Application

Documentation for Parrot Libre AI IDE application.

Note

Parrot is in early development, not ready for end users.

Requirements

Parrot requires a current GNU/Linux system, such as Debian Stable.

Download

The current version of Parrot can be downloaded here:

Alternatively, releases are also available on this server:

Run

It is an AppImage, so it needs to be made executable, then run:

chmod +x Parrot.AppImage
./Parrot.AppImage

Status

At present, not all parts of Parrot are ready.

Main components:

  • IDE (text/code editor).

  • IDE AI extension.

  • AI extension server.

  • AI inference engine server.

  • AI model.

  • Search server.

What nominally works:

  • IDE (text/code editor).

  • IDE AI extension.

  • AI extension server.

What is missing:

  • AI inference engine server.

  • AI model.

  • Search server.

AI Server

Parrot is missing the built-in AI inference server. It can be made to work, by running your own AI server and model. That would require, in sum:

  • Find a model.

  • git clone https://github.com/ggerganov/llama.cpp

  • Build llama.cpp.

  • Run llama.cpp server.

Running the server like this, with GGUF model (see: TheBloke).

cd llama.cpp/build
./bin/server -c 4096 --host 127.0.0.1 -t 16 mymodel.gguf

Another example, using the first GPU:

export CUDA_VISIBLE_DEVICES=0

cd ~/devel/ggerganov/llama.cpp/build
./bin/server \
    --n-gpu-layers 1024 \
    -c 8192 \
    --host 0.0.0.0 \
    --port 8080 \
    -t 16 \
    -m /srv/ml/huggingface/TheBloke/Phind-CodeLlama-34B-v2-GGUF/phind-codellama-34b-v2.Q8_0.gguf

Note

Parrot documentation is written in English and uses AI machine translation for other languages.