feat: add OpenAI protocol support and enhance AI provider configuration

domain/ai/__init__.py

@@ -22,8 +22,11 @@ from .types import (
     AIModelUpdate,
     AIProviderCreate,
     AIProviderUpdate,
+    OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    OPENAI_PROTOCOL_RESPONSES,
     VectorDBConfigPayload,
     normalize_capabilities,
+    normalize_openai_protocol,
 )
 from .vector_providers import (
     BaseVectorProvider,
@@ -58,6 +61,9 @@ __all__ = [
     "get_provider_class",
     "ABILITIES",
     "normalize_capabilities",
+    "normalize_openai_protocol",
+    "OPENAI_PROTOCOL_CHAT_COMPLETIONS",
+    "OPENAI_PROTOCOL_RESPONSES",
     "AIDefaultsUpdate",
     "AIModelCreate",
     "AIModelUpdate",

domain/ai/api.py

@@ -34,7 +34,7 @@ async def list_providers_endpoint(
 @audit(
     action=AuditAction.CREATE,
     description="创建 AI 提供商",
-    body_fields=["name", "identifier", "provider_type", "api_format", "base_url", "logo_url"],
+    body_fields=["name", "identifier", "provider_type", "api_format", "base_url", "logo_url", "extra_config"],
     redact_fields=["api_key"],
 )
 @router_ai.post("/providers")
@@ -61,7 +61,7 @@ async def get_provider(
 @audit(
     action=AuditAction.UPDATE,
     description="更新 AI 提供商",
-    body_fields=["name", "provider_type", "api_format", "base_url", "logo_url", "api_key"],
+    body_fields=["name", "provider_type", "api_format", "base_url", "logo_url", "api_key", "extra_config"],
     redact_fields=["api_key"],
 )
 @router_ai.put("/providers/{provider_id}")

domain/ai/inference.py

@@ -6,6 +6,7 @@ from urllib.parse import parse_qsl, urlencode, urlsplit, urlunsplit
 
 from models.database import AIModel, AIProvider
 from .service import AIProviderService
+from .types import OPENAI_PROTOCOL_RESPONSES, normalize_openai_protocol
 
 
 provider_service = AIProviderService
@@ -227,6 +228,248 @@ def _is_azure_openai(provider: AIProvider) -> bool:
     return ".openai.azure.com" in base_url
 
