Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly
Sign up
Appearance settings

feat: Add support for openai realtime api in adk #2720

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
ac-machache wants to merge 2 commits into google:main
base: main
Choose a base branch
Loading
from ac-machache:feat/add-support-for-openai-realtime
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
51 changes: 51 additions & 0 deletions src/google/adk/agents/base_agent.py
View file Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -283,6 +283,41 @@ async def _run_with_trace() -> AsyncGenerator[Event, None]:
async for event in agen:
yield event

@final
async def run_realtime(
self,
parent_context: InvocationContext,
) -> AsyncGenerator[Event, None]:
"""Entry method to run an agent via video/audio-based conversation.

Args:
parent_context: InvocationContext, the invocation context of the parent
agent.

Yields:
Event: the events generated by the agent.
"""

async def _run_with_trace() -> AsyncGenerator[Event, None]:
with tracer.start_as_current_span(f'agent_run [{self.name}]'):
ctx = self._create_invocation_context(parent_context)

if event := await self.__handle_before_agent_callback(ctx):
yield event
if ctx.end_invocation:
return

async with Aclosing(self._run_realtime_impl(ctx)) as agen:
async for event in agen:
yield event

if event := await self.__handle_after_agent_callback(ctx):
yield event

async with Aclosing(_run_with_trace()) as agen:
async for event in agen:
yield event

async def _run_async_impl(
self, ctx: InvocationContext
) -> AsyncGenerator[Event, None]:
Expand Down Expand Up @@ -315,6 +350,22 @@ async def _run_live_impl(
)
yield # AsyncGenerator requires having at least one yield statement

async def _run_realtime_impl(
self, ctx: InvocationContext
) -> AsyncGenerator[Event, None]:
"""Core logic to run this agent via video/audio-based conversation.

Args:
ctx: InvocationContext, the invocation context for this agent.

Yields:
Event: the events generated by the agent.
"""
raise NotImplementedError(
f'_run_realtime_impl for {type(self)} is not implemented.'
)
yield # AsyncGenerator requires having at least one yield statement

@property
def root_agent(self) -> BaseAgent:
"""Gets the root agent of this agent."""
Expand Down
26 changes: 26 additions & 0 deletions src/google/adk/agents/llm_agent.py
View file Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,10 +40,14 @@
from ..events.event import Event
from ..flows.llm_flows.auto_flow import AutoFlow
from ..flows.llm_flows.base_llm_flow import BaseLlmFlow
from ..flows.llm_flows.openai_llm_flow import OpenAILlmFlow
from ..flows.llm_flows.openai_llm_flow import OpenAutoFlow
from ..flows.llm_flows.openai_llm_flow import OpenSingleFlow
from ..flows.llm_flows.single_flow import SingleFlow
from ..models.base_llm import BaseLlm
from ..models.llm_request import LlmRequest
from ..models.llm_response import LlmResponse
from ..models.openai_llm import OpenAIRealtime
from ..models.registry import LLMRegistry
from ..planners.base_planner import BasePlanner
from ..tools.base_tool import BaseTool
Expand Down Expand Up @@ -300,6 +304,17 @@ async def _run_live_impl(
if ctx.end_invocation:
return

@override
async def _run_realtime_impl(
self, ctx: InvocationContext
) -> AsyncGenerator[Event, None]:
async with Aclosing(self._open_llm_flow.run_live(ctx)) as agen:
async for event in agen:
self.__maybe_save_output_to_state(event)
yield event
if ctx.end_invocation:
return

@property
def canonical_model(self) -> BaseLlm:
"""The resolved self.model field as BaseLlm.
Expand Down Expand Up @@ -443,6 +458,17 @@ def _llm_flow(self) -> BaseLlmFlow:
else:
return AutoFlow()

@property
def _open_llm_flow(self) -> OpenAILlmFlow:
if (
self.disallow_transfer_to_parent
and self.disallow_transfer_to_peers
and not self.sub_agents
):
return OpenSingleFlow()
else:
return OpenAutoFlow()

def __maybe_save_output_to_state(self, event: Event):
"""Saves the model output to state if needed."""
# skip if the event was authored by some other agent (e.g. current agent
Expand Down
28 changes: 28 additions & 0 deletions src/google/adk/agents/run_config.py
View file Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@
from enum import Enum
import logging
import sys
from typing import Any
from typing import Optional

from google.genai import types
Expand Down Expand Up @@ -107,3 +108,30 @@ def validate_max_llm_calls(cls, value: int) -> int:
)

return value

# ---------------- OpenAI Realtime (vendor-specific) ----------------
openai_realtime_session: Optional[Any] = None
"""Optional raw OpenAI Realtime session settings.

These will be forwarded into the OpenAI Realtime `session.update` payload.
Examples:
{"modalities": ["text"], "turn_detection": {"type": "none"}, "voice": "ash"}

Accepts either a Python dict or a JSON string. Prefer generic fields
(response_modalities, realtime_input_config) when possible. Use this for
OpenAI-specific knobs not covered by the generic config.

Example (dict) with audio configuration:
{
"modalities": ["audio", "text"],
"voice": "ash",
"input_audio_format": "pcm16",
"output_audio_format": "pcm16",
"input_audio_transcription": {"model": "gpt-4o-mini-transcribe", "language": "fr"},
"turn_detection": {"type": "semantic_vad"}
}

Notes:
- If you pass a JSON string instead of a dict, it will be parsed the same
way (single quotes are not valid JSON; use double quotes).
"""
26 changes: 26 additions & 0 deletions src/google/adk/flows/llm_flows/basic.py
View file Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@

from __future__ import annotations

import logging
from typing import AsyncGenerator
from typing import Generator

Expand All @@ -27,6 +28,8 @@
from ...models.llm_request import LlmRequest
from ._base_llm_processor import BaseLlmRequestProcessor

logger = logging.getLogger('google_adk.' + __name__)


class _BasicLlmRequestProcessor(BaseLlmRequestProcessor):

Expand Down Expand Up @@ -82,6 +85,29 @@ async def run_async(
invocation_context.run_config.session_resumption
)

# Vendor-specific: OpenAI Realtime session settings passthrough
# These will be consumed by the OpenAI provider when building session.update
openai_session = getattr(
invocation_context.run_config, 'openai_realtime_session', None
)
if openai_session is not None:
if llm_request.config.labels is None:
llm_request.config.labels = {}
# Prefer dict passthrough; if a JSON string is provided, pass as-is.
payload = openai_session
try:
# Normalize Pydantic/GenAI types to plain JSON if available
if hasattr(types, 'any_serialize'):
payload = types.any_serialize(openai_session)
except Exception as e:
logger.warning(
'Failed to any_serialize openai_realtime_session; passing raw'
' payload. %s',
e,
exc_info=True,
)
llm_request.config.labels['adk_openai_session_json'] = payload

# TODO: handle tool append here, instead of in BaseTool.process_llm_request.

return
Expand Down
Loading

AltStyle によって変換されたページ (->オリジナル) /