Skip to content

Build and run agents you can see, understand and trust.

fix(middleware): use configured id factory for TTS audio blocks (#1930)

YingchaoX134647a2a5f1
diff --git a/src/agentscope/middleware/_tts_middleware.py b/src/agentscope/middleware/_tts_middleware.py
index cff4502ee..32c1d9560 100644
--- a/src/agentscope/middleware/_tts_middleware.py
+++ b/src/agentscope/middleware/_tts_middleware.py
@@ -1,10 +1,10 @@
# -*- coding: utf-8 -*-
"""Middleware that turns reasoning text into speech and injects it as
``DATA_BLOCK_*`` events into the agent's event stream."""
-import uuid
from typing import TYPE_CHECKING, AsyncGenerator, Callable
from ._base import MiddlewareBase
+from .._utils._common import _generate_id
from ..event import (
DataBlockDeltaEvent,
DataBlockEndEvent,
@@ -163,7 +163,7 @@ class TTSMiddleware(MiddlewareBase):
return
if audio_block_id is None:
- audio_block_id = uuid.uuid4().hex
+ audio_block_id = _generate_id()
audio_media_type = media_type
yield DataBlockStartEvent(
reply_id=agent.state.reply_id,
diff --git a/tests/tts_middleware_test.py b/tests/tts_middleware_test.py
index 3f296c631..0b94be346 100644
--- a/tests/tts_middleware_test.py
+++ b/tests/tts_middleware_test.py
@@ -5,6 +5,7 @@ from typing import Any, AsyncGenerator
from unittest import IsolatedAsyncioTestCase
from unittest.mock import AsyncMock, MagicMock
+from agentscope import set_id_factory
from agentscope.event import (
DataBlockDeltaEvent,
DataBlockEndEvent,
@@ -158,6 +159,43 @@ class TestTTSMiddlewareNonRealtime(IsolatedAsyncioTestCase):
)
tts.synthesize.assert_not_called()
+ async def test_audio_block_id_uses_configured_id_factory(self) -> None:
+ """TTS audio DATA_BLOCK events use the configured ID factory."""
+ import agentscope._utils._common as common
+
+ # pylint: disable=protected-access
+ saved_factory = common._id_factory
+ self.addCleanup(setattr, common, "_id_factory", saved_factory)
+ set_id_factory(lambda: "custom-audio-id")
+
+ tts = MagicMock(spec=TTSModelBase)
+ tts.realtime = False
+ tts.__aenter__ = AsyncMock(return_value=tts)
+ tts.__aexit__ = AsyncMock(return_value=None)
+ tts.synthesize = AsyncMock(
+ return_value=_make_tts_response("AAAA"),
+ )
+
+ middleware = TTSMiddleware(tts_model=tts)
+ agent = _make_agent_stub()
+
+ async def next_handler(**_kwargs: Any) -> AsyncGenerator:
+ yield TextBlockDeltaEvent(
+ reply_id="reply-1",
+ block_id="blk-1",
+ delta="hi",
+ )
+ yield TextBlockEndEvent(reply_id="reply-1", block_id="blk-1")
+
+ emitted = [
+ evt async for evt in middleware.on_reply(agent, {}, next_handler)
+ ]
+
+ self.assertEqual(
+ [evt.block_id for evt in emitted[2:]],
+ ["custom-audio-id"] * 3,
+ )
+
async def test_streaming_output_multiple_chunks(self) -> None:
"""When synthesize() returns an async generator, each chunk produces
a DATA_BLOCK_DELTA under the same block_id."""