 
+def _openai_protocol(provider: AIProvider) -> str:
+    extra = provider.extra_config if isinstance(provider.extra_config, dict) else {}
+    return normalize_openai_protocol(extra.get("openai_protocol"))
+
+
+def _content_to_text(content: Any) -> str:
+    if content is None:
+        return ""
+    if isinstance(content, str):
+        return content
+    if isinstance(content, list):
+        parts: List[str] = []
+        for part in content:
+            if not isinstance(part, dict):
+                continue
+            text = part.get("text")
+            if isinstance(text, str):
+                parts.append(text)
+        return "".join(parts)
+    try:
+        return json.dumps(content, ensure_ascii=False)
+    except TypeError:
+        return str(content)
+
+
+def _openai_content_to_responses_input(content: Any) -> Any:
+    if content is None:
+        return ""
+    if isinstance(content, str):
+        return content
+    if not isinstance(content, list):
+        return _content_to_text(content)
+
+    blocks: List[Dict[str, Any]] = []
+    for part in content:
+        if not isinstance(part, dict):
+            continue
+        part_type = part.get("type")
+        if part_type in {"text", "input_text"} and isinstance(part.get("text"), str):
+            blocks.append({"type": "input_text", "text": part["text"]})
+            continue
+        if part_type == "image_url":
+            image_url = part.get("image_url")
+            url = image_url.get("url") if isinstance(image_url, dict) else image_url
+            if not isinstance(url, str) or not url.strip():
+                continue
+            block: Dict[str, Any] = {"type": "input_image", "image_url": url}
+            detail = image_url.get("detail") if isinstance(image_url, dict) else None
+            if isinstance(detail, str) and detail.strip():
+                block["detail"] = detail
+            blocks.append(block)
+            continue
+        if part_type == "input_image" and isinstance(part.get("image_url"), str):
+            block = {"type": "input_image", "image_url": part["image_url"]}
+            detail = part.get("detail")
+            if isinstance(detail, str) and detail.strip():
+                block["detail"] = detail
+            blocks.append(block)
+            continue
+        if part_type == "input_file" and isinstance(part.get("file_id") or part.get("filename"), str):
+            blocks.append(dict(part))
+
+    return blocks or ""
+
+
+def _openai_tool_call_to_responses_item(call: Dict[str, Any], idx: int) -> Dict[str, Any] | None:
+    fn = call.get("function")
+    fn = fn if isinstance(fn, dict) else {}
+    name = fn.get("name")
+    if not isinstance(name, str) or not name.strip():
+        return None
+
+    raw_args = fn.get("arguments")
+    if isinstance(raw_args, dict):
+        args_text = json.dumps(raw_args, ensure_ascii=False)
+    elif isinstance(raw_args, str):
+        args_text = raw_args
+    else:
+        args_text = ""
+
+    return {
+        "type": "function_call",
+        "call_id": str(call.get("id") or f"call_{idx}"),
+        "name": name,
+        "arguments": args_text,
+    }
+
+
+def _openai_messages_to_responses_input(messages: List[Dict[str, Any]]) -> Tuple[str, List[Dict[str, Any]]]:
+    instructions: List[str] = []
+    input_items: List[Dict[str, Any]] = []
+
+    for msg in messages:
+        if not isinstance(msg, dict):
+            continue
+        role = msg.get("role")
+        if role in {"system", "developer"}:
+            text = _content_to_text(msg.get("content")).strip()
+            if text:
+                instructions.append(text)
+            continue
+
+        if role == "tool":
+            content = msg.get("content")
+            output = content if isinstance(content, str) else _content_to_text(content)
+            call_id = str(msg.get("tool_call_id") or "").strip()
+            if call_id:
+                input_items.append({
+                    "type": "function_call_output",
+                    "call_id": call_id,
+                    "output": output,
+                })
+            elif output:
+                input_items.append({"role": "user", "content": output})
+            continue
+
+        if role == "user":
+            input_items.append({
+                "role": "user",
+                "content": _openai_content_to_responses_input(msg.get("content")),
+            })
+            continue
+
+        if role == "assistant":
