Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix(elevenlabs): preserve API key on restart

harry070301def3b95b4c
diff --git a/app/services/voice.py b/app/services/voice.py
index 7a1176e6..3ef5fb85 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -252,6 +252,18 @@ def is_elevenlabs_voice(voice_name: str) -> bool:
return (voice_name or "").startswith("elevenlabs:")
+def get_elevenlabs_api_key() -> str:
+ """
+ 读取 ElevenLabs TTS 使用的 API Key。
+
+ 配置文件优先,环境变量仅作为未配置时的后备来源。WebUI 和配乐功能已经
+ 支持 ``ELEVENLABS_API_KEY``,TTS 必须使用相同规则,否则仅通过容器环境
+ 变量部署时,音色列表可正常加载,真正合成语音却会误报未配置 Key。
+ """
+ configured_key = str(config.elevenlabs.get("api_key", "") or "").strip()
+ return configured_key or os.getenv("ELEVENLABS_API_KEY", "").strip()
+
+
def is_chatterbox_voice(voice_name: str) -> bool:
return (voice_name or "").startswith("chatterbox:")
@@ -1288,7 +1300,7 @@ def elevenlabs_tts(
logger.error("ElevenLabs TTS text is empty")
return None
- api_key = config.elevenlabs.get("api_key", "")
+ api_key = get_elevenlabs_api_key()
if not api_key:
logger.error("ElevenLabs API key is not set")
return None
diff --git a/test/services/test_voice.py b/test/services/test_voice.py
index 23126e46..5e4c8a09 100644
--- a/test/services/test_voice.py
+++ b/test/services/test_voice.py
@@ -1024,7 +1024,10 @@ class TestElevenLabsVoice(unittest.TestCase):
@patch("app.services.voice.config")
def test_elevenlabs_tts_no_api_key(self, mock_config):
mock_config.elevenlabs.get.return_value = ""
- result = vs.elevenlabs_tts("Hello", "abc123", "/tmp/test.mp3")
+ # Key 解析包含环境变量回退,测试必须显式清空宿主环境,避免开发机或 CI
+ # 恰好设置 ELEVENLABS_API_KEY 后改变“未配置”的测试前提。
+ with patch.dict(os.environ, {}, clear=True):
+ result = vs.elevenlabs_tts("Hello", "abc123", "/tmp/test.mp3")
self.assertIsNone(result)
@patch("app.services.voice.config")
@@ -1033,6 +1036,42 @@ class TestElevenLabsVoice(unittest.TestCase):
result = vs.elevenlabs_tts(" ", "abc123", "/tmp/test.mp3")
self.assertIsNone(result)
+ def test_elevenlabs_api_key_prefers_config(self):
+ with (
+ patch.object(vs.config, "elevenlabs", {"api_key": "config-key"}),
+ patch.dict(os.environ, {"ELEVENLABS_API_KEY": "env-key"}),
+ ):
+ self.assertEqual(vs.get_elevenlabs_api_key(), "config-key")
+
+ def test_elevenlabs_api_key_falls_back_to_environment(self):
+ with (
+ patch.object(vs.config, "elevenlabs", {"api_key": ""}),
+ patch.dict(os.environ, {"ELEVENLABS_API_KEY": " env-key "}),
+ ):
+ self.assertEqual(vs.get_elevenlabs_api_key(), "env-key")
+
+ def test_elevenlabs_api_key_matches_music_service(self):
+ """TTS 和配乐共用同一账号配置,两条生成链路必须解析出相同 Key。"""
+ from app.services import elevenlabs_music
+
+ for configured_key, env_key in (("config-key", "env-key"), ("", "env-key")):
+ with self.subTest(configured_key=configured_key):
+ with (
+ patch.object(
+ vs.config, "elevenlabs", {"api_key": configured_key}
+ ),
+ patch.object(
+ elevenlabs_music.config,
+ "elevenlabs",
+ {"api_key": configured_key},
+ ),
+ patch.dict(os.environ, {"ELEVENLABS_API_KEY": env_key}),
+ ):
+ self.assertEqual(
+ vs.get_elevenlabs_api_key(),
+ elevenlabs_music.get_api_key(),
+ )
+
if __name__ == "__main__":
