Provider/model settings are now explicit app config, not profile runtime patches.
Provider credentials and engine settings must now be wired explicitly:
StepSettings
gp.engines.fromConfig(...)
apiType
model
apiKey
baseURL
Profiles no longer carry provider credentials or engine-setting patches. They only contribute runtime metadata such as:
system_prompt
tools
middlewares
Example JS:
const gp = require("geppetto"); const engine = gp.engines.fromConfig({ apiType: "openai", model: "gpt-4.1-mini", apiKey: process.env.OPENAI_API_KEY, });
Agent documentation | LLMs.txt | Sitemap