+            text = _content_to_text(msg.get("content"))
+            if text:
+                input_items.append({"role": "assistant", "content": text})
+            tool_calls = msg.get("tool_calls")
+            if isinstance(tool_calls, list):
+                for idx, call in enumerate(tool_calls):
+                    if not isinstance(call, dict):
+                        continue
+                    item = _openai_tool_call_to_responses_item(call, idx)
+                    if item:
+                        input_items.append(item)
+
+    if not input_items:
+        input_items.append({"role": "user", "content": ""})
+
+    return "\n\n".join(instructions).strip(), input_items
+
+
+def _openai_tools_to_responses(tools: List[Dict[str, Any]] | None) -> List[Dict[str, Any]] | None:
+    if not tools:
+        return None
+    out: List[Dict[str, Any]] = []
+    for tool in tools:
+        if not isinstance(tool, dict) or tool.get("type") != "function":
+            continue
+        fn = tool.get("function")
+        fn = fn if isinstance(fn, dict) else {}
+        name = fn.get("name")
+        if not isinstance(name, str) or not name.strip():
+            continue
+        entry: Dict[str, Any] = {
+            "type": "function",
+            "name": name,
+            "description": fn.get("description") if isinstance(fn.get("description"), str) else "",
+            "parameters": fn.get("parameters") if isinstance(fn.get("parameters"), dict) else {},
+        }
+        if isinstance(fn.get("strict"), bool):
+            entry["strict"] = fn["strict"]
+        out.append(entry)
+    return out or None
+
+
+def _openai_tool_choice_to_responses(tool_choice: Any) -> Any | None:
+    if tool_choice is None:
+        return None
+    if isinstance(tool_choice, str):
+        return tool_choice
+    if not isinstance(tool_choice, dict):
+        return None
+    if tool_choice.get("type") == "function":
+        fn = tool_choice.get("function")
+        name = fn.get("name") if isinstance(fn, dict) else tool_choice.get("name")
+        if isinstance(name, str) and name.strip():
+            return {"type": "function", "name": name}
+    return None
+
+
+def _responses_function_call_to_openai(item: Dict[str, Any], idx: int) -> Dict[str, Any] | None:
+    name = item.get("name")
+    if not isinstance(name, str) or not name.strip():
+        return None
+    raw_args = item.get("arguments")
+    if isinstance(raw_args, dict):
+        args_text = json.dumps(raw_args, ensure_ascii=False)
+    elif isinstance(raw_args, str):
+        args_text = raw_args
+    else:
+        args_text = ""
+    return {
+        "id": str(item.get("call_id") or item.get("id") or f"call_{idx}"),
+        "type": "function",
+        "function": {"name": name, "arguments": args_text},
+    }
+
+
+def _responses_content_to_text(content: Any) -> str:
+    if isinstance(content, str):
+        return content
+    if not isinstance(content, list):
+        return ""
+    parts: List[str] = []
+    for block in content:
+        if not isinstance(block, dict):
+            continue
+        if block.get("type") in {"output_text", "text"} and isinstance(block.get("text"), str):
+            parts.append(block["text"])
+    return "".join(parts)
+
+
+def _responses_body_to_openai_message(body: Dict[str, Any]) -> Dict[str, Any]:
+    text_parts: List[str] = []
+    tool_calls: List[Dict[str, Any]] = []
+    output = body.get("output")
+
+    if isinstance(output, list):
+        for idx, item in enumerate(output):
+            if not isinstance(item, dict):
+                continue
+            item_type = item.get("type")
+            if item_type == "message":
+                text = _responses_content_to_text(item.get("content"))
+                if text:
+                    text_parts.append(text)
+                continue
+            if item_type == "function_call":
+                tool_call = _responses_function_call_to_openai(item, idx)
+                if tool_call:
+                    tool_calls.append(tool_call)
+
+    if not text_parts and isinstance(body.get("output_text"), str):
+        text_parts.append(body["output_text"])
+
+    message: Dict[str, Any] = {"role": "assistant", "content": "".join(text_parts)}
+    if tool_calls:
+        message["tool_calls"] = tool_calls
+    return message
+
+
 def _gemini_endpoint(provider: AIProvider, path: str) -> str:
     base = (provider.base_url or "").rstrip("/")
     if not base:
@@ -808,6 +1051,12 @@ async def _chat_stream_with_ollama(
 
