AI assistant

Dreeve ships with an AI-powered workout assistant. To use it, you configure an AI provider under Settings → Integrations.

Caution

Use caution when enabling this feature if your app is publicly accessible.

Important

Dreeve uses the Neuron AI library to interface with AI models. Only providers supported by Neuron AI are compatible. See the full list of supported providers

Configuring an AI provider

Cloud providers

To use a cloud-based AI provider you need an API key and the name of the model you want. Configure the key in the AI_API_KEY environment variable, then pick the provider and enter the model name in the settings.

Locally hosted Ollama

You can also run a local model using Ollama. Start by configuring a Docker container for Ollama:

docker-compose.yml

services:
    ollama:
        image: ollama/ollama:latest
        container_name: 'dreeve-ollama'
        tty: true
        restart: unless-stopped
        volumes:
            - .:/code
            - ./ollama:/root/.ollama
        environment:
            - OLLAMA_KEEP_ALIVE=24h
            - OLLAMA_HOST=0.0.0.0
        ports:
            - '11434:11434'
        networks:
            - dreeve-network

Tip

Looking for more advanced setups? Check out this Ollama Docker GitHub repository.

Next, download the model you want to use. For example, to run llama3.2:

> docker compose exec ollama ollama pull llama3.2

Finally, enable the integration, choose the ollama provider, and set:

• Model: the model you pulled, e.g. llama3.2.
• URL: the Ollama API root, e.g. http://dreeve-ollama:11434/api. Use the container name when Ollama runs in the same Docker network as Dreeve. The URL must include the scheme and end with /api.

Important

Make sure you’re running the latest version of Ollama. Streaming responses with tooling has been added on May 28, 2025.

OpenRouter

OpenRouter provides access to hundreds of models (OpenAI, Anthropic, Google and more) through a single API key, including free-tier models. This makes it easy to get started or test different models.

To use OpenRouter, pick the openAILike provider under Settings → Integrations and set:

• Base URI: https://openrouter.ai/api/v1
• Model: e.g. anthropic/claude-sonnet-4.5. See openrouter.ai/models.

and put your OpenRouter API key in AI_API_KEY in your .env.

Tip

OpenRouter offers free models which are great for testing without any cost.

Your AI workout assistant

Via the UI

Via the CLI

> docker compose exec app bin/console app:ai:agent-chat

This will prompt you with a message like the following:

Pre-defining chat commands

The app allows you to pre-define chat commands that can be used by the AI assistant. This is useful for questions that might be asked frequently, such as

Please analyze my most recent ride with regard to aspects such as heart rate, power (if available). Please give me an assessment of my performance level and possible improvements for future training sessions?

Each command has a name and the message it expands to. For example:

Command	Message
analyse-last-workout	You are my bike trainer. Please analyze my most recent ride with regard to aspects such as heart rate, power (if available). Please give me an assessment of my performance level and possible improvements for future training sessions.
compare-last-two-weeks	You are my bike trainer. Please compare my workouts and performance of the last 7 days with the 7 days before and give a short assessment.

You can then use /analyse-last-workout and /compare-last-two-weeks in your chat with the AI assistant: