Code

What this surface does

The Code surface documents the OpenAI-compatible integration path for Calypso RAG.

It exposes the grounded agent through standard request shapes such as:

• Chat Completions
• Responses

What makes this integration different

The request body stays intentionally thin. You call the hosted model by ID and let the server resolve the saved policy at runtime.

That means the important setup work happens before code:

• knowledge is configured
• the agent policy is saved
• Playground has validated the behavior

Readiness checklist

Before integrating from code, confirm:

• an active project API key exists
• the target agent is configured correctly
• the knowledge base is ready
• the chosen model ID is the one you actually want to call

Typical model choices

• calypso-rag-agent for the default workspace policy
• calypso-rag-agent:{profile_id} for a named profile

What the docs examples should teach

Examples should make these ideas explicit:

• where to set the base URL
• how to pass the project API key
• how to choose between Chat Completions and Responses
• how to swap between the default model and named profiles

Best practice

Treat Playground as your policy and grounding validator, and treat the Code surface as the production connection layer.

If the output is wrong in code but also wrong in Playground, the problem is usually policy or knowledge, not the SDK.