`OpenAIRealtimeAgentAdapter`

Connects directly to wss://api.openai.com/v1/realtime. Use this when the OpenAI Realtime model itself is the agent under test, or as a voice-enabled user simulator.

Constructor

python

import scenario
from scenario.config.voice_models import OPENAI_REALTIME_MODEL
from scenario.types import AgentRole
 
adapter = scenario.OpenAIRealtimeAgentAdapter(
    model=OPENAI_REALTIME_MODEL,
    voice="alloy",
    instructions="You are a helpful assistant. Keep responses brief.",
    role=AgentRole.AGENT,
)

api_key / apiKey is read from OPENAI_API_KEY by default; pass it explicitly to override. tools accepts function-calling schemas.

To use the OpenAI Realtime model as a voice user simulator instead of the agent under test, set role to the user role (role=AgentRole.USER / role: AgentRole.USER).

Capabilities

streaming_transcripts	native_vad	dtmf	interruption	input_formats	output_formats
✅	✅	❌	✅	`pcm16/24000`	`pcm16/24000`

Worked examples

Python: openai_realtime_agent.py — agent under test with a judge step; openai_realtime_user.py — Realtime as a voice user simulator; getting_started.py — minimal scenario from the getting-started page.
TypeScript: openai-realtime-agent.test.ts — agent under test; openai-realtime-user.test.ts — Realtime as a voice user simulator.