# python -m unittest test.services.test_voice.TestVoiceService.test_azure_tts_v1
diff --git a/test/services/test_webui_tts_settings.py b/test/services/test_webui_tts_settings.py
index 9d92adf9..cce8d7ac 100644
--- a/test/services/test_webui_tts_settings.py
+++ b/test/services/test_webui_tts_settings.py
@@ -1,4 +1,5 @@
import json
+import os
from pathlib import Path
from unittest.mock import patch
@@ -94,3 +95,62 @@ def test_tts_provider_inputs_render_the_standardized_labels():
assert not getattr(api_key_input.proto, "help", "")
assert [str(item.value) for item in app.exception] == []
+
+
+def test_elevenlabs_reconnect_restores_saved_key_before_loading_voices():
+ """
+ 服务重启后浏览器可能重放空密码状态;WebUI 应保留配置并在当前 rerun 就用
+ 已保存的 Key 加载音色,而不是只避免写空、却继续以空 Key 请求服务。
+ """
+ test_config = dict(config.elevenlabs, api_key="saved-key")
+ test_ui = dict(
+ config.ui,
+ voice_mode="tts",
+ tts_server="elevenlabs",
+ voice_name="",
+ )
+
+ with (
+ patch.object(config, "elevenlabs", test_config),
+ patch.object(config, "ui", test_ui),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.object(voice, "get_elevenlabs_voices", return_value=[]) as get_voices,
+ ):
+ app = AppTest.from_file(str(WEBUI_MAIN), default_timeout=30)
+ app.session_state["ui_language"] = "en"
+ app.session_state["elevenlabs_api_key_input"] = ""
+ app.run()
+
+ assert test_config["api_key"] == "saved-key"
+ assert app.session_state["elevenlabs_api_key_input"] == "saved-key"
+ assert get_voices.call_count >= 1
+ assert all(call.args == ("saved-key",) for call in get_voices.call_args_list)
+ assert [str(item.value) for item in app.exception] == []
+
+
+def test_elevenlabs_environment_key_is_used_without_persisting_it():
+ """环境变量可以驱动音色加载,但不能被 WebUI 自动复制进 config.toml。"""
+ test_config = dict(config.elevenlabs, api_key="")
+ test_ui = dict(
+ config.ui,
+ voice_mode="tts",
+ tts_server="elevenlabs",
+ voice_name="",
+ )
+
+ with (
+ patch.object(config, "elevenlabs", test_config),
+ patch.object(config, "ui", test_ui),
+ patch.object(config, "try_save_config", return_value=True),
+ patch.dict(os.environ, {"ELEVENLABS_API_KEY": "env-key"}),
+ patch.object(voice, "get_elevenlabs_voices", return_value=[]) as get_voices,
+ ):
+ app = AppTest.from_file(str(WEBUI_MAIN), default_timeout=30)
+ app.session_state["ui_language"] = "en"
+ app.run()
+
+ assert test_config["api_key"] == ""
+ assert app.session_state["elevenlabs_api_key_input"] == "env-key"
+ assert get_voices.call_count >= 1
+ assert all(call.args == ("env-key",) for call in get_voices.call_args_list)
+ assert [str(item.value) for item in app.exception] == []
diff --git a/webui/Main.py b/webui/Main.py
index 3a9ac095..f386944e 100644
--- a/webui/Main.py
+++ b/webui/Main.py
@@ -2935,6 +2935,45 @@ def _get_reusable_full_voice_preview(params, voice_mode: str) -> dict | None:
}
+def _sync_elevenlabs_api_key_input():
+ """
+ 同步 ElevenLabs 密码控件、持久化配置和环境变量,并返回当前有效 Key。
+
+ Streamlit 在浏览器标签页连接到重启后的服务时,可能重放一个空的密码控件
+ 状态。这个空值无法与用户主动清空可靠区分,因此当配置文件或环境变量仍有
+ Key 时,优先恢复有效值,防止空状态覆盖配置并确保本次 rerun 能立即加载
+ 音色。需要彻底删除 Key 时应修改配置文件或环境变量,避免重连误判。
+ """
+ widget_key = "elevenlabs_api_key_input"