 
 async def _describe_with_openai(provider: AIProvider, model: AIModel, base64_image: str, detail: str) -> str:
+    if _openai_protocol(provider) == OPENAI_PROTOCOL_RESPONSES:
+        return await _describe_with_openai_responses(provider, model, base64_image, detail)
+    return await _describe_with_openai_chat_completions(provider, model, base64_image, detail)
+
+
+async def _describe_with_openai_chat_completions(provider: AIProvider, model: AIModel, base64_image: str, detail: str) -> str:
     url = _openai_endpoint(provider, "/chat/completions")
     payload = {
         "model": model.name,
@@ -834,6 +1083,32 @@ async def _describe_with_openai(provider: AIProvider, model: AIModel, base64_ima
         return body["choices"][0]["message"]["content"]
 
 
+async def _describe_with_openai_responses(provider: AIProvider, model: AIModel, base64_image: str, detail: str) -> str:
+    url = _openai_endpoint(provider, "/responses")
+    payload = {
+        "model": model.name,
+        "input": [
+            {
+                "role": "user",
+                "content": [
+                    {
+                        "type": "input_image",
+                        "image_url": f"data:image/jpeg;base64,{base64_image}",
+                        "detail": detail,
+                    },
+                    {"type": "input_text", "text": "描述这个图片"},
+                ],
+            }
+        ],
+    }
+    async with httpx.AsyncClient(timeout=60.0) as client:
+        response = await client.post(url, headers=_openai_headers(provider), json=payload)
+        response.raise_for_status()
+        body = response.json()
+        message = _responses_body_to_openai_message(body if isinstance(body, dict) else {})
+        return str(message.get("content") or "")
+
+
 async def _describe_with_anthropic(provider: AIProvider, model: AIModel, base64_image: str, detail: str) -> str:
     url = _anthropic_endpoint(provider, "/messages")
     detail_text = f"描述这个图片，细节等级：{detail}"
@@ -1080,6 +1355,37 @@ async def _chat_with_openai(
     tool_choice: Any | None,
     temperature: float | None,
     timeout: float,
+) -> Dict[str, Any]:
+    if _openai_protocol(provider) == OPENAI_PROTOCOL_RESPONSES:
+        return await _chat_with_openai_responses(
+            provider,
+            model,
+            messages,
+            tools=tools,
+            tool_choice=tool_choice,
+            temperature=temperature,
+            timeout=timeout,
+        )
+    return await _chat_with_openai_chat_completions(
+        provider,
+        model,
+        messages,
+        tools=tools,
+        tool_choice=tool_choice,
+        temperature=temperature,
+        timeout=timeout,
+    )
+
+
+async def _chat_with_openai_chat_completions(
+    provider: AIProvider,
+    model: AIModel,
+    messages: List[Dict[str, Any]],
+    *,
+    tools: List[Dict[str, Any]] | None,
+    tool_choice: Any | None,
+    temperature: float | None,
+    timeout: float,
 ) -> Dict[str, Any]:
     url = _openai_endpoint(provider, "/chat/completions")
     payload: Dict[str, Any] = {
@@ -1106,6 +1412,41 @@ async def _chat_with_openai(
     return message
 
 
+async def _chat_with_openai_responses(
+    provider: AIProvider,
+    model: AIModel,
+    messages: List[Dict[str, Any]],
+    *,
+    tools: List[Dict[str, Any]] | None,
+    tool_choice: Any | None,
+    temperature: float | None,
+    timeout: float,
+) -> Dict[str, Any]:
+    url = _openai_endpoint(provider, "/responses")
+    instructions, input_items = _openai_messages_to_responses_input(messages)
+    payload: Dict[str, Any] = {
+        "model": model.name,
+        "input": input_items,
+    }
+    if instructions:
+        payload["instructions"] = instructions
+    response_tools = _openai_tools_to_responses(tools)
+    if response_tools:
+        payload["tools"] = response_tools
+        payload["tool_choice"] = _openai_tool_choice_to_responses(tool_choice) or "auto"
+    if temperature is not None:
+        payload["temperature"] = float(temperature)
+
+    async with httpx.AsyncClient(timeout=timeout) as client:
+        response = await client.post(url, headers=_openai_headers(provider), json=payload)
+        response.raise_for_status()
+        body = response.json()
+
+    if not isinstance(body, dict):
+        raise RuntimeError("Responses 接口返回格式异常")
+    return _responses_body_to_openai_message(body)
+
+
 async def chat_completion_stream(
     messages: List[Dict[str, Any]],
     *,
@@ -1174,6 +1515,40 @@ async def _chat_stream_with_openai(
     tool_choice: Any | None,
     temperature: float | None,
     timeout: float,
+) -> AsyncIterator[Dict[str, Any]]:
+    if _openai_protocol(provider) == OPENAI_PROTOCOL_RESPONSES:
+        async for event in _chat_stream_with_openai_responses(
+            provider,
+            model,
+            messages,
+            tools=tools,
+            tool_choice=tool_choice,
+            temperature=temperature,
+            timeout=timeout,
+        ):
+            yield event
+        return
+    async for event in _chat_stream_with_openai_chat_completions(
+        provider,
+        model,
+        messages,
+        tools=tools,
+        tool_choice=tool_choice,
+        temperature=temperature,
+        timeout=timeout,
+    ):
+        yield event
+
+
+async def _chat_stream_with_openai_chat_completions(
+    provider: AIProvider,
+    model: AIModel,
+    messages: List[Dict[str, Any]],
+    *,
+    tools: List[Dict[str, Any]] | None,
+    tool_choice: Any | None,
+    temperature: float | None,
+    timeout: float,
 ) -> AsyncIterator[Dict[str, Any]]:
     url = _openai_endpoint(provider, "/chat/completions")
     payload: Dict[str, Any] = {
@@ -1273,3 +1648,100 @@ async def _chat_stream_with_openai(
         message["tool_calls"] = tool_calls
 
