feat(studio): auto-detect working provider transport settings

packages/cli/src/commands/doctor.ts

@@ -8,6 +8,33 @@ import {
   inspectNodeRuntimePinFiles,
 } from "../runtime-requirements.js";
 
+function buildDoctorProbePlans(
+  preferredApiFormat: "chat" | "responses" | undefined,
+  preferredStream: boolean | undefined,
+): Array<{ apiFormat: "chat" | "responses"; stream: boolean }> {
+  const plans: Array<{ apiFormat: "chat" | "responses"; stream: boolean }> = [];
+  const seen = new Set<string>();
+  const push = (apiFormat: "chat" | "responses", stream: boolean) => {
+    const key = `${apiFormat}:${stream ? "1" : "0"}`;
+    if (seen.has(key)) return;
+    seen.add(key);
+    plans.push({ apiFormat, stream });
+  };
+
+  if (preferredApiFormat) {
+    push(preferredApiFormat, preferredStream ?? false);
+    push(preferredApiFormat, !(preferredStream ?? false));
+  }
+  const alternate = preferredApiFormat === "responses" ? "chat" : "responses";
+  push(alternate, false);
+  push(alternate, true);
+  push("chat", false);
+  push("chat", true);
+  push("responses", false);
+  push("responses", true);
+  return plans;
+}
+
 export const doctorCommand = new Command("doctor")
   .description("Check environment and project health")
   .option("--repair-node-runtime", "Write .nvmrc and .node-version pinned to Node 22 for this project")
@@ -188,17 +215,47 @@ export const doctorCommand = new Command("doctor")
           detail: `provider=${llmConfig.provider} model=${llmConfig.model} stream=${llmConfig.stream ?? true} baseUrl=${llmConfig.baseUrl}`,
         });
 
-        const client = createLLMClient(llmConfig);
         log("\n  [..] Testing API connectivity...");
-        const response = await chatCompletion(client, llmConfig.model, [
-          { role: "user", content: "Say OK" },
-        ], { maxTokens: 16 });
+
+        let connected = false;
+        let detectedDetail = "";
+        let lastError = "Unknown error";
+        const plans = llmConfig.provider === "openai"
+          ? buildDoctorProbePlans(llmConfig.apiFormat, llmConfig.stream)
+          : [{ apiFormat: (llmConfig.apiFormat ?? "chat") as "chat" | "responses", stream: llmConfig.stream ?? true }];
+
+        for (const plan of plans) {
+          try {
+            const client = createLLMClient({
+              ...llmConfig,
+              apiFormat: plan.apiFormat,
+              stream: plan.stream,
+            });
+            const response = await chatCompletion(client, llmConfig.model, [
+              { role: "user", content: "Say OK" },
+            ], { maxTokens: 16 });
+
+            connected = true;
+            detectedDetail = `OK (model: ${llmConfig.model}, apiFormat=${plan.apiFormat}, stream=${plan.stream}, tokens: ${response.usage.totalTokens})`;
+            break;
+          } catch (error) {
+            lastError = error instanceof Error ? error.message : String(error);
+          }
+        }
 
         checks.push({
           name: "API Connectivity",
-          ok: true,
-          detail: `OK (model: ${llmConfig.model}, tokens: ${response.usage.totalTokens})`,
+          ok: connected,
+          detail: connected ? detectedDetail : lastError.split("\n")[0]!,
         });
+
+        if (!connected && llmConfig.provider === "openai") {
+          checks.push({
+            name: "  Hint",
+            ok: false,
+            detail: "当前已自动尝试 chat/responses 与流式开关组合；如果仍失败，问题更可能在模型名、baseUrl 路径或服务商兼容性本身。",
+          });
+        }
       }
     } catch (e) {
       const errMsg = String(e);

packages/studio/src/api/server.test.ts

@@ -459,6 +459,47 @@ describe("createStudioServer daemon lifecycle", () => {
     );
   });
 
