qwenproxy-cli 1.0.25 → 1.0.27

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "qwenproxy-cli",
-  "version": "1.0.25",
+  "version": "1.0.27",
   "description": "High-performance OpenAI & Anthropic compatible API gateway for Qwen with multi-account rotation, interactive TUI, and resilient tool calling.",
   "main": "src/index.ts",
   "bin": {
@@ -24,8 +24,8 @@
     "reset": "npx tsx src/reset-cooldowns.ts",
     "purge": "npx tsx src/delete-chats.ts",
     "test": "npm run test:mock && npm run test:live",
-    "test:mock": "tsx --test --test-concurrency=1 --env-file=.env.test src/tests/account-concurrency.test.ts src/tests/concurrency-latest-wins.test.ts src/tests/mutex.test.ts src/tests/context-truncation.test.ts src/tests/streaming-thinking-summary.test.ts src/tests/tool-call-guard.test.ts src/tests/utils-logger-full.test.ts src/tests/logger-toolcall-errors.test.ts src/tests/logger-toolcall-disabled.test.ts src/tests/utils-json-full.test.ts src/tests/utils-session-guard-full.test.ts src/tests/core-logic-full.test.ts src/tests/core-logic-accounts.test.ts src/tests/chat-validation-full.test.ts src/tests/chat-logic-full.test.ts src/tests/account-stickiness.test.ts src/tests/accounts-security.test.ts src/tests/advanced.test.ts src/tests/auth-playwright.test.ts src/tests/captcha-solver.test.ts src/tests/cache-scan-regex.test.ts src/tests/chat-cleanup.test.ts src/tests/chromium-args.test.ts src/tests/config-current.test.ts src/tests/context-meter.test.ts src/tests/concurrency.test.ts src/tests/delta.test.ts src/tests/fingerprint.test.ts src/tests/human-behavior.test.ts src/tests/index.test.ts src/tests/json-recovery.test.ts src/tests/jsonFix.test.ts src/tests/memory-cache.test.ts src/tests/model-alias.test.ts src/tests/model-registry-thinking.test.ts src/tests/models-endpoint.test.ts src/tests/multimodal.test.ts src/tests/oss-upload.test.ts src/tests/parser.test.ts src/tests/personalization-flow.test.ts src/tests/personalization-required.test.ts src/tests/retry-policy.test.ts src/tests/rotation.test.ts src/tests/server-lifecycle.test.ts src/tests/session-keeper.test.ts src/tests/session-state.test.ts src/tests/single-account-mode.test.ts src/tests/stream-retry.test.ts src/tests/tool-calls-endpoint.test.ts src/tests/tool-instructions-cache.test.ts src/tests/session-id.test.ts src/tests/watchdog-ram.test.ts src/tests/tool-recovery-t1t2.test.ts src/tests/abort-silent-t4.test.ts src/tests/t3-close-tag-escape.test.ts src/tests/t5-tool-cap.test.ts src/tests/logger-redaction.test.ts src/tests/stall-timeout-retryable.test.ts src/tests/t6-missing-quote-array-close.test.ts src/tests/watchdog-rss.test.ts src/tests/logical-thread-batch.test.ts src/tests/chat-pool-style.test.ts src/tests/queue-wait-cap.test.ts src/tests/context-eviction-order.test.ts src/tests/chat-in-progress-settle.test.ts src/tests/stream-lock-leak.test.ts src/tests/stream-semaphore.test.ts src/tests/waf-isolation.test.ts src/tests/legacy-completions.test.ts src/tests/response-format.test.ts src/tests/cors-headers.test.ts src/tests/stream-lock-terminal-cancel.test.ts src/tests/parser-truncated-tool-call.test.ts src/tests/metrics-prometheus.test.ts src/tests/context-death-cleanup.test.ts src/tests/personalization-direct-fetch.test.ts src/tests/completion-headers-parity.test.ts src/tests/parallel-escape-repro.test.ts src/tests/chat-in-progress-scale-cost.test.ts src/tests/temp-chat-mode.test.ts src/tests/profile-reset.test.ts src/tests/error-classification.test.ts src/tests/quota-cooldown.test.ts src/tests/account-ready-gate.test.ts src/tests/tool-cap-notice.test.ts src/tests/interaction-flows.test.ts src/tests/headers-gate-cooldown-fallback.test.ts src/tests/unwrapped-tool-calls.test.ts src/tests/anthropic.test.ts src/tests/sync-clients.test.ts src/tests/cli.test.ts src/tests/paths.test.ts src/tests/tool-manifest.test.ts src/tests/tui.test.ts src/tests/personalization-deadline.test.ts",
-    "test:mock:fast": "tsx --test --test-concurrency=2 --env-file=.env.test src/tests/account-concurrency.test.ts src/tests/concurrency-latest-wins.test.ts src/tests/mutex.test.ts src/tests/context-truncation.test.ts src/tests/streaming-thinking-summary.test.ts src/tests/tool-call-guard.test.ts src/tests/utils-logger-full.test.ts src/tests/logger-toolcall-errors.test.ts src/tests/logger-toolcall-disabled.test.ts src/tests/utils-json-full.test.ts src/tests/utils-session-guard-full.test.ts src/tests/core-logic-full.test.ts src/tests/core-logic-accounts.test.ts src/tests/chat-validation-full.test.ts src/tests/chat-logic-full.test.ts src/tests/account-stickiness.test.ts src/tests/accounts-security.test.ts src/tests/advanced.test.ts src/tests/auth-playwright.test.ts src/tests/captcha-solver.test.ts src/tests/cache-scan-regex.test.ts src/tests/chat-cleanup.test.ts src/tests/chromium-args.test.ts