refactor: reorganize directory structure and fix Qwen 3.5 compatibility (#14)

* feat(provider): introduce Qwen error handling and refactor provider structure

- Added `error.ts` for Qwen error response schema and handling.
- Created `provider.ts` to encapsulate QwenProvider functionality, including model creation methods.
- Updated `index.ts` to export new provider structure and removed deprecated references.
- Removed `map-qwen-finish-reason.ts` as it is no longer needed.
- Added comprehensive tests for the new provider structure and error handling.

* chore: lint fix

Author: hairyf

src/__tests__/integration/reranking.integration.test.ts

@@ -6,7 +6,7 @@
  */
 import { rerank } from "ai"
 import { describe, expect, it } from "vitest"
-import { createQwen } from "../../qwen-provider"
+import { createQwen } from "../../provider"
 import "dotenv/config"
 
 // Skip if no API key

…tests__/qwen-chat-language-model.test.ts src/__tests__/models/chat.test.tssrc/__tests__/qwen-chat-language-model.test.ts renamed to src/__tests__/models/chat.test.ts src/__tests__/qwen-chat-language-model.test.ts renamed to src/__tests__/models/chat.test.ts

@@ -2,8 +2,8 @@
 import type { LanguageModelV3Prompt } from "@ai-sdk/provider"
 import { convertReadableStreamToArray } from "@ai-sdk/provider-utils/test"
 import { beforeEach, describe, expect, it, vi } from "vitest"
-import { QwenChatLanguageModel } from "../qwen-chat-language-model"
-import { createQwen } from "../qwen-provider"
+import { QwenChatLanguageModel } from "../../models/chat"
+import { createQwen } from "../../provider"
 
 const TEST_PROMPT: LanguageModelV3Prompt = [
   { role: "user", content: [{ type: "text", text: "Hello" }] },
@@ -301,6 +301,7 @@ describe("doGenerate", () => {
 
     expect(requestBody).toStrictEqual({
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
     })
   })
@@ -316,6 +317,7 @@ describe("doGenerate", () => {
 
     expect(requestBody).toStrictEqual({
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
       user: "test-user-id",
     })
@@ -335,6 +337,7 @@ describe("doGenerate", () => {
 
     expect(requestBody).toStrictEqual({
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
       someCustomOption: "test-value",
     })
@@ -354,6 +357,7 @@ describe("doGenerate", () => {
 
     expect(requestBody).toStrictEqual({
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
     })
   })
@@ -385,6 +389,7 @@ describe("doGenerate", () => {
 
     expect(requestBody).toStrictEqual({
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
       tools: [
         {
@@ -557,6 +562,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
       })
     })
@@ -573,6 +579,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         response_format: { type: "json_object" },
       })
@@ -616,6 +623,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         response_format: { type: "json_object" },
       })
@@ -668,6 +676,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         response_format: {
           type: "json_schema",
@@ -727,6 +736,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         response_format: {
           type: "json_schema",
@@ -776,6 +786,7 @@ describe("doGenerate", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         response_format: {
           type: "json_object",
@@ -793,7 +804,7 @@ describe("doGenerate", () => {
     })
 
     expect(request).toMatchObject({
-      body: "{\"model\":\"qwen-chat\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}",
+      body: "{\"model\":\"qwen-chat\",\"extra_body\":{\"enable_thinking\":false},\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}]}",
     })
   })
 })
@@ -1125,7 +1136,10 @@ describe("doStream", () => {
 
     const parts = await convertReadableStreamToArray(stream)
 
-    expect(parts.find(p => p.type === "error")).toBeFalsy()
+    expect(parts).toContainEqual({
+      type: "error",
+      error: "InternalServerError: list index out of range",
+    })
     expect(parts).toContainEqual({
       type: "tool-input-end",
       id: "call_recover_1",
@@ -1138,15 +1152,15 @@ describe("doStream", () => {
     })
     expect(parts).toContainEqual({
       type: "finish",
-      finishReason: { unified: "tool-calls", raw: "tool_calls" },
+      finishReason: { unified: "other", raw: undefined },
       usage: {
         inputTokens: { total: undefined, noCache: undefined, cacheRead: undefined, cacheWrite: undefined },
         outputTokens: { total: undefined, text: undefined, reasoning: undefined },
       },
     })
   })
 
