AI SDKs
This category contains documentation topics explaining the LaunchDarkly AI SDKs.
The LaunchDarkly AI SDKs are designed for use with AgentControl. They are in a pre-1.0 release and under active development.
Python AI SDK is in open beta
The current Python and Node.js AI SDKs are higher-level, opinionated SDKs that reduce the setup required to call a model, run judges, and orchestrate agent graphs. They are in open beta. To get started, read the Python AI SDK reference or Node.js AI SDK reference.
We recommend using these SDKs for new applications. They are not drop-in replacements for the Python AI SDK (Legacy) or Node.js AI SDK (Legacy), which are now in maintenance mode. If you are updating an existing integration, read Migrate from the legacy AI SDKs.
About AI SDKs
While LaunchDarkly server-side, client-side, and edge SDKs focus on interacting with feature flags, the LaunchDarkly AI SDKs interact with AgentControl configs. Configs are resources that you create in LaunchDarkly and use to customize, test, and roll out new large language models (LLMs) within your generative AI applications.
You can use the AI SDKs to customize each config based on the context that you provide. This means both the message and the model evaluation are specific to each end user, at runtime. You can also use the AI SDKs to record metrics from your AI model generation, including duration and tokens.
AI metrics and traces are recorded differently. AI metrics are captured through the AI SDK tracker methods. If you use the LaunchDarkly observability plugin to capture traces, run your model calls inside an active span so LaunchDarkly can associate traces with the evaluated config.
Additional resources
You can follow development or contribute on GitHub with the following links.