     yield {"type": "message", "message": message, "finish_reason": finish_reason}
+
+
+async def _chat_stream_with_openai_responses(
+    provider: AIProvider,
+    model: AIModel,
+    messages: List[Dict[str, Any]],
+    *,
+    tools: List[Dict[str, Any]] | None,
+    tool_choice: Any | None,
+    temperature: float | None,
+    timeout: float,
+) -> AsyncIterator[Dict[str, Any]]:
+    url = _openai_endpoint(provider, "/responses")
+    instructions, input_items = _openai_messages_to_responses_input(messages)
+    payload: Dict[str, Any] = {
+        "model": model.name,
+        "input": input_items,
+        "stream": True,
+    }
+    if instructions:
+        payload["instructions"] = instructions
+    response_tools = _openai_tools_to_responses(tools)
+    if response_tools:
+        payload["tools"] = response_tools
+        payload["tool_choice"] = _openai_tool_choice_to_responses(tool_choice) or "auto"
+    if temperature is not None:
+        payload["temperature"] = float(temperature)
+
+    content_parts: List[str] = []
+    output_items: Dict[int, Dict[str, Any]] = {}
+    completed_body: Dict[str, Any] | None = None
+    current_event: str | None = None
+
+    async with httpx.AsyncClient(timeout=timeout) as client:
+        async with client.stream("POST", url, headers=_openai_headers(provider), json=payload) as response:
+            response.raise_for_status()
+            async for line in response.aiter_lines():
+                if not line:
+                    continue
+                if line.startswith("event:"):
+                    current_event = line[6:].strip()
+                    continue
+                if not line.startswith("data:"):
+                    continue
+                data = line[5:].strip()
+                if not data or data == "[DONE]":
+                    continue
+                try:
+                    chunk = json.loads(data)
+                except json.JSONDecodeError:
+                    continue
+                if not isinstance(chunk, dict):
+                    continue
+
+                event_type = chunk.get("type") or current_event
+                if event_type == "response.output_text.delta":
+                    delta = chunk.get("delta")
+                    if isinstance(delta, str) and delta:
+                        content_parts.append(delta)
+                        yield {"type": "delta", "delta": delta}
+                    continue
+
+                if event_type in {"response.output_item.added", "response.output_item.done"}:
+                    item = chunk.get("item")
+                    idx = chunk.get("output_index")
+                    if isinstance(item, dict) and isinstance(idx, int):
+                        output_items[idx] = item
+                    continue
+
+                if event_type == "response.completed":
+                    response_body = chunk.get("response")
+                    if isinstance(response_body, dict):
+                        completed_body = response_body
+                    elif isinstance(chunk.get("output"), list):
+                        completed_body = chunk
+                    break
+
+                if event_type == "error":
+                    error = chunk.get("error") if isinstance(chunk.get("error"), dict) else chunk
+                    raise RuntimeError(str(error.get("message") if isinstance(error, dict) else error))
+
+    if completed_body is not None:
+        message = _responses_body_to_openai_message(completed_body)
+        if not message.get("content") and content_parts:
+            message["content"] = "".join(content_parts)
+        yield {"type": "message", "message": message, "finish_reason": None}
+        return
+
+    if output_items:
+        body = {"output": [output_items[idx] for idx in sorted(output_items.keys())]}
+        message = _responses_body_to_openai_message(body)
+        if not message.get("content") and content_parts:
+            message["content"] = "".join(content_parts)
+        yield {"type": "message", "message": message, "finish_reason": None}
+        return
+
+    yield {"type": "message", "message": {"role": "assistant", "content": "".join(content_parts)}, "finish_reason": None}

domain/ai/types.py

@@ -3,6 +3,19 @@ from typing import Any, Dict, Iterable, List, Optional
 from pydantic import BaseModel, Field, field_validator
 
