opencode-qwen-cli-auth 2.3.1 → 2.3.3

package/README.md

@@ -67,7 +67,7 @@ The plugin stores each successful login in the multi-account store and can auto-
 | `ENABLE_PLUGIN_REQUEST_LOGGING=1` | Enable request logging to file | Optional |
 | `OPENCODE_QWEN_ENABLE_CLI_FALLBACK=1` | Enable CLI fallback on quota error | Optional |
 | `OPENCODE_QWEN_ACCOUNTS_PATH` | Override multi-account store path (must be inside `~/.qwen`) | Optional |
-| `OPENCODE_QWEN_QUOTA_COOLDOWN_MS` | Cooldown for exhausted accounts | Default: `1800000` (30 min) |
+| `OPENCODE_QWEN_QUOTA_COOLDOWN_MS` | Cooldown for exhausted accounts | Default: `86400000` (24h) |
 
 ### Debug & Logging
 

package/README.vi.md

@@ -67,7 +67,7 @@ Plugin sẽ lưu từng lần đăng nhập thành công vào kho đa tài kho
 | `ENABLE_PLUGIN_REQUEST_LOGGING=1` | Bật ghi log request ra file | Tùy chọn |
 | `OPENCODE_QWEN_ENABLE_CLI_FALLBACK=1` | Bật tính năng gọi CLI khi hết quota | Tùy chọn |
 | `OPENCODE_QWEN_ACCOUNTS_PATH` | Ghi đè đường dẫn kho đa tài khoản (phải nằm trong `~/.qwen`) | Tùy chọn |
-| `OPENCODE_QWEN_QUOTA_COOLDOWN_MS` | Thời gian cooldown cho tài khoản đã hết quota | Mặc định: `1800000` (30 phút) |
+| `OPENCODE_QWEN_QUOTA_COOLDOWN_MS` | Thời gian cooldown cho tài khoản đã hết quota | Mặc định: `86400000` (24 giờ) |
 
 ### Debug & Logging
 

package/dist/index.js

@@ -22,7 +22,7 @@ import { PROVIDER_ID, AUTH_LABELS, DEVICE_FLOW, PORTAL_HEADERS } from "./lib/con
 import { logError, logInfo, logWarn, LOGGING_ENABLED } from "./lib/logger.js";
 