src/tests/config-current.test.ts src/tests/context-meter.test.ts src/tests/concurrency.test.ts src/tests/delta.test.ts src/tests/fingerprint.test.ts src/tests/human-behavior.test.ts src/tests/index.test.ts src/tests/json-recovery.test.ts src/tests/jsonFix.test.ts src/tests/memory-cache.test.ts src/tests/model-alias.test.ts src/tests/model-registry-thinking.test.ts src/tests/models-endpoint.test.ts src/tests/multimodal.test.ts src/tests/oss-upload.test.ts src/tests/parser.test.ts src/tests/personalization-flow.test.ts src/tests/personalization-required.test.ts src/tests/retry-policy.test.ts src/tests/rotation.test.ts src/tests/server-lifecycle.test.ts src/tests/session-keeper.test.ts src/tests/session-state.test.ts src/tests/single-account-mode.test.ts src/tests/stream-retry.test.ts src/tests/tool-calls-endpoint.test.ts src/tests/tool-instructions-cache.test.ts src/tests/session-id.test.ts src/tests/watchdog-ram.test.ts src/tests/tool-recovery-t1t2.test.ts src/tests/abort-silent-t4.test.ts src/tests/t3-close-tag-escape.test.ts src/tests/t5-tool-cap.test.ts src/tests/logger-redaction.test.ts src/tests/stall-timeout-retryable.test.ts src/tests/t6-missing-quote-array-close.test.ts src/tests/watchdog-rss.test.ts src/tests/logical-thread-batch.test.ts src/tests/chat-pool-style.test.ts src/tests/queue-wait-cap.test.ts src/tests/context-eviction-order.test.ts src/tests/chat-in-progress-settle.test.ts src/tests/stream-lock-leak.test.ts src/tests/stream-semaphore.test.ts src/tests/waf-isolation.test.ts src/tests/legacy-completions.test.ts src/tests/response-format.test.ts src/tests/cors-headers.test.ts src/tests/stream-lock-terminal-cancel.test.ts src/tests/parser-truncated-tool-call.test.ts src/tests/metrics-prometheus.test.ts src/tests/context-death-cleanup.test.ts src/tests/personalization-direct-fetch.test.ts src/tests/completion-headers-parity.test.ts src/tests/parallel-escape-repro.test.ts src/tests/chat-in-progress-scale-cost.test.ts src/tests/temp-chat-mode.test.ts src/tests/profile-reset.test.ts src/tests/error-classification.test.ts src/tests/quota-cooldown.test.ts src/tests/account-ready-gate.test.ts src/tests/tool-cap-notice.test.ts src/tests/interaction-flows.test.ts src/tests/headers-gate-cooldown-fallback.test.ts src/tests/unwrapped-tool-calls.test.ts src/tests/anthropic.test.ts src/tests/sync-clients.test.ts src/tests/cli.test.ts src/tests/paths.test.ts src/tests/tool-manifest.test.ts src/tests/tui.test.ts src/tests/personalization-deadline.test.ts",
+    "test:mock": "tsx --test --test-concurrency=1 --env-file=.env.test src/tests/account-concurrency.test.ts src/tests/concurrency-latest-wins.test.ts src/tests/mutex.test.ts src/tests/context-truncation.test.ts src/tests/streaming-thinking-summary.test.ts src/tests/tool-call-guard.test.ts src/tests/utils-logger-full.test.ts src/tests/logger-toolcall-errors.test.ts src/tests/logger-toolcall-disabled.test.ts src/tests/utils-json-full.test.ts src/tests/utils-session-guard-full.test.ts src/tests/core-logic-full.test.ts src/tests/core-logic-accounts.test.ts src/tests/chat-validation-full.test.ts src/tests/chat-logic-full.test.ts src/tests/account-stickiness.test.ts src/tests/accounts-security.test.ts src/tests/advanced.test.ts src/tests/auth-playwright.test.ts src/tests/captcha-solver.test.ts src/tests/cache-scan-regex.test.ts src/tests/chat-cleanup.test.ts src/tests/chat-cleanup-auto.test.ts src/tests/chromium-args.test.ts src/tests/config-current.test.ts src/tests/context-meter.test.ts src/tests/concurrency.test.ts src/tests/delta.test.ts src/tests/fingerprint.test.ts src/tests/human-behavior.test.ts src/tests/index.test.ts src/tests/json-recovery.test.ts src/tests/jsonFix.test.ts src/tests/memory-cache.test.ts src/tests/model-alias.test.ts src/tests/model-registry-thinking.test.ts src/tests/models-endpoint.test.ts src/tests/multimodal.test.ts src/tests/oss-upload.test.ts src/tests/parser.test.ts src/tests/personalization-flow.test.ts src/tests/personalization-required.test.ts src/tests/retry-policy.test.ts src/tests/rotation.test.ts src/tests/server-lifecycle.test.ts src/tests/session-keeper.test.ts src/tests/session-state.test.ts src/tests/single-account-mode.test.ts src/tests/stream-retry.test.ts src/tests/tool-calls-endpoint.test.ts src/tests/tool-instructions-cache.test.ts src/tests/session-id.test.ts src/tests/watchdog-ram.test.ts src/tests/tool-recovery-t1t2.test.ts src/tests/abort-silent-t4.test.ts src/tests/t3-close-tag-escape.test.ts src/tests/t5-tool-cap.test.ts src/tests/logger-redaction.test.ts src/tests/stall-timeout-retryable.test.ts src/tests/t6-missing-quote-array-close.test.ts src/tests/watchdog-rss.test.ts src/tests/logical-thread-batch.test.ts src/tests/chat-pool-style.test.ts src/tests/queue-wait-cap.test.ts src/tests/context-eviction-order.test.ts