 ABILITIES = ["chat", "vision", "embedding", "rerank", "voice", "tools"]
+OPENAI_PROTOCOL_CHAT_COMPLETIONS = "chat_completions"
+OPENAI_PROTOCOL_RESPONSES = "responses"
+OPENAI_PROTOCOLS = {OPENAI_PROTOCOL_CHAT_COMPLETIONS, OPENAI_PROTOCOL_RESPONSES}
+OPENAI_PROTOCOL_ALIASES = {
+    "chat": OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    "chat_completion": OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    "chat_completions": OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    "chat/completions": OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    "/chat/completions": OPENAI_PROTOCOL_CHAT_COMPLETIONS,
+    "response": OPENAI_PROTOCOL_RESPONSES,
+    "responses": OPENAI_PROTOCOL_RESPONSES,
+    "/responses": OPENAI_PROTOCOL_RESPONSES,
+}
 
 
 def normalize_capabilities(items: Optional[Iterable[str]]) -> List[str]:
@@ -16,6 +29,34 @@ def normalize_capabilities(items: Optional[Iterable[str]]) -> List[str]:
     return normalized
 
 
+def normalize_openai_protocol(value: Any) -> str:
+    if value is None:
+        return OPENAI_PROTOCOL_CHAT_COMPLETIONS
+    key = str(value).strip().lower().replace("-", "_").replace(".", "_")
+    if not key:
+        return OPENAI_PROTOCOL_CHAT_COMPLETIONS
+    normalized = OPENAI_PROTOCOL_ALIASES.get(key)
+    if normalized:
+        return normalized
+    normalized = OPENAI_PROTOCOL_ALIASES.get(key.replace("_", "/"))
+    if normalized:
+        return normalized
+    if key in OPENAI_PROTOCOLS:
+        return key
+    raise ValueError("openai_protocol must be 'chat_completions' or 'responses'")
+
+
+def normalize_provider_extra_config(config: Optional[dict]) -> Optional[dict]:
+    if config is None:
+        return None
+    if not isinstance(config, dict):
+        raise ValueError("extra_config must be an object")
+    normalized = dict(config)
+    if "openai_protocol" in normalized:
+        normalized["openai_protocol"] = normalize_openai_protocol(normalized.get("openai_protocol"))
+    return normalized
+
+
 class AIProviderBase(BaseModel):
     name: str
     identifier: str = Field(..., pattern=r"^[a-z0-9_\-\.]+$")
@@ -34,6 +75,11 @@ class AIProviderBase(BaseModel):
             raise ValueError("api_format must be 'openai', 'gemini', 'anthropic', or 'ollama'")
         return fmt
 
+    @field_validator("extra_config")
+    @classmethod
+    def normalize_extra_config(cls, value: Optional[dict]) -> Optional[dict]:
+        return normalize_provider_extra_config(value)
+
 
 class AIProviderCreate(AIProviderBase):
     pass
@@ -58,6 +104,11 @@ class AIProviderUpdate(BaseModel):
             raise ValueError("api_format must be 'openai', 'gemini', 'anthropic', or 'ollama'")
         return fmt
 
