Initial commit

2026-08-18 10:03:35 +08:00
commit 8e4424b82a
11 changed files with 1524 additions and 0 deletions
@@ -0,0 +1,56 @@
+model_list:
+    - model_name: gemini-flash-latest
+      litellm_params:
+          model: vertex_ai/gemini-flash-latest
+          api_key: os.environ/GEMINI_API_KEY
+
+    - model_name: gemini-flash-lite-latest
+      litellm_params:
+          model: vertex_ai/gemini-flash-lite-latest
+          api_key: os.environ/GEMINI_API_KEY
+
+    - model_name: gemini-pro-latest
+      litellm_params:
+          model: vertex_ai/gemini-pro-latest
+          api_key: os.environ/GEMINI_API_KEY
+
+    - model_name: chatgpt-pro
+      litellm_params:
+          model: openai/chatgpt-pro
+          api_base: http://cli-proxy-api:8317/v1
+          api_key: os.environ/CLIPROXY_API_KEY
+
+    # # 供 Open WebUI 的图片生成功能调用。不要把这些模型当普通聊天模型直接在对话里选择，
+    # # 否则走 /chat/completions 流式返回大块 base64 时，前端可能报 Chunk too big。
+    # - model_name: gemini-3-pro-image-preview
+    #   litellm_params:
+    #       model: gemini/gemini-3-pro-image-preview
+    #       api_key: os.environ/GEMINI_API_KEY
+
+litellm_settings:
+    drop_params: true
+
+router_settings:
+    num_retries: 3
+    timeout: 60
+    fallbacks:
+        - {
+              'gemini-flash-lite-latest':
+                  ['gemini-flash-latest', 'gemini-pro-latest'],
+          }
+        - { 'gemini-flash-latest': ['gemini-pro-latest'] }
+
+#
+# - model_name: gemini-flash-enterprise
+#   litellm_params:
+#       model: vertex_ai/gemini-2.5-flash
+#       vertex_project: 'gen-lang-client-0934968879'
+#       vertex_location: 'asia-southeast1'
+#       # vertex_credentials: '/app/litellm/gen-lang-client-0934968879-03bf3a88d4dd.json' # 容器内路径
+
+# - model_name: gemini-pro-enterprise
+#   litellm_params:
+#       model: vertex_ai/gemini-2.5-pro
+#       vertex_project: 'gen-lang-client-0934968879'
+#       vertex_location: 'asia-southeast1'
+# vertex_credentials: '/app/litellm/gen-lang-client-0934968879-03bf3a88d4dd.json' # 容器内路径
@@ -0,0 +1,45 @@
+<service>
+    <id>ai-stack-litellm</id>
+    <name>AI Stack - LiteLLM</name>
+    <description>LiteLLM proxy for Gemini models, provides OpenAI-compatible API on port 4000</description>
+
+    <!--
+        使用 WinSW 的 %BASE% 相对路径。
+        %BASE% = WinSW.exe 所在目录（本文件所在目录：...\litellm）。
+        现在使用专属 venv：..\litellm\venv，支持任意盘符部署。
+    -->
+    <executable>%BASE%\venv\Scripts\litellm.exe</executable>
+    <arguments>--config %BASE%\config.yaml --port 4000</arguments>
+    <workingdirectory>%BASE%\..</workingdirectory>
+
+    <!-- Gemini API Key 从 Windows 系统环境变量读取；install.ps1/update.ps1 会从 .env 同步 -->
+    <env name="GEMINI_API_KEY" value="%GEMINI_API_KEY%"/>
+
+    <!--
+        如果访问 Gemini 需要代理，取消注释并修改代理地址
+    -->
+    <!-- <env name="HTTP_PROXY" value="http://127.0.0.1:7890"/> -->
+    <!-- <env name="HTTPS_PROXY" value="http://127.0.0.1:7890"/> -->
+    <!-- <env name="ALL_PROXY" value="http://127.0.0.1:7890"/> -->
+
+    <!-- 日志，滚动模式防止无限增长 -->
+    <logpath>%BASE%\..\logs\litellm</logpath>
+    <log mode="roll">
+        <sizeThreshold>10240</sizeThreshold>
+        <keepFiles>5</keepFiles>
+    </log>
+
+    <!-- 开机自动启动，延迟启动避免网络未就绪 -->
+    <startmode>Automatic</startmode>
+    <delayedAutoStart>true</delayedAutoStart>
+
+    <!-- 失败自动重启，10 秒后重试，连续失败也持续重启 -->
+    <onfailure action="restart" delay="10 sec"/>
+    <resetfailure>1 hour</resetfailure>
+
+    <!-- 停止等待时间 -->
+    <stoptimeout>15 sec</stoptimeout>
+
+    <!-- 不显示控制台窗口 -->
+    <hidewindow>true</hidewindow>
+</service>