+ configured_key = str(config.elevenlabs.get("api_key", "") or "").strip()
+ env_key = os.getenv("ELEVENLABS_API_KEY", "").strip()
+ effective_key = configured_key or env_key
+ had_widget_state = widget_key in st.session_state
+ entered_key = str(st.session_state.get(widget_key, "") or "").strip()
+
+ if not entered_key and effective_key:
+ # 重连后的空状态不能覆盖有效凭证,同时必须在渲染音色列表之前恢复,
+ # 否则配置文件虽然没有被清空,当前页面仍会使用空 Key 请求 ElevenLabs。
+ st.session_state[widget_key] = effective_key
+ entered_key = effective_key
+ if had_widget_state:
+ logger.debug("restored ElevenLabs API key after empty session replay")
+ elif not had_widget_state:
+ # 先初始化再创建控件,避免同时传 value 和 session_state 触发 Streamlit
+ # 的默认值冲突警告;没有任何 Key 时初始化为空即可。
+ st.session_state[widget_key] = entered_key
+
+ if entered_key and entered_key != effective_key:
+ # 用户主动输入的新值才落入 config.toml。环境变量作为有效值回填时不会
+ # 被复制到文件,容器或部署平台注入的密钥仍只保留在运行环境中。
+ for cache_key in list(st.session_state.keys()):
+ if str(cache_key).startswith("elevenlabs_voices_"):
+ del st.session_state[cache_key]
+ _set_runtime_config("elevenlabs", "api_key", entered_key)
+
+ return entered_key
+
+
def _render_elevenlabs_api_key_input(label_key):
"""
渲染 ElevenLabs TTS 与配乐共用的唯一 API Key 输入状态。
@@ -2943,26 +2982,13 @@ def _render_elevenlabs_api_key_input(label_key):
后渲染的输入框还会覆盖共享配置。这里统一使用一个 key,并集中处理环境变量
回填、配置更新和音色缓存失效,确保界面显示与后台任务始终读取同一个值。
"""
- configured_key = str(config.elevenlabs.get("api_key", "") or "").strip()
- effective_key = configured_key or os.getenv("ELEVENLABS_API_KEY", "").strip()
- entered_key = st.text_input(
+ _sync_elevenlabs_api_key_input()
+ return st.text_input(
tr(label_key),
- value=effective_key,
type="password",
key="elevenlabs_api_key_input",
).strip()
- if entered_key != effective_key:
- for cache_key in list(st.session_state.keys()):
- if str(cache_key).startswith("elevenlabs_voices_"):
- del st.session_state[cache_key]
-
- # 环境变量仅用于当前进程,不在用户未修改时自动复制到 config.toml。
- # 已有配置或用户主动修改输入时才更新本机配置,与 Sonilo 行为保持一致。
- if configured_key or entered_key != effective_key:
- _set_runtime_config("elevenlabs", "api_key", entered_key)
- return entered_key
-
def _render_background_music_settings(params, elevenlabs_api_key_rendered=False):
"""渲染背景音乐来源与音量设置,并返回本次待保存的上传文件。"""
@@ -3270,17 +3296,9 @@ def _render_audio_settings(panel, params):
# 获取 Xiaomi MiMo TTS 的预置音色列表
filtered_voices = voice.get_mimo_voices()
elif selected_tts_server == "elevenlabs":
- # Read from session_state first so the API key is available before
- # the Play Voice button runs (which is earlier in the script than
- # the API key text_input widget).
- saved_elevenlabs_api_key = st.session_state.get(
- "elevenlabs_api_key_input",
- config.elevenlabs.get("api_key", ""),
- )
- if saved_elevenlabs_api_key:
- _set_runtime_config(
- "elevenlabs", "api_key", saved_elevenlabs_api_key
- )
+ # 音色列表位于 Key 输入框之前渲染,必须先统一恢复重连状态并读取
+ # 配置/环境变量,否则页面会用空 Key 加载并缓存空音色列表。
+ saved_elevenlabs_api_key = _sync_elevenlabs_api_key_input()
cache_key = f"elevenlabs_voices_{saved_elevenlabs_api_key}"
if cache_key not in st.session_state:
st.session_state[cache_key] = voice.get_elevenlabs_voices(