src/tests/chat-in-progress-settle.test.ts src/tests/stream-lock-leak.test.ts src/tests/stream-semaphore.test.ts src/tests/waf-isolation.test.ts src/tests/legacy-completions.test.ts src/tests/response-format.test.ts src/tests/cors-headers.test.ts src/tests/stream-lock-terminal-cancel.test.ts src/tests/parser-truncated-tool-call.test.ts src/tests/metrics-prometheus.test.ts src/tests/context-death-cleanup.test.ts src/tests/personalization-direct-fetch.test.ts src/tests/completion-headers-parity.test.ts src/tests/parallel-escape-repro.test.ts src/tests/chat-in-progress-scale-cost.test.ts src/tests/temp-chat-mode.test.ts src/tests/profile-reset.test.ts src/tests/error-classification.test.ts src/tests/quota-cooldown.test.ts src/tests/account-ready-gate.test.ts src/tests/tool-cap-notice.test.ts src/tests/interaction-flows.test.ts src/tests/headers-gate-cooldown-fallback.test.ts src/tests/unwrapped-tool-calls.test.ts src/tests/anthropic.test.ts src/tests/sync-clients.test.ts src/tests/cli.test.ts src/tests/paths.test.ts src/tests/tool-manifest.test.ts src/tests/tui.test.ts src/tests/tui-settings.test.ts src/tests/personalization-deadline.test.ts",
+    "test:mock:fast": "tsx --test --test-concurrency=2 --env-file=.env.test src/tests/account-concurrency.test.ts src/tests/concurrency-latest-wins.test.ts src/tests/mutex.test.ts src/tests/context-truncation.test.ts src/tests/streaming-thinking-summary.test.ts src/tests/tool-call-guard.test.ts src/tests/utils-logger-full.test.ts src/tests/logger-toolcall-errors.test.ts src/tests/logger-toolcall-disabled.test.ts src/tests/utils-json-full.test.ts src/tests/utils-session-guard-full.test.ts src/tests/core-logic-full.test.ts src/tests/core-logic-accounts.test.ts src/tests/chat-validation-full.test.ts src/tests/chat-logic-full.test.ts src/tests/account-stickiness.test.ts src/tests/accounts-security.test.ts src/tests/advanced.test.ts src/tests/auth-playwright.test.ts src/tests/captcha-solver.test.ts src/tests/cache-scan-regex.test.ts src/tests/chat-cleanup.test.ts src/tests/chat-cleanup-auto.test.ts src/tests/chromium-args.test.ts src/tests/config-current.test.ts src/tests/context-meter.test.ts src/tests/concurrency.test.ts src/tests/delta.test.ts src/tests/fingerprint.test.ts src/tests/human-behavior.test.ts src/tests/index.test.ts src/tests/json-recovery.test.ts src/tests/jsonFix.test.ts src/tests/memory-cache.test.ts src/tests/model-alias.test.ts src/tests/model-registry-thinking.test.ts src/tests/models-endpoint.test.ts src/tests/multimodal.test.ts src/tests/oss-upload.test.ts src/tests/parser.test.ts src/tests/personalization-flow.test.ts src/tests/personalization-required.test.ts src/tests/retry-policy.test.ts src/tests/rotation.test.ts src/tests/server-lifecycle.test.ts src/tests/session-keeper.test.ts src/tests/session-state.test.ts src/tests/single-account-mode.test.ts src/tests/stream-retry.test.ts src/tests/tool-calls-endpoint.test.ts src/tests/tool-instructions-cache.test.ts src/tests/session-id.test.ts src/tests/watchdog-ram.test.ts src/tests/tool-recovery-t1t2.test.ts src/tests/abort-silent-t4.test.ts src/tests/t3-close-tag-escape.test.ts src/tests/t5-tool-cap.test.ts src/tests/logger-redaction.test.ts src/tests/stall-timeout-retryable.test.ts src/tests/t6-missing-quote-array-close.test.ts src/tests/watchdog-rss.test.ts src/tests/logical-thread-batch.test.ts src/tests/chat-pool-style.test.ts src/tests/queue-wait-cap.test.ts src/tests/context-eviction-order.test.ts src/tests/chat-in-progress-settle.test.ts src/tests/stream-lock-leak.test.ts src/tests/stream-semaphore.test.ts src/tests/waf-isolation.test.ts src/tests/legacy-completions.test.ts src/tests/response-format.test.ts src/tests/cors-headers.test.ts src/tests/stream-lock-terminal-cancel.test.ts src/tests/parser-truncated-tool-call.test.ts src/tests/metrics-prometheus.test.ts src/tests/context-death-cleanup.test.ts src/tests/personalization-direct-fetch.test.ts src/tests/completion-headers-parity.test.ts src/tests/parallel-escape-repro.test.ts src/tests/chat-in-progress-scale-cost.test.ts src/tests/temp-chat-mode.test.ts src/tests/profile-reset.test.ts src/tests/error-classification.test.ts src/tests/quota-cooldown.test.ts src/tests/account-ready-gate.test.ts src/tests/tool-cap-notice.test.ts src/tests/interaction-flows.test.ts src/tests/headers-gate-cooldown-fallback.test.ts src/tests/unwrapped-tool-calls.test.ts src/tests/anthropic.test.ts src/tests/sync-clients.test.ts src/tests/cli.test.ts src/tests/paths.test.ts src/tests/tool-manifest.test.ts src/tests/tui.test.ts src/tests/tui-settings.test.ts src/tests/personalization-deadline.test.ts",
     "test:live": "tsx --test src/tests/concurrentChat.test.ts src/tests/agenticStress.test.ts",
     "benchmark:proxy": "npx tsx src/benchmarks/proxy-baseline.ts",
     "typecheck": "tsc --noEmit"

