provider: openai_api_compatible
label:
  en_US: OpenAI-API-compatible
description:
  en_US: Model providers compatible with OpenAI's API standard, such as LM Studio.
  zh_Hans: 兼容 OpenAI API 的模型供应商，例如 LM Studio 。
icon_small:
  en_US: icon.svg
supported_model_types:
  - llm
  - rerank
  - text-embedding
  - speech2text
  - tts
configurate_methods:
  - customizable-model
model_credential_schema:
  model:
    label:
      en_US: Model Name
      zh_Hans: 模型名称
    placeholder:
      en_US: Enter full model name
      zh_Hans: 输入模型全称
  credential_form_schemas:
    - variable: display_name
      label:
        en_US: Model display name
        zh_Hans: 模型显示名称
      type: text-input
      required: false
      placeholder:
        zh_Hans: 模型在界面的显示名称
        en_US: The display name of the model in the interface.
    - variable: api_key
      label:
        zh_Hans: API Key
        en_US: API Key
      type: secret-input
      required: false
      placeholder:
        zh_Hans: 在此输入您的 API Key
        en_US: Enter your API Key
    - variable: language
      label:
        en_US: Language
        zh_Hans: 语言
      type: text-input
      required: false
      show_on:
        - variable: __model_type
          value: speech2text
      placeholder:
        zh_Hans: ISO-639-1 语言代码，例如 zh、en、ru；留空则由模型自动检测
        en_US: ISO-639-1 language code, e.g. zh, en, ru. Leave empty to let the model auto-detect.
    - variable: initial_prompt
      label:
        en_US: Initial Prompt
        zh_Hans: 初始提示
      type: text-input
      required: false
      show_on:
        - variable: __model_type
          value: speech2text
      placeholder:
        zh_Hans: 模型在对话开始时的初始提示
        en_US: The initial prompt for the model at the start of the conversation.
    - variable: endpoint_url
      label:
        zh_Hans: API Base URL
        en_US: API Base URL
      type: text-input
      required: true
      placeholder:
        zh_Hans: Base URL, e.g. https://api.openai.com/v1
        en_US: Base URL, e.g. https://api.openai.com/v1
    - variable: rerank_endpoint_url
      label:
        zh_Hans: 重排序端点 URL
        en_US: Rerank Endpoint URL
      type: text-input
      required: false
      show_on:
        - variable: __model_type
          value: rerank
      placeholder:
        zh_Hans: 可选完整 URL；留空使用 API Base URL + /rerank；请勿包含凭据或令牌
        en_US: Optional complete URL; defaults to API Base URL + /rerank; do not include credentials or tokens
    - variable: endpoint_model_name
      label:
        zh_Hans: API endpoint中的模型名称
        en_US: model name for API endpoint
      type: text-input
      required: false
      placeholder:
        zh_Hans: endpoint model name, e.g. chatgpt4.0
        en_US: endpoint model name, e.g. chatgpt4.0
    - variable: mode
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 对话类型
        en_US: Completion mode
      type: select
      required: false
      default: chat
      placeholder:
        zh_Hans: 选择对话类型
        en_US: Select completion mode
      options:
        - value: completion
          label:
            en_US: Completion
            zh_Hans: 补全
        - value: chat
          label:
            en_US: Chat
            zh_Hans: 对话
    - variable: context_size
      label:
        zh_Hans: 模型上下文长度
        en_US: Model context size
      required: true
      show_on:
        - variable: __model_type
          value: llm
      type: text-input
      default: "4096"
      placeholder:
        zh_Hans: 在此输入您的模型上下文长度
        en_US: Enter your Model context size
    - variable: max_chunks
      label:
        zh_Hans: 模型每批的最大分块数
        en_US: Model max chunks per batch
      required: true
      show_on:
        - variable: __model_type
          value: text-embedding
      type: text-input
      default: "1"
      placeholder:
        zh_Hans: 在此输入您的模型每批次的最大分块数量
        en_US: Enter your Model maximum chunks per batch size
    - variable: context_size
      label:
        zh_Hans: 模型上下文长度
        en_US: Model context size
      required: true
      show_on:
        - variable: __model_type
          value: text-embedding
      type: text-input
      default: "4096"
      placeholder:
        zh_Hans: 在此输入您的模型上下文长度
        en_US: Enter your Model context size
    - variable: encoding_format
      label:
        zh_Hans: Embedding encoding format
        en_US: Embedding encoding format
      required: false
      show_on:
        - variable: __model_type
          value: text-embedding
      type: select
      default: not_set
      options:
        - value: not_set
          label:
            zh_Hans: Not set
            en_US: Not set
        - value: float
          label:
            zh_Hans: float
            en_US: float
      help:
        zh_Hans: Set this to float for strict OpenAI-compatible gateways such as LiteLLM that require an explicit embedding encoding_format.
