Skip to main content

​
Deploy from CLI

​
Launch an Agent Using the Base Runtime

To deploy an agent using the base runtime from the command line, you can use the following command: 
lucid launch --runtime base --model gpt-4o --prompt "You are a helpful agent" --target docker
This command launches a pre-built Docker image of the agent runtime, specifically ghcr.io/lucid-fdn/agent-runtime:v1.0.0. This image is designed to work with any OpenAI-compatible endpoint for inference, which you can specify using the PROVIDER_URL environment variable.

​
Configuration

The deployment is configured through several environment variables:
  • LUCID_MODEL: Specifies the model to be used.
  • LUCID_PROMPT: Defines the prompt for the agent.
  • LUCID_TOOLS: Lists any additional tools or configurations.

​
Inference and Receipts

  • Inference: Handled via the PROVIDER_URL, allowing you to connect to any compatible endpoint.
  • Receipts: Managed through the LUCID_API_URL, which is decoupled from the inference process to ensure flexibility and modularity.

​
Verification

The deployment process includes full verification to ensure that the agent is functioning as expected. This ensures reliability and consistency in the agent’s performance.