3 
ollama pull llava

If you want to use Groq, ChatGPT or Claude, you will need to get the API keys.

Installing

To install the project, follow these steps:

1. Clone the repository:

   git clone https://github.com/drforse/not-llama-fs.git

2. Navigate to the project directory:

   cd not-llama-fs

3. Install requirements

   pip install -r requirements.txt

   or with poetry

   poetry install 

Usage

Demo command

To see the demo of the resulting file structure, run the command:

For local llama3+llava (requires their installation! check prerequisites section for more details)

python -m app demo "path/to/directory/with/files/to/organize" --producer ollama 

For groq

python -m app demo "path/to/directory/with/files/to/organize" --producer groq --apikey "your-groq-api-key" 

For OpenAI (ChatGPT)

python -m app demo "path/to/directory/with/files/to/organize" --producer openai --apikey "your-openai-api-key"

For Claude

python -m app demo "path/to/directory/with/files/to/organize" --producer claude --apikey "your-claude-api-key" 

Real run command

To actually copy/move and organize files, run the command:

For local llama3+llava (requires their installation! check prerequisites section for more details)

python -m app create_fs "path/to/directory/with/files/to/organize" --dest-path "destination/path" --producer ollama 

And for other producers similarly, the same as in the demo command, but with create_fs instead of demo and with the destination path.

More settings for your run:

--text-model: model for text files
Defaults are:

• llama3 for local ollama models
• llama3-70b-8192 for groq
• gpt-4o for openai
• claude-3-haiku-20240307 for claude

--image-model: model for image files
Defaults are:

• llava for local ollama models
• claude-3-haiku-20240307 for claude
• text model is used for groq (setting ignored, code here is not completely alright)
• text model is used for openai (setting ignored, code here is not completely alright)

--move: by default create_fs copies files to new locations, but it can be changed with move flag, which makes it move the files insteadof copying

--pdf-as-images: by default producers translate pdf to text and analyze text unless there is no text found, but if pdf-as-images is specified than it analyzes it as images

Differences from LlamaFS

• NotLlamaFS is much better written, and has a better structure.
• NotLlamaFS works, and documentations is intended to fix not-understandable parts of the launch.
• NotLlamaFS is developed by a Windows's user, so it's more likely to work on Windows.
• NotLlamaFS not implementing an API, but instead it's a cli tool.
• NotLlamaFS supports not only local llama (moondream) and groq, but also ChatGPT and Claude.
• NotLlamaFS doesn't mix producers (moondream, groq, chatgpt, claude), but uses what it's told to use.
• NotLlamaFS won't have a frontend, but it's planned to have a GUI.
• NotLlamaFS doesn't try to say "its extremely fast because of using obvious cache for not reprocessing the same file" :D

What's next for NotLlamaFS

• Better configuration
• Updating file structure in system
• Better support for different models
• Better support for different file types
• Better documentation
• Improved prompts
• User-friendly interface (GUI)
• Custom rules for organizing files

Credits

https://github.com/iyaja/llama-fs - for the idea, inspiration and making me angry (for nothing working!) enough to write this.