Aider-setup

Set up an Aider / Ollama conda environment in a container

Usage

Warning! This repository is only designed for use as a git submodule to add Aider - Ollama tools to containers. It is not intended for end-user applications. If you want to use Aider and Ollama as an end user, please consult the Aider documentation and the Ollama documentation.

Installation

1. ./1_reinstall_miniforge3.sh. This will remove any existing $HOME/miniforge3 installation and install a fresh copy.

2. ./2_recreate_conda_env.sh. This will create a conda virtual environment Aider containing

   • Python 3 and pip,
   • Ollama,
   • Aider, and
   • Shell scripts to manage the Ollama server.

This script will automatically detect whether an NVIDIA GPU is available and install the GPU-enabled ollama package if it is.

Operation

1. Enter the virtual environment: conda activate aider. On most Linux shells your shell prompt will change to remind you which environment you are in.

2. Start the Ollama server: start_ollama_server.sh. After the server starts, all of the pulled models will be listed.

3. Pull some models. I've whiteboard-interviewed a few well-known models and selected a few to test on your own applications. If you have at least 8 GB of RAM, run small_models.sh. This takes a fair amount of time but you only have to do it once. If you have at least 16 GB of RAM, also run large_models.sh. These scripts will also list the installed models.

4. If you forget which models you have, type list_models.sh. If there's a model you want to use that's not in my scripts, type

ollama pull <name of model>

In theory you don't need to do this; if you ollama run a model that hasn't been downloaded, ollama will download it, but it's better to download models up front rather than interrupt your workflow later.

5. You can now interact with the models directly. Type

ollama run <name of model>.

6. Start Aider: aider.

7. Stopping the Ollama server: stop_ollama_server.sh.