+  it("auto-falls back to a non-stream probe in doctor checks when the first transport returns empty", async () => {
+    const freshConfig = {
+      ...cloneProjectConfig(),
+      llm: {
+        ...cloneProjectConfig().llm,
+        model: "claude-sonnet-4-6",
+        baseUrl: "https://timesniper.club",
+        stream: true,
+        apiFormat: "chat",
+      },
+    };
+    loadProjectConfigMock.mockResolvedValue(freshConfig);
+    createLLMClientMock.mockImplementation(((cfg: unknown) => cfg) as any);
+    chatCompletionMock.mockImplementation(async (client: any) => {
+      if (client.stream === false) {
+        return {
+          content: "pong",
+          usage: { promptTokens: 1, completionTokens: 1, totalTokens: 2 },
+        };
+      }
+      throw new Error("LLM returned empty response from stream");
+    });
+
+    const { createStudioServer } = await import("./server.js");
+    const app = createStudioServer(freshConfig as never, root);
+
+    const response = await app.request("http://localhost/api/v1/doctor");
+    expect(response.status).toBe(200);
+    await expect(response.json()).resolves.toMatchObject({
+      llmConnected: true,
+    });
+    expect(createLLMClientMock).toHaveBeenCalledWith(expect.objectContaining({
+      stream: true,
+      apiFormat: "chat",
+    }));
+    expect(createLLMClientMock).toHaveBeenCalledWith(expect.objectContaining({
+      stream: false,
+      apiFormat: "chat",
+    }));
+  });
+
   it("reloads latest llm config for radar scans without restarting the studio server", async () => {
     const startupConfig = {
       ...cloneProjectConfig(),
@@ -682,6 +723,102 @@ describe("createStudioServer daemon lifecycle", () => {
     });
   });
 
+  it("auto-detects a working custom combination when /models is unavailable", async () => {
+    await writeFile(join(root, "inkos.json"), JSON.stringify({
+      ...projectConfig,
+      llm: {
+        configSource: "env",
+        services: [
+          { service: "custom", name: "MiniMax", baseUrl: "https://api.minimax.com/v1" },
+        ],
+      },
+    }, null, 2), "utf-8");
+    await writeFile(join(root, ".env"), [
+      "INKOS_LLM_MODEL=MiniMax-M2.7",
+      "INKOS_LLM_BASE_URL=https://api.minimax.com/v1",
+      "INKOS_LLM_API_KEY=sk-minimax",
+    ].join("\n"), "utf-8");
+
+    createLLMClientMock.mockImplementation(((cfg: unknown) => cfg) as any);
+    chatCompletionMock.mockImplementation(async (client: any) => {
+      if (client.apiFormat === "chat" && client.stream === false) {
+        return {
+          content: "pong",
+          usage: { promptTokens: 1, completionTokens: 1, totalTokens: 2 },
+        };
+      }
+      throw new Error("LLM returned empty response from stream");
+    });
+
+    const fetchMock = vi.fn().mockResolvedValue({
+      ok: false,
+      status: 404,
+      text: async () => "404 page not found",
+    });
+    vi.stubGlobal("fetch", fetchMock as typeof fetch);
+
+    const { createStudioServer } = await import("./server.js");
+    const app = createStudioServer(cloneProjectConfig() as never, root);
+
+    const response = await app.request("http://localhost/api/v1/services/custom%3AMiniMax/test", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({
+        apiKey: "sk-minimax",
+        baseUrl: "https://api.minimax.com/v1",
+        apiFormat: "chat",
+        stream: true,
+      }),
+    });
+
+    expect(response.status).toBe(200);
+    await expect(response.json()).resolves.toMatchObject({
+      ok: true,
+      selectedModel: "MiniMax-M2.7",
+      detected: {
+        apiFormat: "chat",
+        stream: false,
+        modelsSource: "fallback",
+      },
+      models: [{ id: "MiniMax-M2.7", name: "MiniMax-M2.7" }],
+    });
+  });
+
+  it("falls back to the detected/default model when custom /models is unavailable", async () => {
+    await writeFile(join(root, "inkos.json"), JSON.stringify({
+      ...projectConfig,
+      llm: {
+        defaultModel: "MiniMax-M2.7",
+        services: [
+          { service: "custom", name: "MiniMax", baseUrl: "https://api.minimax.com/v1", apiFormat: "chat", stream: false },
+        ],
+      },
+    }, null, 2), "utf-8");
+    getServiceApiKeyMock.mockResolvedValue("sk-minimax");
+
+    const fetchMock = vi.fn().mockResolvedValue({
+      ok: false,
+      status: 404,
+      text: async () => "404 page not found",
+    });
+    vi.stubGlobal("fetch", fetchMock as typeof fetch);
+    createLLMClientMock.mockImplementation(((cfg: unknown) => cfg) as any);
+    chatCompletionMock.mockResolvedValue({
+      content: "pong",
+      usage: { promptTokens: 1, completionTokens: 1, totalTokens: 2 },
+    });
+
+    const { createStudioServer } = await import("./server.js");
+    const app = createStudioServer(cloneProjectConfig() as never, root);
+
+    const response = await app.request("http://localhost/api/v1/services/custom%3AMiniMax/models");
+
+    expect(response.status).toBe(200);
+    await expect(response.json()).resolves.toMatchObject({
+      models: [{ id: "MiniMax-M2.7", name: "MiniMax-M2.7" }],
+    });
+  });
+
   it("returns stored service secret for detail page rehydration", async () => {
     loadSecretsMock.mockResolvedValue({
       services: {