 /** Request timeout for chat completions in milliseconds */
-const CHAT_REQUEST_TIMEOUT_MS = 30000;
+const CHAT_REQUEST_TIMEOUT_MS = 120000;
 /** Maximum number of retry attempts for failed requests */
 const CHAT_MAX_RETRIES = 3;
 /** Output token cap for coder-model (64K tokens) */
@@ -415,19 +415,19 @@ function sanitizeOutgoingPayload(payload) {
 function createQuotaDegradedPayload(payload) {
     const degraded = { ...payload };
     let changed = false;
-    // Remove tool-related fields
-    if ("tools" in degraded) {
-        delete degraded.tools;
-        changed = true;
-    }
-    if ("tool_choice" in degraded) {
-        delete degraded.tool_choice;
-        changed = true;
-    }
-    if ("parallel_tool_calls" in degraded) {
-        delete degraded.parallel_tool_calls;
-        changed = true;
-    }
+    // Remove tool-related fields (skip removing tools so Agents don't break)
+    // if ("tools" in degraded) {
+    //     delete degraded.tools;
+    //     changed = true;
+    // }
+    // if ("tool_choice" in degraded) {
+    //     delete degraded.tool_choice;
+    //     changed = true;
+    // }
+    // if ("parallel_tool_calls" in degraded) {
+    //     delete degraded.parallel_tool_calls;
+    //     changed = true;
+    // }
     // Disable streaming
     if (degraded.stream !== false) {
         degraded.stream = false;
@@ -595,7 +595,7 @@ function createSseResponseChunk(data) {
  * @param {boolean} streamMode - Whether to return streaming response
  * @returns {Response} Formatted completion response
  */
-function makeQwenCliCompletionResponse(model, content, context, streamMode) {
+function makeQwenCliCompletionResponse(model, content, context, streamMode, abortSignal) {
     if (LOGGING_ENABLED) {
         logInfo("Qwen CLI fallback returned completion", {
             request_id: context.requestId,
@@ -609,7 +609,7 @@ function makeQwenCliCompletionResponse(model, content, context, streamMode) {
         const encoder = new TextEncoder();
         const stream = new ReadableStream({
             start(controller) {
-                // Send first chunk with content
+                // Send first chunk with empty content
                 controller.enqueue(encoder.encode(createSseResponseChunk({
                     id: completionId,
                     object: "chat.completion.chunk",
@@ -618,28 +618,51 @@ function makeQwenCliCompletionResponse(model, content, context, streamMode) {
                     choices: [
                         {
                             index: 0,
-                            delta: { role: "assistant", content },
+                            delta: { role: "assistant", content: "" },
                             finish_reason: null,
                         },
                     ],
                 })));
-                // Send stop chunk
-                controller.enqueue(encoder.encode(createSseResponseChunk({
-                    id: completionId,
-                    object: "chat.completion.chunk",
-                    created,
-                    model,
-                    choices: [
-                        {
-                            index: 0,
-                            delta: {},
-                            finish_reason: "stop",
-                        },
-                    ],
-                })));
-                // Send DONE marker
-                controller.enqueue(encoder.encode("data: [DONE]\n\n"));
-                controller.close();
+
+                const CHUNK_SIZE = 15;
+                const DELAY_MS = 20;
+                let position = 0;
+
+                function pushNextChunk() {
+                    if (abortSignal?.aborted) {
+                        try { controller.close(); } catch (e) { }
+                        return;
+                    }
+
+                    if (position >= content.length) {
+                        // Send stop chunk
+                        controller.enqueue(encoder.encode(createSseResponseChunk({
+                            id: completionId,
+                            object: "chat.completion.chunk",
+                            created,
+                            model,
+                            choices: [{ index: 0, delta: {}, finish_reason: "stop" }],
+                        })));
+                        controller.enqueue(encoder.encode("data: [DONE]\n\n"));
+                        try { controller.close(); } catch (e) { }
+                        return;
+                    }
+
+                    const nextSlice = content.slice(position, position + CHUNK_SIZE);
+                    position += CHUNK_SIZE;
+
+                    controller.enqueue(encoder.encode(createSseResponseChunk({
+                        id: completionId,
+                        object: "chat.completion.chunk",
+                        created,
+                        model,
+                        choices: [{ index: 0, delta: { content: nextSlice }, finish_reason: null }],
+                    })));
+
+                    setTimeout(pushNextChunk, DELAY_MS);
+                }
+
+                pushNextChunk();
             },
         });
         return new Response(stream, {
@@ -701,12 +724,8 @@ async function runQwenCliFallback(payload, context, abortSignal) {
             command: QWEN_CLI_COMMAND,
         });
     }
-    if (requiresShellExecution(QWEN_CLI_COMMAND)) {
-        return {
-            ok: false,
-            reason: "cli_shell_execution_blocked_for_security",
-        };
-    }
+    // Use secure spawn logic across ALL OS, allowing .cmd locally on Windows by injecting shell correctly.
+    const isShellRequired = requiresShellExecution(QWEN_CLI_COMMAND);
     return await new Promise((resolve) => {
         let settled = false;
         let stdout = "";
@@ -736,7 +755,7 @@ async function runQwenCliFallback(payload, context, abortSignal) {
         }
         try {
             child = spawn(QWEN_CLI_COMMAND, args, {
-                shell: false,
+                shell: isShellRequired,
                 windowsHide: true,
                 stdio: ["ignore", "pipe", "pipe"],
             });
@@ -791,7 +810,7 @@ async function runQwenCliFallback(payload, context, abortSignal) {
             if (content) {
                 finalize({
                     ok: true,
-                    response: makeQwenCliCompletionResponse(model, content, context, streamMode),
+                    response: makeQwenCliCompletionResponse(model, content, context, streamMode, abortSignal),
                 });
                 return;
             }
@@ -964,7 +983,7 @@ async function failFastFetch(input, init) {
                     attempt: retryAttempt + 1,
                 });
             }
-const RETRYABLE_STATUS_CODES = [429, 500, 502, 503, 504];
+            const RETRYABLE_STATUS_CODES = [429, 500, 502, 503, 504];
             if (RETRYABLE_STATUS_CODES.includes(response.status)) {
                 if (response.status === 429) {
                     const firstBody = await response.text().catch(() => "");

package/dist/lib/auth/auth.js

@@ -30,7 +30,7 @@ const LOCK_MAX_ATTEMPTS = 20;
 /** Account schema version for ~/.qwen/oauth_accounts.json */
 const ACCOUNT_STORE_VERSION = 1;
 /** Default cooldown when account hits insufficient_quota */
-const DEFAULT_QUOTA_COOLDOWN_MS = 30 * 60 * 1000;
+const DEFAULT_QUOTA_COOLDOWN_MS = 24 * 60 * 60 * 1000;
 
 /**
  * Checks if an error is an AbortError (from AbortController)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-qwen-cli-auth",
-  "version": "2.3.1",
+  "version": "2.3.3",
   "description": "Qwen OAuth authentication plugin for opencode - use your Qwen account instead of API keys",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",