launchdarkly-ai-openai-agents

OpenAI handler for launchdarkly-ai-server using the OpenAI Agents SDK (openai-agents). Delegates the full agentic loop — tool calls, retries, and orchestration — to the Agents SDK.

provides_for: ['OpenAI', 'agent'] — matches flag variations where provider.name is "OpenAI" and meta.mode is "agent".

Installation

pip install launchdarkly-ai-server launchdarkly-ai-openai-agents

Set OPENAI_API_KEY in your environment (the OpenAI SDK reads it automatically).

Usage

With config()

import asyncio
from launchdarkly_ai_server import config, shutdown
from launchdarkly_ai_openai_agents import create_openai_agent_handler

async def main():
    result = await config(
        key="my-ai-config-flag",
        handler=create_openai_agent_handler(),
        tool_handlers={"search": lambda q: "..."},
    ).invoke("Summarize today's changelog", {"kind": "user", "key": "user-123"})

    print(result.response)
    await shutdown()

asyncio.run(main())

Tool handlers may be sync or async. Sync handlers run on the event-loop thread, so blocking I/O stalls the agent. Keep graph __handoff_* handlers synchronous: routing records the selected edge on the call itself, and moving them onto asyncio.to_thread would break that.

Convenience wrapper

import asyncio
from launchdarkly_ai_openai_agents import openai_agents

async def main():
    user_input = "Summarize today's changelog"
    result = await openai_agents(
        user_input,
        {"kind": "user", "key": "user-123"},
        {"key": "my-ai-config-flag"},
        variables={"user_input": user_input},
    )
    print(result.response)

asyncio.run(main())

Agent graphs — openai_graph()

Runs a LaunchDarkly agent graph with the OpenAI agent handler pre-bound. Equivalent to calling the base graph() with handlers=[create_openai_agent_handler()]. See the core client docs for the full graph() API.

import asyncio
from launchdarkly_ai_openai_agents import openai_graph

async def main():
    result = await openai_graph("support-graph").invoke(
        "I was double charged",
        {"kind": "user", "key": "user-123"},
    )
    print(result["response"])

asyncio.run(main())

Native graph adapter — to_openai_agents()

Converts a resolve_graph() result into a framework-native OpenAI Agents swarm (post-order traversal: leaves → root). Children are wired as handoffs; the root is run with Runner.run.

import asyncio
from launchdarkly_ai_server import resolve_graph
from launchdarkly_ai_openai_agents import to_openai_agents

async def main():
    ctx = {"kind": "user", "key": "user-123"}
    result = await to_openai_agents(
        resolve_graph("support-graph", context=ctx),
        {"tool_handlers": registry.tools, "context": ctx},
    ).call("I was double charged")
    print(result["response"])

asyncio.run(main())

How It Works

• Uses the system prompt and tools defined in your LaunchDarkly flag config.
• Template placeholders ({{variable}}) in the prompt are substituted using variables before the call.
• The OpenAI Agents SDK manages the full agentic loop — tool dispatch, re-prompting, and termination — automatically.
• Emits an OTel span and LaunchDarkly telemetry for every call.

Choosing Between openai-agents and openai-messages

Environment Variables