services: deepclaude: image: ghcr.io/erlichliu/deepclaude:latest ports: - "8000:8000" environment: ALLOW_API_KEY: your_allow_api_key # 请替换为你的 ALLOW_API_KEY ALLOW_ORIGINS: "*" # 请替换为你的 允许的域名,例如 "http://localhost:3000,https://chat.example.com" 或 "*" DEEPSEEK_API_KEY: your_deepseek_api_key # 请替换为你的 DeepSeek API KEY DEEPSEEK_API_URL: https://api.deepseek.com/v1/chat/completions # 如果使用 SiliconFlow,则使用 https://api.siliconflow.cn/v1/chat/completions DEEPSEEK_MODEL: deepseek-reasoner # 如果使用 SiliconFlow,则使用 deepseek-ai/DeepSeek-R1 IS_ORIGIN_REASONING: true # 请替换为你的 DeepSeek 推理过程格式配置 # Claude 配置 CLAUDE_API_KEY: your_claude_api_key # 请替换为你的 Claude API KEY CLAUDE_MODEL: claude-3-5-sonnet-20241022 CLAUDE_PROVIDER: anthropic # 可选值: anthropic, openrouter, oneapi CLAUDE_API_URL: https://api.anthropic.com/v1/messages # 如果使用 OpenRouter 或 OneAPI,请替换为相应的 API URL # OpenAI 兼容模型配置(例如 Gemini) OPENAI_COMPOSITE_API_KEY: your_openai_compatible_api_key # 请替换为你的 API KEY OPENAI_COMPOSITE_API_URL: your_openai_compatible_api_url # 请替换为目标服务的 API URL,例如 Gemini 的 URL OPENAI_COMPOSITE_MODEL: your_model_name # 请替换为目标服务的模型名称,例如 gemini-2.0-flash LOG_LEVEL: INFO # 可选值:DEBUG, INFO, WARNING, ERROR, CRITICAL restart: always # 建议添加 restart policy,确保容器意外退出后可以自动重启