Refactor AI provider model creation and configuration

- Removed individual provider creation logic in favor of a unified `createLanguageModel` function that handles multiple provider types.
- Introduced support for custom API formats (OpenAI and Anthropic) in provider configurations.
- Updated the AIProviderDialog component to allow users to select the API format for compatible providers.
- Removed the deprecated thinking mode utility functions and integrated thinking options directly into provider configurations.
- Enhanced error handling and connection testing for providers.
- Updated storage functions to handle sidepanel thinking mode state more robustly.
- Removed unused Qwen compatibility functions and streamlined provider handling.

Author: lcomplete

app/extension/src/__tests__/openAICompatibleStream.test.ts

@@ -2,73 +2,10 @@ import {
   getOpenAICompatibleBaseUrl,
   getOllamaBaseUrl,
   getOllamaOpenAIBaseUrl,
-  isDashScopeCompatibleBaseUrl,
-  usesRawOpenAICompatibleStream,
 } from "../ai/openAICompatibleProviders";
+import { getEffectiveApiFormat } from "../ai/types";
 
 describe("providers helpers", () => {
-  it("detects DashScope-compatible endpoints", () => {
-    expect(
-      isDashScopeCompatibleBaseUrl(
-        "https://dashscope.aliyuncs.com/compatible-mode/v1"
-      )
-    ).toBe(true);
-    expect(
-      isDashScopeCompatibleBaseUrl(
-        "https://dashscope-intl.aliyuncs.com/compatible-mode/v1"
-      )
-    ).toBe(true);
-    expect(isDashScopeCompatibleBaseUrl("https://api.openai.com/v1")).toBe(
-      false
-    );
-  });
-
-  it("uses raw OpenAI-compatible streaming only for providers that need it", () => {
-    expect(
-      usesRawOpenAICompatibleStream({
-        type: "qwen",
-        enabled: true,
-        apiKey: "test",
-        baseUrl: "",
-        enabledModels: ["qwen3-max"],
-        updatedAt: Date.now(),
-      })
-    ).toBe(true);
-
-    expect(
-      usesRawOpenAICompatibleStream({
-        type: "openai",
-        enabled: true,
-        apiKey: "test",
-        baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1",
-        enabledModels: ["qwen3-max"],
-        updatedAt: Date.now(),
-      })
-    ).toBe(true);
-
-    expect(
-      usesRawOpenAICompatibleStream({
-        type: "openai",
-        enabled: true,
-        apiKey: "test",
-        baseUrl: "https://api.openai.com/v1",
-        enabledModels: ["gpt-4.1"],
-        updatedAt: Date.now(),
-      })
-    ).toBe(false);
-
-    expect(
-      usesRawOpenAICompatibleStream({
-        type: "google",
-        enabled: true,
-        apiKey: "test",
-        baseUrl: "",
-        enabledModels: ["gemini-2.5-flash"],
-        updatedAt: Date.now(),
-      })
-    ).toBe(false);
-  });
-
   it("returns configured or default OpenAI-compatible base url", () => {
     expect(
       getOpenAICompatibleBaseUrl({
@@ -106,4 +43,25 @@ describe("providers helpers", () => {
       "http://localhost:11434/v1"
     );
   });
+
+  it("falls back to the provider native format when no override is given", () => {
+    expect(getEffectiveApiFormat({ type: "qwen" })).toBe("openai");
+    expect(getEffectiveApiFormat({ type: "anthropic" })).toBe("anthropic");
+  });
+
+  it("honours the user-selected api format for flexible providers", () => {
+    expect(
+      getEffectiveApiFormat({ type: "qwen", apiFormat: "anthropic" })
+    ).toBe("anthropic");
+    expect(
+      getEffectiveApiFormat({ type: "zhipu", apiFormat: "anthropic" })
+    ).toBe("anthropic");
+  });
+
+  it("ignores an api format override on providers that do not allow it", () => {
+    // openai has a fixed native format; the override should be ignored.
+    expect(
+      getEffectiveApiFormat({ type: "openai", apiFormat: "anthropic" })
+    ).toBe("openai");
+  });
 });

app/extension/src/__tests__/providers.test.ts

@@ -1,43 +1,21 @@
-import { AIProviderConfig, ProviderType, PROVIDER_REGISTRY } from "./types";
+import { AIProviderConfig, PROVIDER_REGISTRY } from "./types";
 
-const ALWAYS_RAW_OPENAI_COMPATIBLE_PROVIDERS = new Set<ProviderType>([
-  "qwen",
-  "zhipu",
-  "minimax",
-]);
-
-export function isDashScopeCompatibleBaseUrl(baseUrl?: string): boolean {
-  const normalizedBaseUrl = baseUrl?.toLowerCase() || "";
+function trimTrailingSlash(url: string): string {
+  return url.replace(/\/+$/, "");
+}
 
+export function getProviderBaseUrl(config: AIProviderConfig): string | undefined {
   return (
-    normalizedBaseUrl.includes("dashscope.aliyuncs.com/compatible-mode") ||
-    normalizedBaseUrl.includes("dashscope-intl.aliyuncs.com/compatible-mode")
+    config.baseUrl ||
+    PROVIDER_REGISTRY[config.type]?.defaultBaseUrl ||
+    undefined
   );
 }
 
-export function usesRawOpenAICompatibleStream(
-  config: AIProviderConfig
-): boolean {
-  if (ALWAYS_RAW_OPENAI_COMPATIBLE_PROVIDERS.has(config.type)) {
-    return true;
-  }
-
-  if (config.type === "openai") {
-    return isDashScopeCompatibleBaseUrl(getOpenAICompatibleBaseUrl(config));
-  }
-
-  return false;
-}
-
-export function getOpenAICompatibleBaseUrl(
-  config: AIProviderConfig
-): string | undefined {
-  return config.baseUrl || PROVIDER_REGISTRY[config.type]?.defaultBaseUrl || undefined;
-}
-
-function trimTrailingSlash(url: string): string {
-  return url.replace(/\/+$/, "");
-}
+/**
+ * @deprecated use {@link getProviderBaseUrl}. Kept for call sites still being migrated.
+ */
+export const getOpenAICompatibleBaseUrl = getProviderBaseUrl;
 
 export function getOllamaBaseUrl(baseUrl?: string): string {
   const normalizedBaseUrl = trimTrailingSlash(
@@ -50,7 +28,5 @@ export function getOllamaBaseUrl(baseUrl?: string): string {
 }
 
 export function getOllamaOpenAIBaseUrl(baseUrl?: string): string {
-  const normalizedBaseUrl = getOllamaBaseUrl(baseUrl);
-
-  return `${normalizedBaseUrl}/v1`;
+  return `${getOllamaBaseUrl(baseUrl)}/v1`;
 }