hello-agentcontrol-python

Hello LaunchDarkly for the Python Agent Control SDK.

Thin sample that wires every provider handler and asks LaunchDarkly which one to use at runtime. Observability (OpenTelemetry) is installed and auto-configured via init_client().

Prerequisites

• Python 3.12+
• uv

Clone

git clone https://github.com/launchdarkly/hello-agentcontrol-python.git
cd hello-agentcontrol-python

Setup

# Install deps and create a blank .env
uv run setup.py

# Or pass your LaunchDarkly keys and have setup write them into .env:
uv run setup.py <LD_SDK_KEY> <CONFIG_KEY>

Example:

uv run setup.py sdk-xxxxxxxxxxxxxxxxxxxxxxxx my-ai-config

This syncs dependencies and creates .env from .env.example if missing. When both arguments are provided, LD_SDK_KEY and CONFIG_KEY are inserted (or updated) in .env. Provide both keys, or neither.

Still fill in provider keys as needed:

• OPENAI_API_KEY and/or ANTHROPIC_API_KEY — whichever providers your config may select

Run

uv run main.py "What is feature flagging?"

LaunchDarkly chooses the provider and model from your AI config. Traces export to LaunchDarkly Observability automatically (no extra app code).