+    @field_validator("extra_config")
+    @classmethod
+    def normalize_extra_config(cls, value: Optional[dict]) -> Optional[dict]:
+        return normalize_provider_extra_config(value)
+
 
 class AIModelBase(BaseModel):
     name: str

web/src/api/aiProviders.ts

@@ -1,6 +1,7 @@
 import request from './client';
 
 export type AIAbility = 'chat' | 'vision' | 'embedding' | 'rerank' | 'voice' | 'tools';
+export type OpenAIProtocol = 'chat_completions' | 'responses';
 
 export interface AIProviderPayload {
   name: string;

web/src/i18n/locales/en.json

@@ -516,6 +516,7 @@
   "Enter identifier": "Enter identifier",
   "Only lowercase letters, numbers, dash, dot and underscore are allowed": "Only lowercase letters, numbers, dash, dot and underscore are allowed",
   "API Format": "API Format",
+  "OpenAI Protocol": "OpenAI Protocol",
   "Base URL": "Base URL",
   "Enter base url": "Enter base URL",
   "Optional, can also be provided per request": "Optional, can also be provided per request",

web/src/i18n/locales/zh.json

@@ -515,6 +515,7 @@
   "Enter identifier": "请输入标识符",
   "Only lowercase letters, numbers, dash, dot and underscore are allowed": "仅允许小写字母、数字、连字符、点和下划线",
   "API Format": "API 格式",
+  "OpenAI Protocol": "OpenAI 协议",
   "Base URL": "基础 URL",
   "Enter base url": "请输入基础 URL",
   "Optional, can also be provided per request": "可选，也可在请求时提供",

web/src/pages/SystemSettingsPage/components/AiSettingsTab.tsx

@@ -48,6 +48,7 @@ import type {
   AIModelPayload,
   AIProvider,
   AIProviderPayload,
+  OpenAIProtocol,
 } from '../../../api/aiProviders';
 import {
   createModel,
@@ -90,6 +91,11 @@ interface ProviderTemplate {
 }
 
 const abilityOrder: AIAbility[] = ['chat', 'vision', 'embedding', 'rerank', 'voice', 'tools'];
+const defaultOpenAIProtocol: OpenAIProtocol = 'chat_completions';
+
+function normalizeOpenAIProtocol(value: unknown): OpenAIProtocol {
+  return value === 'responses' ? 'responses' : defaultOpenAIProtocol;
+}
 
 const abilityInfo: Record<AIAbility, { icon: ReactNode; label: string; color: string; description: string }> = {
   chat: {
@@ -242,6 +248,7 @@ type AIProviderFormValues = {
   name?: string;
   identifier?: string;
   api_format: AIProviderPayload['api_format'];
+  openai_protocol?: OpenAIProtocol;
   base_url?: string;
   api_key?: string;
   logo_url?: string;
@@ -275,12 +282,19 @@ export default function AiSettingsTab() {
   const [addingRemoteModels, setAddingRemoteModels] = useState<boolean>(false);
   const [modelModalTab, setModelModalTab] = useState<'remote' | 'manual'>('remote');
   const [remoteSearchKeyword, setRemoteSearchKeyword] = useState<string>('');
+  const providerApiFormat = Form.useWatch('api_format', providerForm);
   const capabilitiesValue = Form.useWatch('capabilities', modelForm);
   const showEmbeddingDimensions = useMemo(() => {
     const capabilities = Array.isArray(capabilitiesValue) ? capabilitiesValue : [];
     return capabilities.includes('embedding') || capabilities.includes('rerank');
   }, [capabilitiesValue]);
 