package/src/api/server.ts

@@ -22,6 +22,11 @@ import { AuthError, NotFoundError } from "../core/errors.js";
 import type { QwenAccount } from "../core/accounts.js";
 import { isAuthMockEnabled } from "../services/auth-playwright.js";
 
+import {
+  hookServerConsoleForLogging,
+  getServerLogHistory,
+  subscribeServerLogStream,
+} from "../core/server-log-buffer.ts";
 // Module-level state (initialized in startServer)
 let cache: MemoryCache | undefined;
 let watchdog: Watchdog | undefined;
@@ -131,12 +136,23 @@ app.use("*", async (c, next) => {
     String(Math.max(0, ratelimit.tokens - 1)),
   );
   c.header("x-ratelimit-reset-tokens", "0");
-
-  metrics.increment("requests.total");
+  const isProbe =
+    c.req.path === "/health" ||
+    c.req.path === "/metrics" ||
+    c.req.path === "/logs" ||
+    c.req.path.startsWith("/logs") ||
+    c.req.path.startsWith("/diagnostics") ||
+    c.req.path === "/favicon.ico";
+
+  if (!isProbe) {
+    metrics.increment("requests.total");
+  }
   const start = Date.now();
   await next();
   const duration = Date.now() - start;
-  metrics.histogram("latency.request", duration);
+  if (!isProbe) {
+    metrics.histogram("latency.request", duration);
+  }
   c.header("X-Response-Time", `${duration}ms`);
   c.header("openai-processing-ms", String(duration));
 });
@@ -272,6 +288,22 @@ app.get("/health", async (c) => {
     activeAccounts: (await import("../services/playwright.js")).getActivePlaywrightAccountIds(),
     metrics: {
       cache: await cache?.getStats(),
+      requestsTotal: Number(metrics.get("requests.total")?.value ?? 0),
+      requestsErrors: Number(metrics.get("requests.errors")?.value ?? 0),
+      latencyAvgMs: (() => {
+        const hist = metrics.get("latency.request")?.value;
+        if (hist && typeof hist === "object" && (hist as any).count > 0) {
+          return Math.round((hist as any).sum / (hist as any).count);
+        }
+        return 0;
+      })(),
+      deltasCount: Number(metrics.get("requests.delta")?.value ?? 0),
+      fullReplaysCount: Number(metrics.get("requests.full")?.value ?? 0),
+      toolCallsCount: Number(metrics.get("toolcalls.total")?.value ?? 0),
+      toolCallsRecovered: Number(metrics.get("toolcalls.recovered")?.value ?? 0),
+      captchasDetected: Number(metrics.get("captcha.challenges.detected")?.value ?? 0),
+      captchasSolved: Number(metrics.get("captcha.solves.succeeded")?.value ?? 0),
+      chatsCleaned: Number(metrics.get("chats.cleaned")?.value ?? 0),
     },
   });
 });
@@ -305,9 +337,54 @@ app.get("/metrics", (c) => {
   });
 });
 