packages/studio/src/api/server.ts

@@ -142,6 +142,17 @@ interface EnvConfigStatus {
   effectiveSource: "project" | "global" | null;
 }
 
+interface ServiceProbeResult {
+  ok: boolean;
+  models: Array<{ id: string; name: string }>;
+  selectedModel?: string;
+  apiFormat?: "chat" | "responses";
+  stream?: boolean;
+  baseUrl?: string;
+  modelsSource?: "api" | "fallback";
+  error?: string;
+}
+
 function broadcast(event: string, data: unknown): void {
   for (const handler of subscribers) {
     handler(event, data);
@@ -311,6 +322,167 @@ async function resolveConfiguredServiceEntry(root: string, serviceId: string): P
   }
 }
 
+function buildProbePlans(
+  preferredApiFormat: "chat" | "responses" | undefined,
+  preferredStream: boolean | undefined,
+): Array<{ apiFormat: "chat" | "responses"; stream: boolean }> {
+  const candidates: Array<{ apiFormat: "chat" | "responses"; stream: boolean }> = [];
+  const seen = new Set<string>();
+  const push = (apiFormat: "chat" | "responses", stream: boolean) => {
+    const key = `${apiFormat}:${stream ? "1" : "0"}`;
+    if (seen.has(key)) return;
+    seen.add(key);
+    candidates.push({ apiFormat, stream });
+  };
+
+  if (preferredApiFormat) {
+    push(preferredApiFormat, preferredStream ?? false);
+    push(preferredApiFormat, !(preferredStream ?? false));
+  }
+  const alternate = preferredApiFormat === "responses" ? "chat" : "responses";
+  push(alternate, false);
+  push(alternate, true);
+  push("chat", false);
+  push("chat", true);
+  push("responses", false);
+  push("responses", true);
+  return candidates;
+}
+
+function buildModelCandidates(args: {
+  preferredModel?: string;
+  configModel?: string;
+  envModel?: string | null;
+  discoveredModels: Array<{ id: string; name: string }>;
+}): string[] {
+  const seen = new Set<string>();
+  const candidates: string[] = [];
+  const push = (value: string | null | undefined) => {
+    if (!value || value.trim().length === 0) return;
+    const id = value.trim();
+    if (seen.has(id)) return;
+    seen.add(id);
+    candidates.push(id);
+  };
+
+  push(args.preferredModel);
+  push(args.configModel);
+  push(args.envModel ?? undefined);
+  for (const model of args.discoveredModels) push(model.id);
+  push("gpt-5.4");
+  push("gpt-4o");
+  push("claude-sonnet-4-6");
+  push("MiniMax-M2.7");
+  push("kimi-k2.5");
+  return candidates;
+}
+
+async function fetchModelsFromServiceBaseUrl(
+  baseUrl: string,
+  apiKey: string,
+): Promise<{ models: Array<{ id: string; name: string }>; error?: string }> {
+  const modelsUrl = baseUrl.replace(/\/$/, "") + "/models";
+  try {
+    const res = await fetch(modelsUrl, {
+      headers: { Authorization: `Bearer ${apiKey}` },
+      signal: AbortSignal.timeout(10_000),
+    });
+    if (!res.ok) {
+      const body = await res.text().catch(() => "");
+      return { models: [], error: `服务商返回 ${res.status}: ${body.slice(0, 200)}` };
+    }
