.______  .______ .______  .___ 
: __   \ \____  |:      \ : __|
|  \____|/  ____||   .   || : |
|   :  \ \      ||   :   ||   |
|   |___\ \__:__||___|   ||   |
|___|        :       |___||___|
             •                 

Run r2ai in local, without internet or leaking any data. Use your CPU/GPU/NPU and interact with r2 using natural language.

The current implementation is based on llama-cpp and the default model is CodeLlama-CherryPop

--pancake

Installation

pip3 install rich inquirer llama-cpp tokentrim
r2pm -i r2ai

Execution

There are 4 different ways to run r2ai:

• Standalone and interactive: r2pm -r r2ai
• Batch mode: `r2ai '-r act as a calculator' '3+3=?'
• From radare2 (requires r2pm -ci rlang-python): r2 -c 'r2ai -h'
• Using r2pipe: #!pipe python main.py

Scripting

You can interact with r2ai from standalone python, from r2pipe via r2 keeping a global state or using the javascript intrepreter embedded inside radare2.

• examples/conversation.r2.js - load two models and make them talk to each other

Development/Testing

Just run make .. or well python main.py /path/to/file

It's also possible to install it with conda, which is the recommended way on Macs.

• https://developer.apple.com/metal/pytorch/

curl -O https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh
sh Miniconda3-latest-MacOSX-arm64.sh

conda install pytorch torchvision torchaudio -c pytorch-nightly
conda run pip install inquirer rich appdirs huggingface_hub tokentrim llama-cpp-python

Features

• Prompt the language model without internet requirements
• Slurp file contents and make actions on that
• Embed the output of an r2 command and ask the LM to resolve questions
• Define different system-level assistant role
• Set environment variables to provide context to the language model
• Live with repl and batch mode from cli or r2 prompt
• Accessible as an r2lang-python plugin, keeps session state inside radare2
• Scriptable from bash, r2pipe, and javascript (r2papi)
• Use different models, dynamically adjust query template
  • Load multiple models and make them talk between them

TODO

• add "undo" command