+app.get("/logs", (c) => {
+  return c.json(getServerLogHistory());
+});
+
+app.get("/logs/live", (c) => {
+  const encoder = new TextEncoder();
+  return c.body(
+    new ReadableStream({
+      start(controller) {
+        const past = getServerLogHistory();
+        for (const entry of past) {
+          try {
+            controller.enqueue(encoder.encode(`data: ${JSON.stringify(entry)}\n\n`));
+          } catch {}
+        }
+
+        const unsubscribe = subscribeServerLogStream((entry) => {
+          try {
+            controller.enqueue(encoder.encode(`data: ${JSON.stringify(entry)}\n\n`));
+          } catch {
+            unsubscribe();
+          }
+        });
+
+        c.req.raw.signal.addEventListener("abort", () => {
+          unsubscribe();
+        });
+      },
+    }),
+    200,
+    {
+      "Content-Type": "text/event-stream",
+      "Cache-Control": "no-cache",
+      Connection: "keep-alive",
+    },
+  );
+});
+
 app.onError((err, c) => {
   const requestId = c.req.header("X-Request-Id") || "unknown";
-  metrics.increment("requests.errors");
+  const isProbe =
+    c.req.path === "/health" ||
+    c.req.path === "/metrics" ||
+    c.req.path.startsWith("/diagnostics") ||
+    c.req.path === "/favicon.ico";
+  if (!isProbe) {
+    metrics.increment("requests.errors");
+  }
   logger.error("API Error", {
     requestId,
     error: err instanceof Error ? err.message : String(err),
@@ -599,6 +676,7 @@ export async function startServer(options?: {
   installSignalHandlers?: boolean;
   showBanner?: boolean;
 }): Promise<StartedServerInfo> {
+  hookServerConsoleForLogging();
   if (server) {
     if (options?.installSignalHandlers !== false) installSignalHandlers();
     return buildStartedServerInfo();
@@ -814,6 +892,9 @@ export async function startServer(options?: {
       await import("../core/account-concurrency.ts");
     startLeaseSweepTimer();
 
+    const { scheduleStartupChatCleanup } =
+      await import("../services/chat-cleanup.ts");
+    scheduleStartupChatCleanup();
     const serverInstance = serve({
       fetch: app.fetch,
       port: config.server.port,

package/src/core/config.ts

@@ -161,7 +161,9 @@ const envSchema = z
     CONTEXT_METER_WINDOW_TOKENS: z.string().default("0"),
     CONTEXT_METER_REPORT_USAGE: z.string().default("true"),
     DELETE_ALL_CHATS_ON_SHUTDOWN: z.string().default("false"),
-    // Keep idle account pages alive (subtlePageActivity + occasional reload).
+    AUTO_CLEAN_CHATS_ON_STARTUP: z.string().default("true"),
+    AUTO_CLEAN_ORPHAN_CHATS: z.string().default("true"),
+    AUTO_CLEAN_CHAT_MAX_AGE_HOURS: z.string().default("24"),
     // The Baxia WAF scores live page behavior (pointer/scroll events, open
     // session) — an account whose page sits frozen for minutes returns a low
     // trust score and gets TMD-challenged on the next request. On by default;
@@ -367,7 +369,12 @@ export const config = {
       parseInt(env.QWEN_MAX_PERSONALIZATION_BYTES),
     ),
     deleteAllChatsOnShutdown: env.DELETE_ALL_CHATS_ON_SHUTDOWN === "true",
-    /** Send the captured bx-ua/bx-umidtoken headers (real client does NOT). */
+    autoCleanChatsOnStartup: env.AUTO_CLEAN_CHATS_ON_STARTUP !== "false",
+    autoCleanOrphanChats: env.AUTO_CLEAN_ORPHAN_CHATS !== "false",
+    autoCleanChatMaxAgeHours: Math.max(
+      1,
+      parseInt(env.AUTO_CLEAN_CHAT_MAX_AGE_HOURS) || 24,
+    ),
     sendBxUa: env.QWEN_SEND_BX_UA === "true",
     /** Deployed web bundle version sent as the `version` API header. */
     webVersion: env.QWEN_WEB_VERSION,

package/src/core/metrics.ts

@@ -32,8 +32,13 @@ export class Metrics extends EventEmitter {
       // Core request metrics
       ["requests.total", "counter", "Total requests processed"],
       ["requests.errors", "counter", "Total request errors"],
+      ["requests.delta", "counter", "Thread-native delta requests"],
+      ["requests.full", "counter", "Full context replay or initial requests"],
       ["latency.request", "histogram", "Request latency (ms)"],
-
+      ["toolcalls.total", "counter", "Total tool calls processed"],
+      ["toolcalls.malformed", "counter", "Malformed tool calls detected"],
+      ["toolcalls.recovered", "counter", "Malformed tool calls recovered/healed"],
+      ["chats.cleaned", "counter", "Remote chats automatically cleaned"],
       // Stream metrics
       ["streams.active", "gauge", "Active SSE streams"],
       ["streams.errors", "counter", "Stream errors"],

package/src/core/paths.ts

@@ -170,6 +170,13 @@ export function getSyncStatePath(customDataDir?: string): string {
   return path.join(customDataDir || getDataDir(), "sync-state.json");
 }
 
