AI Configuration¶

Configure local AI processing with Ollama.

Model Selection¶

Available Models¶

Model	Size	RAM Required	Best For
`tinyllama`	637 MB	2 GB	Quick tests
`phi3`	2.2 GB	4 GB	Light usage
`llama3.2:3b`	2.0 GB	4 GB	Balanced
`llama3.2:7b`	4.7 GB	8 GB	Quality
`mistral`	4.1 GB	8 GB	General purpose

Setting the Model¶

OLLAMA_MODEL=llama3.2:3b

Download Models Manually¶

docker exec ollama ollama pull llama3.2:3b

GPU Acceleration¶

NVIDIA GPUs¶

Ensure NVIDIA Container Toolkit is installed:

# Install toolkit
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt update && sudo apt install -y nvidia-container-toolkit
sudo systemctl restart docker

Verify GPU Access¶

docker exec ollama nvidia-smi

Disabling AI¶

For resource-constrained environments:

# Via flag
./start.sh --no-ai

# Via environment
DISABLE_ALL_AI_FUNCTIONALITY=true

Custom Prompts¶

AI perspectives are generated using configurable prompts. Example prompts are provided out of the box, and you can add your own using either of the following approaches.

Option 1: `prompts.json` (legacy, still functional)¶

Create a prompts.json file in the prompts folder at the root of the application (see prompts_example.json in the same folder for the expected structure). This folder is mounted as a Docker volume, so changes are picked up without rebuilding images.

Option 2: Prompts API (recommended)¶

Using the TruSpace API, you can create, read, update, or delete prompts — persisted in the SQLite database rather than a static file. The frontend for managing prompts through the UI is not yet complete, but you can drive the API directly with curl.

See scripts/prompts/examples.sh for a full worked example, including the login flow. The Prompts API is protected — only registered users can view and edit prompts, and the example script illustrates the required login/cookie handling.