+    const json = await res.json() as { data?: Array<{ id: string }> };
+    return {
+      models: (json.data ?? []).map((m) => ({ id: m.id, name: m.id })),
+    };
+  } catch (error) {
+    return {
+      models: [],
+      error: error instanceof Error ? error.message : String(error),
+    };
+  }
+}
+
+async function probeServiceCapabilities(args: {
+  root: string;
+  service: string;
+  apiKey: string;
+  baseUrl: string;
+  preferredApiFormat?: "chat" | "responses";
+  preferredStream?: boolean;
+  preferredModel?: string;
+}): Promise<ServiceProbeResult> {
+  const rawConfig = await loadRawConfig(args.root).catch(() => ({} as Record<string, unknown>));
+  const llm = (rawConfig.llm as Record<string, unknown> | undefined) ?? {};
+  const envConfig = await readEnvConfigStatus(args.root);
+  const envModel = envConfig.effectiveSource === "project"
+    ? envConfig.project.model
+    : envConfig.effectiveSource === "global"
+      ? envConfig.global.model
+      : null;
+
+  const modelsResponse = await fetchModelsFromServiceBaseUrl(args.baseUrl, args.apiKey);
+  const discoveredModels = modelsResponse.models;
+  const modelCandidates = buildModelCandidates({
+    preferredModel: args.preferredModel,
+    configModel: typeof llm.defaultModel === "string" ? llm.defaultModel : typeof llm.model === "string" ? llm.model : undefined,
+    envModel,
+    discoveredModels,
+  });
+
+  if (modelCandidates.length === 0) {
+    return {
+      ok: false,
+      models: [],
+      error: "无法自动确定模型，请先填写可用模型或提供支持 /models 的服务端点。",
+    };
+  }
+
+  let lastError = modelsResponse.error ?? "自动探测失败";
+
+  for (const model of modelCandidates) {
+    for (const plan of buildProbePlans(args.preferredApiFormat, args.preferredStream)) {
+      const client = createLLMClient({
+        provider: args.service === "anthropic" ? "anthropic" : "openai",
+        service: isCustomServiceId(args.service) ? "custom" : args.service,
+        configSource: "studio",
+        baseUrl: args.baseUrl,
+        apiKey: args.apiKey.trim(),
+        model,
+        temperature: 0.7,
+        maxTokens: 64,
+        thinkingBudget: 0,
+        apiFormat: plan.apiFormat,
+        stream: plan.stream,
+      } as ProjectConfig["llm"]);
+
+      try {
+        await chatCompletion(client, model, [{ role: "user", content: "ping" }], { maxTokens: 5 });
+        const models = discoveredModels.length > 0
+          ? discoveredModels
+          : [{ id: model, name: model }];
+        return {
+          ok: true,
+          models,
+          selectedModel: model,
+          apiFormat: plan.apiFormat,
+          stream: plan.stream,
+          baseUrl: args.baseUrl,
+          modelsSource: discoveredModels.length > 0 ? "api" : "fallback",
+        };
+      } catch (error) {
+        lastError = error instanceof Error ? error.message : String(error);
+      }
+    }
+  }
+
+  return {
+    ok: false,
+    models: discoveredModels,
+    error: lastError,
+  };
+}
+
 // --- Server factory ---
 