+/**
+ * Path to TUI user preferences file (model, effort, last tab, etc.).
+ */
+export function getTuiSettingsPath(customDataDir?: string): string {
+  return path.join(customDataDir || getDataDir(), "tui-settings.json");
+}
+
 /**
  * Path to user configuration .env file (either local or in global data dir).
  */

package/src/core/server-log-buffer.ts

@@ -0,0 +1,105 @@
+/**
+ * QwenProxy - Server Log Circular Buffer & Live SSE Broadcast
+ * Captures server logs and streams them to connected TUI instances.
+ */
+
+import { stripAnsi } from "../tui/theme.ts";
+
+export interface ServerLogMessage {
+  time: string;
+  level: "INFO" | "WARN" | "ERROR";
+  message: string;
+}
+
+const MAX_LOG_HISTORY = 400;
+const logHistory: ServerLogMessage[] = [];
+const subscribers = new Set<(entry: ServerLogMessage) => void>();
+let consoleHooked = false;
+
+export function recordServerLog(level: "INFO" | "WARN" | "ERROR", text: string): void {
+  if (!text) return;
+  const clean = stripAnsi(text).trim();
+  if (!clean) return;
+
+  const time = new Date().toLocaleTimeString("pt-BR", {
+    hour: "2-digit",
+    minute: "2-digit",
+    second: "2-digit",
+  });
+
+  const lines = clean.split(/\r?\n/);
+  for (const raw of lines) {
+    const line = raw.trim();
+    if (!line) continue;
+    // Filter out raw ASCII box frames and border rows
+    if (/^[+\-=#]{5,}$/.test(line)) continue;
+    if (/^\|\s*\|$/.test(line)) continue;
+    if (line.startsWith("|") && line.endsWith("|")) continue;
+
+    // Filter out box remnants
+    if (
+      line === "QwenProxy" ||
+      line === "OpenAI & Anthropic Compatible API" ||
+      /^Endpoint\s+http/i.test(line) ||
+      /^Port\s+\d+/i.test(line) ||
+      /^Accounts\s+\d+\/\d+/i.test(line) ||
+      /^API Key\s+/i.test(line) ||
+      /^Status\s+●/i.test(line)
+    ) {
+      continue;
+    }
+
+    const entry: ServerLogMessage = { time, level, message: line };
+    logHistory.push(entry);
+    if (logHistory.length > MAX_LOG_HISTORY) {
+      logHistory.shift();
+    }
+
+    for (const sub of subscribers) {
+      try {
+        sub(entry);
+      } catch {}
+    }
+  }
+}
+
+export function getServerLogHistory(): ServerLogMessage[] {
+  return logHistory.slice();
+}
+
+export function subscribeServerLogStream(cb: (entry: ServerLogMessage) => void): () => void {
+  subscribers.add(cb);
+  return () => {
+    subscribers.delete(cb);
+  };
+}
+
+export function hookServerConsoleForLogging(): void {
+  if (consoleHooked) return;
+  consoleHooked = true;
+
+  const origLog = console.log;
+  const origInfo = console.info;
+  const origWarn = console.warn;
+  const origError = console.error;
+
+  console.log = (...args: any[]) => {
+    origLog(...args);
+    recordServerLog("INFO", args.map((a) => (typeof a === "string" ? a : JSON.stringify(a))).join(" "));
+  };
+
+  console.info = (...args: any[]) => {
+    origInfo(...args);
+    recordServerLog("INFO", args.map((a) => (typeof a === "string" ? a : JSON.stringify(a))).join(" "));
+  };
+
+  console.warn = (...args: any[]) => {
+    origWarn(...args);
+    recordServerLog("WARN", args.map((a) => (typeof a === "string" ? a : JSON.stringify(a))).join(" "));
+  };
+
+  console.error = (...args: any[]) => {
+    origError(...args);
+    recordServerLog("ERROR", args.map((a) => (typeof a === "string" ? a : JSON.stringify(a))).join(" "));
+  };
+}

package/src/routes/chat/account.ts

@@ -36,6 +36,7 @@ import {
 } from "../../core/account-concurrency.ts";
 import { isAuthMockEnabled } from "../../services/auth-playwright.ts";
 import { isPlaywrightInitialized, refreshHeaders } from "../../services/playwright.ts";
+import { enqueueOrphanChatDeletion } from "../../services/chat-cleanup.ts";
 import {
 	clearAllSessionsForAccount,
 	createQwenStream,
@@ -1720,12 +1721,15 @@ async function tryCreateStreamWithRetry(
 				// Do NOT persist sticky binding until create succeeds — premature empty
 				// chatSessionId writes make subsequent turns rotate/lose context.
 				if (params.useThreadNative) {
+					const abandonedChatId = params.existingThread?.chatSessionId;
+					if (abandonedChatId) {
+						enqueueOrphanChatDeletion(currentAccountId, abandonedChatId, policy.reason);
+					}
 					params.existingThread = null;
 					params.finalPrompt = params.fullPrompt;
 					params.messageCount = params.fullMessageCount ?? params.messageCount;
 					params.forceNewChat = true;
 				}
-
 				await new Promise((resolve) =>
 					setTimeout(
 						resolve,
@@ -1749,12 +1753,15 @@ async function tryCreateStreamWithRetry(
 			console.warn(
 				`🔄 [Chat] Forcing new chat/full context | reason=${policy.reason}`,
 			);
+			const abandonedChatId = params.existingThread?.chatSessionId;
+			if (abandonedChatId) {
+				enqueueOrphanChatDeletion(currentAccountId, abandonedChatId, policy.reason);
+			}
 			params.existingThread = null;
 			params.finalPrompt = params.fullPrompt;
 			params.messageCount = params.fullMessageCount ?? params.messageCount;
 			params.forceNewChat = true;
 		}
-
 		// Drop files on retry for invalid_input to isolate file-related errors
 		if (policy.dropFiles && params.allFiles.length > 0) {
 			console.warn(