-  it("should retry doStream request on `500 + list index out of range`", async () => {
+  it("should throw on first `500 + list index out of range` for doStream request", async () => {
     let attempts = 0
 
     const provider = createQwen({
@@ -1187,14 +1201,10 @@ describe("doStream", () => {
     })
 
     const model = provider("qwen-chat")
-    const { stream } = await model.doStream({
+    await expect(model.doStream({
       prompt: TEST_PROMPT,
-    })
-
-    const parts = await convertReadableStreamToArray(stream)
-
-    expect(attempts).toBe(2)
-    expect(parts.find(p => p.type === "finish")).toBeTruthy()
+    })).rejects.toThrow("InternalServerError: list index out of range")
+    expect(attempts).toBe(1)
   })
 
   it("should stream tool deltas with V3 format", async () => {
@@ -1422,7 +1432,7 @@ describe("doStream", () => {
       type: "tool-call",
       toolCallId: "chatcmpl-tool-b3b307239370432d9910d4b79b4dbbaa",
       toolName: "searchGoogle",
-      input: "{\"query\": \"latest news on ai\"}",
+      input: "{\"query\":\"latest news on ai\"}",
     })
   })
 
@@ -1534,6 +1544,7 @@ describe("doStream", () => {
         include_usage: true,
       },
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
     })
   })
@@ -1579,6 +1590,7 @@ describe("doStream", () => {
         include_usage: true,
       },
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
       someCustomOption: "test-value",
     })
@@ -1602,6 +1614,7 @@ describe("doStream", () => {
         include_usage: true,
       },
       model: "qwen-chat",
+      extra_body: { enable_thinking: false },
       messages: [{ role: "user", content: "Hello" }],
     })
   })
@@ -1615,7 +1628,7 @@ describe("doStream", () => {
     })
 
     expect(request).toMatchObject({
-      body: "{\"model\":\"qwen-chat\",\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}],\"stream\":true,\"stream_options\":{\"include_usage\":true}}",
+      body: "{\"model\":\"qwen-chat\",\"extra_body\":{\"enable_thinking\":false},\"messages\":[{\"role\":\"user\",\"content\":\"Hello\"}],\"stream\":true,\"stream_options\":{\"include_usage\":true}}",
     })
   })
 })
@@ -1923,6 +1936,7 @@ describe("metadata extraction", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
       })
     })
@@ -1993,6 +2007,7 @@ describe("metadata extraction", () => {
 
       expect(requestBody).toStrictEqual({
         model: "qwen-plus",
+        extra_body: { enable_thinking: false },
         messages: [{ role: "user", content: "Hello" }],
         stream: true,
         stream_options: {

…_/qwen-completion-language-model.test.ts src/__tests__/models/completion.test.tssrc/__tests__/qwen-completion-language-model.test.ts renamed to src/__tests__/models/completion.test.ts src/__tests__/qwen-completion-language-model.test.ts renamed to src/__tests__/models/completion.test.ts

@@ -2,8 +2,8 @@
 import type { LanguageModelV3Prompt } from "@ai-sdk/provider"
 import { convertReadableStreamToArray } from "@ai-sdk/provider-utils/test"
 import { beforeEach, describe, expect, it, vi } from "vitest"
-import { QwenCompletionLanguageModel } from "../qwen-completion-language-model"
-import { createQwen } from "../qwen-provider"
+import { QwenCompletionLanguageModel } from "../../models/completion"
+import { createQwen } from "../../provider"
 
 const TEST_PROMPT: LanguageModelV3Prompt = [
   { role: "user", content: [{ type: "text", text: "Hello" }] },
@@ -394,6 +394,58 @@ describe("doStream", () => {
     })
   })
 