 export function createStudioServer(initialConfig: ProjectConfig, root: string) {
@@ -753,7 +925,7 @@ export function createStudioServer(initialConfig: ProjectConfig, root: string) {
   });
 
   app.put("/api/v1/services/config", async (c) => {
-    const body = await c.req.json<{ services?: unknown; defaultModel?: string; configSource?: LLMConfigSource }>();
+    const body = await c.req.json<{ services?: unknown; defaultModel?: string; configSource?: LLMConfigSource; service?: string }>();
     const config = await loadRawConfig(root);
     config.llm = config.llm ?? {};
     const llm = config.llm as Record<string, unknown>;
@@ -768,6 +940,9 @@ export function createStudioServer(initialConfig: ProjectConfig, root: string) {
     if (body.configSource !== undefined) {
       llm.configSource = normalizeConfigSource(body.configSource);
     }
+    if (body.service !== undefined) {
+      llm.service = body.service;
+    }
     await saveRawConfig(root, config);
     return c.json({ ok: true });
   });
@@ -790,71 +965,31 @@ export function createStudioServer(initialConfig: ProjectConfig, root: string) {
       return c.json({ ok: false, error: `未知服务商: ${service}` }, 400);
     }
 
-    // Call the real /models API — no fallback
-    const modelsUrl = resolvedBaseUrl.replace(/\/$/, "") + "/models";
-    try {
-      const res = await fetch(modelsUrl, {
-        headers: { Authorization: `Bearer ${apiKey.trim()}` },
-        signal: AbortSignal.timeout(10_000),
-      });
-
-      if (!res.ok) {
-        const body = await res.text().catch(() => "");
-        if (res.status === 401 || res.status === 403) {
-          return c.json({ ok: false, error: "API Key 无效，请检查后重试" }, 400);
-        }
-        return c.json({ ok: false, error: `服务商返回 ${res.status}: ${body.slice(0, 200)}` }, 400);
-      }
-
-      const json = await res.json() as { data?: Array<{ id: string; owned_by?: string }> };
-      const models = (json.data ?? []).map((m) => ({ id: m.id, name: m.id }));
-
-      if (models.length === 0) {
-        return c.json({ ok: false, error: "连接成功但未返回可用模型" }, 400);
-      }
-
-      const rawConfig = await loadRawConfig(root).catch(() => ({} as Record<string, unknown>));
-      const preferredModel = typeof (rawConfig.llm as Record<string, unknown> | undefined)?.defaultModel === "string"
-        ? String((rawConfig.llm as Record<string, unknown>).defaultModel)
-        : undefined;
-      const sampleModel = models.find((m) => m.id === preferredModel)?.id
-        ?? models.find((m) => m.id === "gpt-5.4")?.id
-        ?? models[0]?.id;
-      if (sampleModel) {
-        const client = createLLMClient({
-          provider: service === "anthropic" ? "anthropic" : "openai",
-          service: isCustomServiceId(service) ? "custom" : service,
-          configSource: "studio",
-          baseUrl: resolvedBaseUrl,
-          apiKey: apiKey.trim(),
-          model: sampleModel,
-          temperature: 0.7,
-          maxTokens: 64,
-          thinkingBudget: 0,
-          apiFormat: apiFormat ?? "chat",
-          stream: stream ?? true,
-        } as ProjectConfig["llm"]);
-
-        try {
-          await chatCompletion(
-            client,
-            sampleModel,
-            [{ role: "user", content: "ping" }],
-            { maxTokens: 5 },
-          );
-        } catch (error) {
-          const message = error instanceof Error ? error.message : String(error);
-          return c.json({ ok: false, error: `模型列表可读，但生成测试失败：${message}` }, 400);
-        }
-      }
+    const probe = await probeServiceCapabilities({
+      root,
+      service,
+      apiKey: apiKey.trim(),
+      baseUrl: resolvedBaseUrl,
+      preferredApiFormat: apiFormat,
+      preferredStream: stream,
+    });
 