+  useEffect(() => {
+    if (providerApiFormat === 'openai' && !providerForm.getFieldValue('openai_protocol')) {
+      providerForm.setFieldValue('openai_protocol', defaultOpenAIProtocol);
+    }
+  }, [providerApiFormat, providerForm]);
+
   useEffect(() => {
     if (!showEmbeddingDimensions) {
       modelForm.setFieldsValue({ embedding_dimensions: null });
@@ -338,6 +352,7 @@ export default function AiSettingsTab() {
         name: existing.name,
         identifier: existing.identifier,
         api_format: existing.api_format,
+        openai_protocol: normalizeOpenAIProtocol(existing.extra_config?.openai_protocol),
         base_url: existing.base_url ?? undefined,
         api_key: '',
         logo_url: existing.logo_url ?? undefined,
@@ -345,7 +360,7 @@ export default function AiSettingsTab() {
       });
     } else {
       providerForm.resetFields();
-      providerForm.setFieldsValue({ api_format: 'openai' });
+      providerForm.setFieldsValue({ api_format: 'openai', openai_protocol: defaultOpenAIProtocol });
       setSelectedTemplate(null);
       setProviderModal({ open: true, step: 1 });
     }
@@ -364,6 +379,7 @@ export default function AiSettingsTab() {
       name: t(template.nameKey),
       identifier: template.identifier,
       api_format: template.api_format,
+      openai_protocol: template.api_format === 'openai' ? defaultOpenAIProtocol : undefined,
       base_url: template.base_url ?? '',
       api_key: '',
       logo_url: template.logo_url ?? '',
@@ -375,7 +391,7 @@ export default function AiSettingsTab() {
     setProviderModal((prev) => ({ ...prev, step: 1, editing: undefined }));
     setSelectedTemplate(null);
     providerForm.resetFields();
-    providerForm.setFieldsValue({ api_format: 'openai' });
+    providerForm.setFieldsValue({ api_format: 'openai', openai_protocol: defaultOpenAIProtocol });
   };
 
   const handleSubmitProvider = async () => {
@@ -384,6 +400,12 @@ export default function AiSettingsTab() {
     const trimmedApiKey = values.api_key?.trim();
     const trimmedLogoUrl = values.logo_url?.trim();
     const trimmedProviderType = values.provider_type?.trim();
+    const extraConfig = { ...(providerModal.editing?.extra_config ?? {}) };
+    if (values.api_format === 'openai') {
+      extraConfig.openai_protocol = normalizeOpenAIProtocol(values.openai_protocol);
+    } else {
+      delete extraConfig.openai_protocol;
+    }
     const payload: AIProviderPayload = {
       name: (values.name || '').trim(),
       identifier: (values.identifier || '').trim(),
@@ -391,6 +413,7 @@ export default function AiSettingsTab() {
       base_url: trimmedBaseUrl ? trimmedBaseUrl : null,
       logo_url: trimmedLogoUrl ? trimmedLogoUrl : null,
       provider_type: trimmedProviderType ? trimmedProviderType : null,
+      extra_config: Object.keys(extraConfig).length ? extraConfig : null,
     };
     if (trimmedApiKey) {
       payload.api_key = trimmedApiKey;
@@ -1117,16 +1140,30 @@ export default function AiSettingsTab() {
                   label={t('API Format')}
                   rules={[{ required: true }]}
                 >
-                <Select
-                  disabled={!allowFormatChange}
-                  options={[
-                    { value: 'openai', label: 'OpenAI Compatible' },
-                    { value: 'gemini', label: 'Gemini Compatible' },
-                    { value: 'anthropic', label: 'Anthropic Native' },
-                    { value: 'ollama', label: 'Ollama Native' },
-                  ]}
-                />
-              </Form.Item>
+                  <Select
+                    disabled={!allowFormatChange}
+                    options={[
+                      { value: 'openai', label: 'OpenAI Compatible' },
+                      { value: 'gemini', label: 'Gemini Compatible' },
+                      { value: 'anthropic', label: 'Anthropic Native' },
+                      { value: 'ollama', label: 'Ollama Native' },
+                    ]}
+                  />
+                </Form.Item>
+                {providerApiFormat === 'openai' ? (
+                  <Form.Item
+                    name="openai_protocol"
+                    label={t('OpenAI Protocol')}
+                    rules={[{ required: true }]}
+                  >
+                    <Select
+                      options={[
+                        { value: 'chat_completions', label: 'Chat Completions' },
+                        { value: 'responses', label: 'Responses' },
+                      ]}
+                    />
+                  </Form.Item>
+                ) : null}
                 <Form.Item name="base_url" label={t('Base URL')} rules={[{ required: true, message: t('Enter base url') }]}>
                   <Input placeholder="https://" />
                 </Form.Item>