+  it("should throw on retryable errors from stream request", async () => {
+    let fetchCalls = 0
+
+    const provider = createQwen({
+      baseURL: "https://my.api.com/v1/",
+      headers: {
+        Authorization: `Bearer test-api-key`,
+      },
+      fetch: async (_url, init) => {
+        fetchCalls++
+
+        if (init?.body) {
+          requestBody = JSON.parse(init.body as string)
+        }
+
+        if (fetchCalls === 1) {
+          return new Response(JSON.stringify({
+            object: "error",
+            message: "InternalServerError: list index out of range",
+            type: "InternalServerError",
+            param: null,
+            code: null,
+          }), {
+            status: 500,
+            headers: { "content-type": "application/json" },
+          })
+        }
+
+        const encoder = new TextEncoder()
+        const stream = new ReadableStream({
+          start(controller) {
+            for (const chunk of responseChunks) {
+              controller.enqueue(encoder.encode(chunk))
+            }
+            controller.close()
+          },
+        })
+
+        return new Response(stream, {
+          headers: responseHeaders,
+        })
+      },
+    })
+
+    const model = provider.completion("qwen-plus")
+    await expect(model.doStream({
+      prompt: TEST_PROMPT,
+    })).rejects.toThrow("InternalServerError: list index out of range")
+    expect(fetchCalls).toBe(1)
+    expect(requestBody).toMatchObject({ stream: true })
+  })
+
   it("should handle unparsable stream parts", async () => {
     responseChunks = [`data: {unparsable}\n\n`, "data: [DONE]\n\n"]
     const provider = createStreamingTestProvider()

…c/__tests__/qwen-embedding-model.test.ts src/__tests__/models/embedding.test.tssrc/__tests__/qwen-embedding-model.test.ts renamed to src/__tests__/models/embedding.test.ts src/__tests__/qwen-embedding-model.test.ts renamed to src/__tests__/models/embedding.test.ts

@@ -1,6 +1,6 @@
 import { TooManyEmbeddingValuesForCallError } from "@ai-sdk/provider"
 import { beforeEach, describe, expect, it, vi } from "vitest"
-import { createQwen } from "../qwen-provider"
+import { createQwen } from "../../provider"
 
 const dummyEmbeddings = [
   [0.1, 0.2, 0.3, 0.4, 0.5],
@@ -169,4 +169,45 @@ describe("doEmbed", () => {
       async () => await model.doEmbed({ values: many }),
     ).rejects.toBeInstanceOf(TooManyEmbeddingValuesForCallError)
   })
+
+  it("should throw on retryable errors from embed request", async () => {
+    let fetchCalls = 0
+
+    const provider = createQwen({
+      baseURL: "https://my.api.com/v1/",
+      headers: {
+        Authorization: `Bearer test-api-key`,
+      },
+      fetch: async (_url, init) => {
+        fetchCalls++
+        if (init?.body) {
+          requestBody = JSON.parse(init.body as string)
+        }
+
+        if (fetchCalls === 1) {
+          return new Response(JSON.stringify({
+            object: "error",
+            message: "Service Unavailable",
+            type: "ServiceUnavailable",
+            param: null,
+            code: null,
+          }), {
+            status: 503,
+            headers: { "content-type": "application/json" },
+          })
+        }
+
+        return new Response(JSON.stringify(responseBody), {
+          headers: responseHeaders,
+        })
+      },
+    })
+
+    const model = provider.textEmbeddingModel("text-embedding-3-large")
+    await expect(model.doEmbed({ values: testValues })).rejects.toThrow(
+      "Service Unavailable",
+    )
+    expect(fetchCalls).toBe(1)
+    expect(requestBody).toMatchObject({ model: "text-embedding-3-large" })
+  })
 })