package/src/routes/chat/index.ts

@@ -22,6 +22,7 @@ import {
 } from "./streaming.ts";
 import { config, type ChatMode } from "../../core/config.ts";
 import { logger } from "../../core/logger.ts";
+import { metrics } from "../../core/metrics.ts";
 import { getContextMeterHeaders, type ContextMeterMode } from "../../services/context-meter.ts";
 import {
   getLogicalThreadState,
@@ -277,6 +278,11 @@ export async function chatCompletions(c: Context) {
     console.log(
       `📤 [${routeLabel}] Request | req=${reqId} | ${streamResult.activeAccountLabel} | ${body.model} | ${replayed ? parsed.messageCount : msgCount} msg(s) | ${replayed ? fullPromptForRequest.length : finalPrompt.length} chars${replayed ? " | full-replay" : ""} | chat=${streamResult.uiSessionId.substring(0, 12)}${declaredTools.length ? ` | ${declaredTools.length} tool(s)` : ""}${files.length ? ` | ${files.length} file(s)` : ""} | +${Date.now() - reqStartedAt}ms`,
     );
+    if (replayed || !ctx.existingThread) {
+      metrics.increment("requests.full");
+    } else {
+      metrics.increment("requests.delta");
+    }
 
     const onAssistantComplete: ((event: AssistantCompleteEvent) => Promise<void> | void) | undefined = undefined;
 