        en_US: Set this to float for strict OpenAI-compatible gateways such as LiteLLM that require an explicit embedding encoding_format.
    - variable: vision_support
      show_on:
        - variable: __model_type
          value: text-embedding
      label:
        zh_Hans: 视觉支持
        en_US: Vision Support
      type: select
      required: false
      default: no_support
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: context_size
      label:
        zh_Hans: 模型上下文长度
        en_US: Model context size
      required: true
      show_on:
        - variable: __model_type
          value: rerank
      type: text-input
      default: "4096"
      placeholder:
        zh_Hans: 在此输入您的模型上下文长度
        en_US: Enter your Model context size
    - variable: vision_support
      show_on:
        - variable: __model_type
          value: rerank
      label:
        zh_Hans: 视觉支持
        en_US: Vision Support
      type: select
      required: false
      default: no_support
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: max_tokens_to_sample
      label:
        zh_Hans: 最大 token 上限
        en_US: Upper bound for max tokens
      show_on:
        - variable: __model_type
          value: llm
      default: "4096"
      type: text-input
    - variable: agent_thought_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: Thinking Mode Support
        zh_Hans: 思考模式支持
      type: select
      required: false
      default: not_supported
      options:
        - value: only_thinking_supported
          label:
            en_US: Only Thinking Mode
            zh_Hans: 仅支持思考模式
        - value: not_supported
          label:
            en_US: Only Non-Thinking Mode
            zh_Hans: 仅支持非思考模式
        - value: supported
          label:
            en_US: Both Modes
            zh_Hans: 两种模式都支持
      help:
        en_US: Configure whether the model supports thinking mode, non-thinking mode, or both. This will control the availability of the thinking mode toggle.
        zh_Hans: 配置模型是否支持思考模式、非思考模式或两者都支持。这将控制思考模式开关的可用性。
    - variable: web_search_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: Web Search Support
        zh_Hans: 联网搜索支持
      type: select
      required: false
      default: not_supported
      options:
        - value: not_supported
          label:
            en_US: Not Supported
            zh_Hans: 不支持
        - value: tool_standard
          label:
            en_US: Standard Tools Format
            zh_Hans: 标准 Tools 格式
        - value: tool_simple
          label:
            en_US: Simple Tools Format
            zh_Hans: 简单 Tools 格式
        - value: parameter
          label:
            en_US: Parameter Format
            zh_Hans: 参数格式
      help:
        en_US: >
          Configure web search support and format. 
          Standard Tools: {"type": "web_search", "web_search": {"enable": true}} (ZhipuAI, Baichuan). 
          Simple Tools: {"type": "web_search"} (Volcengine). 