-      return c.json({ ok: true, modelCount: models.length, models: models.slice(0, 50) });
-    } catch (err: any) {
-      if (err?.name === "TimeoutError" || err?.name === "AbortError") {
-        return c.json({ ok: false, error: `连接超时：无法访问 ${modelsUrl}` }, 400);
-      }
-      return c.json({ ok: false, error: `连接失败: ${err?.message ?? String(err)}` }, 400);
+    if (!probe.ok) {
+      return c.json({ ok: false, error: probe.error ?? "连接失败" }, 400);
     }
+
+    return c.json({
+      ok: true,
+      modelCount: probe.models.length,
+      models: probe.models.slice(0, 50),
+      selectedModel: probe.selectedModel,
+      detected: {
+        apiFormat: probe.apiFormat,
+        stream: probe.stream,
+        baseUrl: probe.baseUrl,
+        modelsSource: probe.modelsSource,
+      },
+    });
   });
 
   app.put("/api/v1/services/:service/secret", async (c) => {
@@ -888,20 +1023,22 @@ export function createStudioServer(initialConfig: ProjectConfig, root: string) {
     const resolvedBaseUrl = await resolveConfiguredServiceBaseUrl(root, service);
     if (!resolvedBaseUrl) return c.json({ models: [] });
 
-    // Call real API only
-    let models: Array<{ id: string; name: string }> = [];
-    try {
-      const modelsUrl = resolvedBaseUrl.replace(/\/$/, "") + "/models";
-      const res = await fetch(modelsUrl, {
-        headers: { Authorization: `Bearer ${apiKey}` },
-        signal: AbortSignal.timeout(10_000),
-      });
-      if (res.ok) {
-        const json = await res.json() as { data?: Array<{ id: string }> };
-        models = (json.data ?? []).map((m) => ({ id: m.id, name: m.id }));
-      }
-    } catch { /* timeout or network error — return empty */ }
-    return c.json({ models });
+    const rawConfig = await loadRawConfig(root).catch(() => ({} as Record<string, unknown>));
+    const llm = (rawConfig.llm as Record<string, unknown> | undefined) ?? {};
+    const preferredModel = typeof llm.defaultModel === "string" ? llm.defaultModel : undefined;
+    const serviceEntry = await resolveConfiguredServiceEntry(root, service);
+    const probe = await probeServiceCapabilities({
+      root,
+      service,
+      apiKey,
+      baseUrl: resolvedBaseUrl,
+      preferredApiFormat: serviceEntry?.apiFormat,
+      preferredStream: serviceEntry?.stream,
+      preferredModel,
+    });
+    return c.json({
+      models: probe.ok ? probe.models : [],
+    });
   });
 
   // --- Project info ---
@@ -2124,10 +2261,17 @@ export function createStudioServer(initialConfig: ProjectConfig, root: string) {
 
     try {
       const currentConfig = await loadCurrentProjectConfig({ requireApiKey: false });
-      const client = createLLMClient(currentConfig.llm);
-      const { chatCompletion } = await import("@actalk/inkos-core");
-      await chatCompletion(client, currentConfig.llm.model, [{ role: "user", content: "ping" }], { maxTokens: 5 });
-      checks.llmConnected = true;
+      const service = currentConfig.llm.service ?? currentConfig.llm.provider;
+      const probe = await probeServiceCapabilities({
+        root,
+        service,
+        apiKey: currentConfig.llm.apiKey,
+        baseUrl: currentConfig.llm.baseUrl,
+        preferredApiFormat: currentConfig.llm.apiFormat,
+        preferredStream: currentConfig.llm.stream,
+        preferredModel: currentConfig.llm.model,
+      });
+      checks.llmConnected = probe.ok;
     } catch { /* ignore */ }
 