package/src/services/chat-cleanup.ts

@@ -3,14 +3,25 @@ import {
   loadAccounts,
   type QwenAccount,
 } from "../core/accounts.ts";
-import { deleteAllQwenChats } from "./qwen.ts";
+import {
+  deleteAllQwenChats,
+  deleteSingleQwenChat,
+  fetchRemoteQwenChats,
+} from "./qwen.ts";
 import {
   initPlaywrightForAccount,
   isPlaywrightInitialized,
   closeAllPlaywright,
+  getActivePlaywrightAccountIds,
 } from "./playwright.ts";
 import { isAuthMockEnabled } from "./auth-playwright.ts";
-import { maskEmail } from "../core/logger.ts";
+import { maskEmail, logger } from "../core/logger.ts";
+import { config } from "../core/config.ts";
+import { isAccountBusy } from "../core/account-concurrency.ts";
+import { isChatSessionActive } from "./qwen-thread-state.ts";
+import { sleep } from "./human-behavior.ts";
+import { metrics } from "../core/metrics.ts";
+
 export interface DeleteChatsResult {
   attempted: number;
   succeeded: number;
@@ -109,10 +120,171 @@ export async function deleteChatsForConfiguredAccounts(keepBrowserOpen = false):
       });
     }
   }
-
   return {
     attempted: accounts.length,
     succeeded,
     mode: "accounts",
   };
 }
+
+interface OrphanChatEntry {
+  accountId: string;
+  chatId: string;
+  reason: string;
+  enqueuedAt: number;
+}
+
+const orphanChatQueue: OrphanChatEntry[] = [];
+let isProcessingQueue = false;
+
+export function getOrphanQueueLengthForTests(): number {
+  return orphanChatQueue.length;
+}
+
+/**
+ * Enqueues an abandoned/orphan chat for asynchronous deletion in background.
+ * Never blocks the calling request and never deletes a chat actively in use.
+ */
+export function enqueueOrphanChatDeletion(
+  accountId: string,
+  chatId: string,
+  reason = "orphan",
+): void {
+  if (!config.qwen.autoCleanOrphanChats || !accountId || !chatId) return;
+  if (orphanChatQueue.some((e) => e.accountId === accountId && e.chatId === chatId)) {
+    return;
+  }
+  orphanChatQueue.push({
+    accountId,
+    chatId,
+    reason,
+    enqueuedAt: Date.now(),
+  });
+
+  if (!isProcessingQueue) {
+    const timer = setTimeout(() => {
+      void processOrphanChatQueue().catch(() => {});
+    }, 1_000);
+    timer.unref?.();
+  }
+}
+
+async function processOrphanChatQueue(): Promise<void> {
+  if (isProcessingQueue) return;
+  isProcessingQueue = true;
+  try {
+    while (orphanChatQueue.length > 0) {
+      const entry = orphanChatQueue[0];
+
+      // Never delete a chat that is currently actively bound to an ongoing session
+      if (isChatSessionActive(entry.chatId)) {
+        orphanChatQueue.shift();
+        continue;
+      }
+
+      // If the account is actively serving a stream, postpone deletion
+      if (!isAuthMockEnabled() && isAccountBusy(entry.accountId)) {
+        await sleep(2_000);
+        continue;
+      }
+
+      try {
+        const ok = await deleteSingleQwenChat(entry.accountId, entry.chatId);
+        if (ok) {
+          metrics.increment("chats.cleaned");
+          logger.debug("[ChatCleanup] Deleted orphan chat", {
+            accountId: entry.accountId,
+            chatId: entry.chatId,
+            reason: entry.reason,
+          });
+        }
+      } catch (err) {
+        logger.debug("[ChatCleanup] Failed to delete orphan chat", {
+          accountId: entry.accountId,
+          chatId: entry.chatId,
+          error: err instanceof Error ? err.message : String(err),
+        });
+      }
+
+      orphanChatQueue.shift();
+      if (orphanChatQueue.length > 0) {
+        await sleep(300);
+      }
+    }
+  } finally {
+    isProcessingQueue = false;
+  }
+}
+
+export async function drainOrphanChatQueueForTests(): Promise<void> {
+  await processOrphanChatQueue();
+}
+
+/**
+ * Scan an account's remote chat list and delete chats older than maxAgeHours,
+ * skipping any chat that is currently actively bound to a session.
+ */
+export async function cleanOldChatsForAccount(
+  accountId: string,
+  maxAgeHours = config.qwen.autoCleanChatMaxAgeHours,
+): Promise<number> {
+  try {
+    const chats = await fetchRemoteQwenChats(accountId);
+    if (!Array.isArray(chats) || chats.length === 0) return 0;
+
+    const maxAgeMs = maxAgeHours * 60 * 60 * 1000;
+    const now = Date.now();
+    let cleaned = 0;
+
+    for (const chat of chats) {
+      if (!chat?.id) continue;
+      if (isChatSessionActive(chat.id)) continue;
+
+      const updatedMs =
+        typeof chat.updated_at === "number"
+          ? (chat.updated_at > 1e11 ? chat.updated_at : chat.updated_at * 1000)
+          : new Date(chat.updated_at).getTime();
+
+      if (now - updatedMs >= maxAgeMs) {
+        const ok = await deleteSingleQwenChat(accountId, chat.id);
+        if (ok) {
+          cleaned++;
+          metrics.increment("chats.cleaned");
+          await sleep(300);
+        }
+      }
+    }
+
+    if (cleaned > 0) {
+      console.log(
+        `🧹 [ChatCleanup] Remotely cleaned ${cleaned} old chat(s) (> ${maxAgeHours}h) for account ${accountId}`,
+      );
+    }
+    return cleaned;
+  } catch (error) {
+    logger.debug("[ChatCleanup] cleanOldChatsForAccount error", {
+      accountId,
+      error: error instanceof Error ? error.message : String(error),
+    });
+    return 0;
+  }
+}
+
+/**
+ * Startup hook: runs 15s after startup in the background, cleaning old chats
+ * for active accounts without holding up the server or user requests.
+ */
+export function scheduleStartupChatCleanup(): void {
+  if (!config.qwen.autoCleanChatsOnStartup || isAuthMockEnabled()) return;
+
+  const timer = setTimeout(async () => {
+    try {
+      const activeIds = getActivePlaywrightAccountIds();
+      for (const accountId of activeIds) {
+        await cleanOldChatsForAccount(accountId);
+        await sleep(1_000);
+      }
+    } catch {}
+  }, 15_000);
+  timer.unref?.();
+}

package/src/services/qwen-thread-state.ts

@@ -376,6 +376,43 @@ export function clearAllSessionsForAccount(accountId: string): void {
   );
 }
 
+/**
+ * Returns true if a chatSessionId is currently actively bound to any non-expired session.
+ * Prevents auto-cleanup routines from deleting conversations currently in use.
+ */
+export function isChatSessionActive(chatId: string): boolean {
+  if (!chatId) return false;
+  const now = Date.now();
+  for (const entry of logicalThreadStates.values()) {
+    if (entry.chatSessionId === chatId && now - entry.timestamp <= SESSION_TTL_MS) {
+      return true;
+    }
+  }
+  const sessionEntry = sessionStates.get(chatId);
+  if (sessionEntry && now - sessionEntry.timestamp <= SESSION_TTL_MS) {
+    return true;
+  }
+  return false;
+}
+
+/**
+ * Invalidate and remove all local cached/database records for a given chatId.
+ */
+export function removeSessionByChatId(chatId: string): void {
+  if (!chatId) return;
+  sessionStates.delete(chatId);
+  for (const [key, entry] of logicalThreadStates.entries()) {
+    if (entry.chatSessionId === chatId) {
+      logicalThreadStates.delete(key);
+      logicalThreadDirty.delete(key);
+    }
+  }
+  try {
+    const db = getDatabase();
+    db.prepare("DELETE FROM logical_thread_states WHERE chat_session_id = ?").run(chatId);
+  } catch {}
+}
+
 export function getSessionParent(
   sessionId: string,
   accountId?: string,