          Parameter: {"web_search": true} as a top-level parameter.
        zh_Hans: >
          配置联网搜索支持和格式。
          标准 Tools 格式: {"type": "web_search", "web_search": {"enable": true}} (智谱AI、百川)。
          简单 Tools 格式: {"type": "web_search"} (火山引擎)。
          参数格式: {"web_search": true} 作为顶层参数。
    - variable: compatibility_mode
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: Compatibility mode
        zh_Hans: 兼容模式
      type: select
      required: false
      default: strict
      options:
        - value: strict
          label:
            en_US: OpenAI compatible
            zh_Hans: 严格（OpenAI 兼容）
        - value: extended
          label:
            en_US: Enable extra settings
            zh_Hans: 扩展（启用额外设置）
      help:
        en_US: Choose strict OpenAI compatibility or extended settings that add extra parameters (e.g., thinking, chat_template_kwargs, enable_thinking).
        zh_Hans: 选择严格的 OpenAI 兼容或扩展设置（会添加额外参数，例如 thinking、chat_template_kwargs、enable_thinking）。
    - variable: api_type
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: API Type
        zh_Hans: API 类型
      type: select
      required: false
      default: chat_completions
      options:
        - value: chat_completions
          label:
            en_US: Chat Completions API (/chat/completions)
            zh_Hans: Chat Completions API (/chat/completions)
        - value: responses
          label:
            en_US: Responses API (/responses)
            zh_Hans: Responses API (/responses)
      help:
        en_US: Choose Chat Completions API (standard endpoint) or Responses API (required for OpenAI native web search with domain filtering).
        zh_Hans: 选择 Chat Completions API（标准端点）或 Responses API（适用于包含域名过滤的 OpenAI 原生联网搜索）。
    - variable: token_param_name
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: Token parameter name
        zh_Hans: token 参数名
      type: select
      required: false
      default: auto
      options:
        - value: auto
          label:
            en_US: Auto (detect by model)
            zh_Hans: 自动（按模型判断）
        - value: max_tokens
          label:
            en_US: Use max_tokens
            zh_Hans: 使用 max_tokens
        - value: max_completion_tokens
          label:
            en_US: Use max_completion_tokens
            zh_Hans: 使用 max_completion_tokens
      help:
        en_US: Some OpenAI-compatible endpoints require max_completion_tokens (Responses API). Choose explicitly if auto-detection is wrong.
        zh_Hans: 一些兼容端要求使用 max_completion_tokens（Responses API）。若自动判断不准确，可在此显式指定。
    - variable: user_identity_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: User Identity Support
        zh_Hans: 用户标识支持
      type: select
      required: false
      default: support
      help:
        en_US: >
          Whether the endpoint accepts the optional top-level user parameter.
          Some OpenAI-compatible gateways reject it as an unsupported parameter.
          Choose Not Support to omit it from the request body.
        zh_Hans: >
          端点是否接受可选的顶层 user 参数。
          部分 OpenAI 兼容网关会将其视为不支持的参数而拒绝请求。
          选择「不支持」可将其从请求体中省略。
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: stream_include_usage
      show_on:
        - variable: __model_type
          value: llm
      label:
        en_US: Include Usage in Stream
        zh_Hans: 流式中包含用量
      type: select
      required: false
      default: enabled
      help:
        en_US: >
          Whether to send `stream_options: {include_usage: true}` so the endpoint
          returns prompt/completion token counts in the final stream chunk.
          OpenAI-compatible servers (vLLM, SGLang, llama.cpp, ...) only emit
          `usage` when this is requested; without it, Dify falls back to local
          token estimation that counts only the first prompt message, which
          undercounts multi-message prompts (e.g. a long user prompt after a
          system prompt). Choose Disabled only if the endpoint rejects the
          field.
