Skip to content

armgit / dify

Public Imported from GitHub
0

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.

refactor(api): inject TextToSpeechVoiceQuery into agent TTS voices via the decorator (#43006)

jeongeuna154a4d67401
diff --git a/api/controllers/console/app/audio.py b/api/controllers/console/app/audio.py
index 14e9642408c..889db05569b 100644
--- a/api/controllers/console/app/audio.py
+++ b/api/controllers/console/app/audio.py
@@ -455,14 +455,15 @@ class AgentTextToSpeechVoicesApi(Resource):
@with_current_user
@with_current_tenant_id
@with_session(write=False)
+ @model_validate(TextToSpeechVoiceQuery)
def get(
self,
+ req_data: TextToSpeechVoiceQuery,
session: Session,
current_tenant_id: str,
current_user: Account,
agent_id: UUID,
):
- query = TextToSpeechVoiceQuery.model_validate(request.args.to_dict(flat=True))
enforce_rbac_checks(
tenant_id=current_tenant_id,
account_id=current_user.id,
@@ -474,4 +475,4 @@ class AgentTextToSpeechVoicesApi(Resource):
tenant_id=current_tenant_id,
agent_id=agent_id,
)
- return _get_text_to_speech_voices(tenant_id=app_model.tenant_id, language=query.language)
+ return _get_text_to_speech_voices(tenant_id=app_model.tenant_id, language=req_data.language)
diff --git a/api/tests/unit_tests/controllers/console/app/test_audio.py b/api/tests/unit_tests/controllers/console/app/test_audio.py
index 8fa69b8fbcc..50aded215fa 100644
--- a/api/tests/unit_tests/controllers/console/app/test_audio.py
+++ b/api/tests/unit_tests/controllers/console/app/test_audio.py
@@ -9,7 +9,7 @@ import pytest
from flask import Flask, Response
from sqlalchemy.orm import Session
from werkzeug.datastructures import FileStorage
-from werkzeug.exceptions import Forbidden, InternalServerError
+from werkzeug.exceptions import Forbidden, InternalServerError, UnprocessableEntity
from controllers.console.app import audio as audio_module
from controllers.console.app.audio import (
@@ -540,6 +540,7 @@ def test_agent_text_to_speech_voices_uses_backing_app_and_language(app: Flask, u
):
response = unwrap(api.get)(
api,
+ TextToSpeechVoiceQuery(language="en-US"),
session=unbound_session,
current_tenant_id="tenant-1",
current_user=current_user,
@@ -573,6 +574,7 @@ def test_agent_text_to_speech_voices_does_not_query_provider_without_permission(
):
unwrap(api.get)(
api,
+ TextToSpeechVoiceQuery(language="en-US"),
session=unbound_session,
current_tenant_id="tenant-1",
current_user=_account(),
@@ -583,6 +585,30 @@ def test_agent_text_to_speech_voices_does_not_query_provider_without_permission(
resolve_app.assert_not_called()
+def test_agent_text_to_speech_voices_rejects_missing_language(app: Flask, unbound_session: Session) -> None:
+ agent_id = UUID("019ef3d2-b24c-7803-b428-18b5ee8fb853")
+ api = AgentTextToSpeechVoicesApi()
+ view = unwrap(api.get)
+ # Stop at the innermost decorator so request validation runs without the auth and session layers.
+ handler = unwrap(api.get, stop=lambda f: f.__dict__.get("__wrapped__") is view)
+ with (
+ patch.object(audio_module, "resolve_existing_agent_runtime_app_model") as resolve_app,
+ patch.object(audio_module, "enforce_rbac_checks") as check_access,
+ app.test_request_context(f"/console/api/agent/{agent_id}/text-to-audio/voices", method="GET"),
+ pytest.raises(UnprocessableEntity),
+ ):
+ handler(
+ api,
+ session=unbound_session,
+ current_tenant_id="tenant-1",
+ current_user=_account(),
+ agent_id=agent_id,
+ )
+
+ check_access.assert_not_called()
+ resolve_app.assert_not_called()
+
+
@pytest.mark.parametrize("message_id", [None, "019ef3d2-b24c-7803-b428-18b5ee8fb854"])
def test_agent_text_to_speech_returns_audio_with_scoped_message(
app: Flask, unbound_session: Session, message_id: str | None
@@ -665,7 +691,9 @@ def test_agent_tts_preserves_missing_agent_error(app: Flask, unbound_session: Se
"current_user": _account(),
"agent_id": agent_id,
}
- if not voices:
+ if voices:
+ args["req_data"] = TextToSpeechVoiceQuery(language="en-US")
+ else:
args["req_data"] = TextToSpeechPayload(text="Preview")
with (
patch.object(audio_module, "resolve_existing_agent_runtime_app_model", side_effect=AgentNotFoundError()),