     return c.json(checks);

packages/studio/src/pages/ServiceDetailPage.tsx

@@ -13,6 +13,13 @@ interface ModelInfo {
   readonly name?: string;
 }
 
+interface DetectedConfig {
+  readonly apiFormat?: "chat" | "responses";
+  readonly stream?: boolean;
+  readonly baseUrl?: string;
+  readonly modelsSource?: "api" | "fallback";
+}
+
 // Unified page state
 type ConnectionStatus =
   | { state: "idle" }               // No action taken yet
@@ -57,6 +64,8 @@ export function ServiceDetailPage({ serviceId, nav }: { serviceId: string; nav:
   const [maxTokens, setMaxTokens] = useState("4096");
   const [apiFormat, setApiFormat] = useState<"chat" | "responses">("chat");
   const [stream, setStream] = useState(true);
+  const [detectedModel, setDetectedModel] = useState<string>("");
+  const [detectedConfig, setDetectedConfig] = useState<DetectedConfig | null>(null);
 
   // -- Unified connection status --
   const [status, setStatus] = useState<ConnectionStatus>({ state: "idle" });
@@ -140,7 +149,14 @@ export function ServiceDetailPage({ serviceId, nav }: { serviceId: string; nav:
     setApiKey(trimmedKey);
     setStatus({ state: "testing" });
     try {
-      const result = await fetchJson<{ ok: boolean; models?: ModelInfo[]; modelCount?: number; error?: string }>(
+      const result = await fetchJson<{
+        ok: boolean;
+        models?: ModelInfo[];
+        modelCount?: number;
+        selectedModel?: string;
+        detected?: DetectedConfig;
+        error?: string;
+      }>(
         `/services/${encodeURIComponent(effectiveServiceId)}/test`,
         {
           method: "POST",
@@ -155,6 +171,11 @@ export function ServiceDetailPage({ serviceId, nav }: { serviceId: string; nav:
       );
       if (result.ok) {
         const models = result.models ?? [];
+        if (result.detected?.apiFormat) setApiFormat(result.detected.apiFormat);
+        if (typeof result.detected?.stream === "boolean") setStream(result.detected.stream);
+        if (isCustom && result.detected?.baseUrl) setBaseUrl(result.detected.baseUrl);
+        setDetectedModel(result.selectedModel ?? "");
+        setDetectedConfig(result.detected ?? null);
         setStatus({ state: "connected", models });
         setStoreModels(effectiveServiceId, models); // Write to global store
       } else {
@@ -186,6 +207,8 @@ export function ServiceDetailPage({ serviceId, nav }: { serviceId: string; nav:
         method: "PUT",
         headers: { "Content-Type": "application/json" },
         body: JSON.stringify({
+          service: effectiveServiceId,
+          ...(detectedModel ? { defaultModel: detectedModel } : {}),
           services: [
             {
               service: isCustom ? "custom" : serviceId,
@@ -290,7 +313,10 @@ export function ServiceDetailPage({ serviceId, nav }: { serviceId: string; nav:
           </button>
           {/* Status feedback */}
           {status.state === "connected" && (
-            <span className="text-xs text-emerald-500">连接成功，{models.length} 个模型</span>
+            <span className="text-xs text-emerald-500">
+              连接成功，{models.length} 个模型
+              {detectedModel ? `，已自动匹配 ${detectedModel}${detectedConfig ? ` / ${detectedConfig.apiFormat === "responses" ? "Responses" : "Chat"} / ${detectedConfig.stream ? "流式" : "非流式"}` : ""}` : ""}
+            </span>
           )}
           {status.state === "error" && (
             <span className="text-xs text-destructive">{status.message}</span>