        zh_Hans: >
          是否发送 `stream_options: {include_usage: true}`，让端点在流式最后一片
          中返回 prompt / completion token 数量。OpenAI 兼容服务端（vLLM、SGLang、
          llama.cpp 等）只有在请求中显式要求时才会返回 `usage`；否则 Dify 会
          回退到本地估算，仅对第一条 prompt 消息计 token，导致多消息 prompt
          （例如系统提示后跟很长的用户消息）被少算。仅在端点拒绝该字段时
          选择「禁用」。
      options:
        - value: enabled
          label:
            en_US: Enabled
            zh_Hans: 启用
        - value: disabled
          label:
            en_US: Disabled
            zh_Hans: 禁用
    - variable: function_calling_type
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 函数调用类型
        en_US: Function Call Type
      type: select
      required: false
      default: no_call
      options:
        - value: function_call
          label:
            en_US: Function Call
            zh_Hans: 函数调用
        - value: tool_call
          label:
            en_US: Tool Call
            zh_Hans: 工具调用
        - value: no_call
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: stream_function_calling
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 流式函数调用
        en_US: Stream function calling
      type: select
      required: false
      default: not_supported
      options:
        - value: supported
          label:
            en_US: Support
            zh_Hans: 支持
        - value: not_supported
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: vision_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 视觉支持
        en_US: Vision Support
        ja_JP: 画像入力対応
      type: select
      required: false
      default: no_support
      help:
        en_US: Whether the model can accept image input. Covers IMAGE content only; enable video_support / audio_support / document_support independently for other modalities.
        zh_Hans: 模型是否支持图像输入。仅覆盖 IMAGE 类型内容；视频/音频/文档请分别启用对应的开关。
        ja_JP: モデルが画像入力を受け付けるかどうか。IMAGE タイプのみを対象。動画・音声・文書はそれぞれ独立した設定で有効化してください。
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: video_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 视频支持
        en_US: Video Support
        ja_JP: 動画入力対応
      type: select
      required: false
      default: no_support
      help:
        en_US: Whether the model can accept video input. Covers VIDEO content only; independent from vision_support.
        zh_Hans: 模型是否支持视频输入。仅覆盖 VIDEO 类型内容；与视觉支持相互独立。
        ja_JP: モデルが動画入力を受け付けるかどうか。VIDEO タイプのみを対象。画像入力対応とは独立した設定です。
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: audio_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 音频支持
        en_US: Audio Support
        ja_JP: 音声入力対応
      type: select
      required: false
      default: no_support
      help:
        en_US: Whether the model can accept audio input. Covers AUDIO content only; independent from vision_support.
        zh_Hans: 模型是否支持音频输入。仅覆盖 AUDIO 类型内容；与视觉支持相互独立。
        ja_JP: モデルが音声入力を受け付けるかどうか。AUDIO タイプのみを対象。画像入力対応とは独立した設定です。
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: document_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 文档支持
        en_US: Document Support
        ja_JP: 文書入力対応
      type: select
      required: false
      default: no_support
      help:
        en_US: Whether the model can accept document input such as PDF. Covers DOCUMENT content only; independent from vision_support.
        zh_Hans: 模型是否支持文档输入（如 PDF）。仅覆盖 DOCUMENT 类型内容；与视觉支持相互独立。
        ja_JP: モデルが PDF などの文書入力を受け付けるかどうか。DOCUMENT タイプのみを対象。画像入力対応とは独立した設定です。
      options:
        - value: support
          label:
            en_US: Support
            zh_Hans: 支持
        - value: no_support
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: structured_output_support
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 结构化输出
        en_US: Structured Output
      type: select
      required: false
      default: not_supported
      options:
        - value: supported
          label:
            en_US: Support
            zh_Hans: 支持
        - value: not_supported
          label:
            en_US: Not Support
            zh_Hans: 不支持
    - variable: stream_mode_auth
      show_on:
        - variable: __model_type
          value: llm
      label:
        zh_Hans: 流模式授权
        en_US: Stream mode auth
      type: select
      required: false
      default: not_use
      options:
        - value: use
          label:
            en_US: Use
            zh_Hans: 使用
        - value: not_use
          label:
            en_US: Not Use
            zh_Hans: 不使用
    - variable: stream_mode_delimiter
      label:
        zh_Hans: 流模式返回结果的分隔符
        en_US: Delimiter for streaming results
      show_on:
        - variable: __model_type
          value: llm
      default: '\n\n'
      type: text-input
    - variable: voices
      show_on:
        - variable: __model_type
          value: tts
      label:
        en_US: Available Voices (comma-separated)
        zh_Hans: 可用声音（用英文逗号分隔）
      type: text-input
      required: false
      default: "alloy"
      placeholder:
        en_US: "alloy,echo,fable,onyx,nova,shimmer"
        zh_Hans: "alloy,echo,fable,onyx,nova,shimmer"
      help:
        en_US: "List voice names separated by commas. First voice will be used as default."
        zh_Hans: "用英文逗号分隔的声音列表。第一个声音将作为默认值。"
    - variable: tts_api_format
      show_on:
        - variable: __model_type
          value: tts
      label:
        en_US: TTS API Format
        zh_Hans: TTS API 格式
      type: select
      required: false
      default: openai
      options:
        - value: openai
          label:
            en_US: OpenAI Compatible
            zh_Hans: OpenAI 兼容
        - value: byteplus
          label:
            en_US: BytePlus TTS
            zh_Hans: BytePlus TTS
      help:
        en_US: "Select the TTS API format. Use 'BytePlus TTS' for BytePlus/Volcengine TTS services."
        zh_Hans: "选择 TTS API 格式。使用 'BytePlus TTS' 适配 BytePlus/火山引擎 TTS 服务。"
    - variable: tts_resource_id
      show_on:
        - variable: __model_type
          value: tts
      label:
        en_US: Resource ID (BytePlus)
        zh_Hans: Resource ID (BytePlus)
      type: text-input
      required: false
      placeholder:
        en_US: "e.g. seed-tts-2.0 (optional, defaults to model name)"
        zh_Hans: "例如 seed-tts-2.0（可选，默认使用模型名称）"
      help:
        en_US: "BytePlus TTS Resource ID. If empty, the model name will be used."
        zh_Hans: "BytePlus TTS 的 Resource ID。留空则使用模型名称。"
    - variable: tts_sample_rate
      show_on:
        - variable: __model_type
          value: tts
      label:
        en_US: Sample Rate (BytePlus)
        zh_Hans: 采样率 (BytePlus)
      type: select
      required: false
      default: "24000"
      options:
        - value: "24000"
          label:
            en_US: "24000 Hz"
            zh_Hans: "24000 Hz"
        - value: "16000"
          label:
            en_US: "16000 Hz"
            zh_Hans: "16000 Hz"
        - value: "8000"
          label:
            en_US: "8000 Hz"
            zh_Hans: "8000 Hz"
      help:
        en_US: "Audio sample rate for BytePlus TTS output."
        zh_Hans: "BytePlus TTS 输出的音频采样率。"
    - variable: document_prefix
      label:
        zh_Hans: 文档前缀
        en_US: Document prefix
      required: false
      show_on:
        - variable: __model_type
          value: text-embedding
      type: text-input
      default: ""
      placeholder:
        zh_Hans: "输入您的文档前缀, 例如 document:"
        en_US: "Enter your Document prefix, e.g. document:"
    - variable: query_prefix
      label:
        zh_Hans: 查询前缀
        en_US: Query prefix
      required: false
      show_on:
        - variable: __model_type
          value: text-embedding
      type: text-input
      default: ""
      placeholder:
        zh_Hans: "请输入您的查询前缀, 例如 query:"
        en_US: "Enter your Query prefix, e.g. query:"
extra:
  python:
    provider_source: provider/openai_api_compatible.py
    model_sources:
      - "models/llm/llm.py"
      - "models/text_embedding/text_embedding.py"
      - "models/rerank/rerank.py"
      - "models/speech2text/speech2text.py"
      - "models/tts/tts.py"