fix(core): retry on failed as well as timeout; only auth stops the run (#104)

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

packages/cli/src/i18n.ts

@@ -153,8 +153,12 @@ export interface Messages {
   }): string;
   /** Abort event label (may include a reason). */
   abortLabel(reason?: string): string;
-  /** request_end ended with timeout/malformed: the engine retries (reconnect) carrying already-produced content; attempt is the retry count. */
-  reconnectLabel(status: "timeout" | "malformed", attempt: number): string;
+  /**
+   * request_end ended with a status the engine reconnects on (`failed` / `timeout` /
+   * `malformed` — only `auth` is terminal): the engine retries carrying already-produced
+   * content; attempt is the retry count.
+   */
+  reconnectLabel(status: "failed" | "timeout" | "malformed", attempt: number): string;
   /** compaction start event: indicates compaction in progress (mode is summarize/discard, reason is context/turns/manual). */
   compactionStart(mode: string, reason: string): string;
   /**
@@ -383,7 +387,13 @@ const en: Messages = {
     `[stats] context ${s.context} (${s.contextDelta}) · tokens ${s.tokens} (${s.tokensDelta}) · ${s.elapsed} (${s.elapsedDelta})`,
   abortLabel: (reason) => `[abort]${reason ? `: ${reason}` : ""}`,
   reconnectLabel: (status, attempt) =>
-    `[retry] ${status === "timeout" ? "connection timed out" : "response incomplete or unparseable"}; sending retry #${attempt}…`,
+    `[retry] ${
+      status === "timeout"
+        ? "connection timed out"
+        : status === "malformed"
+          ? "response incomplete or unparseable"
+          : "the model provider returned an error"
+    }; sending retry #${attempt}…`,
   compactionStart: (mode, reason) =>
     mode === "discard"
       ? `[compaction] discarding context (${reason})…`
@@ -571,7 +581,13 @@ const zh: Messages = {
     `[统计信息] 上下文 ${s.context} (${s.contextDelta}) · tokens ${s.tokens} (${s.tokensDelta}) · 用时 ${s.elapsed} (${s.elapsedDelta})`,
   abortLabel: (reason) => `[已中断]${reason ? `：${reason}` : ""}`,
   reconnectLabel: (status, attempt) =>
-    `[重试] ${status === "timeout" ? "连接超时或网络中断" : "响应不完整或无法解析"}，正在发起第 ${attempt} 次重试……`,
+    `[重试] ${
+      status === "timeout"
+        ? "连接超时或网络中断"
+        : status === "malformed"
+          ? "响应不完整或无法解析"
+          : "模型服务返回错误"
+    }，正在发起第 ${attempt} 次重试……`,
   compactionStart: (mode, reason) =>
     mode === "discard"
       ? `[压缩] 正在丢弃旧上下文（${reason}）……`

packages/cli/src/render.ts

@@ -378,8 +378,8 @@ export class StreamRenderer {
    */
   private taskFirstTsMs: number | null = null;
   private taskLastReqEndMs: number | null = null;
-  /** Terminal state (timeout/malformed) of the previous request: the next request_begin is a retry, at which point a notice is printed. */
-  private pendingRetry: "timeout" | "malformed" | null = null;
+  /** Retryable terminal state (failed/timeout/malformed) of the previous request: the next request_begin is a retry, at which point a notice is printed. */
+  private pendingRetry: "failed" | "timeout" | "malformed" | null = null;
   /** Number of retries already initiated (increments on consecutive failures, reset once a request completes normally). */
   private reconnectRun = 0;
 
@@ -693,7 +693,7 @@ export class StreamRenderer {
         this.out.write(`${formatAbort(payload as AbortPayload, this.t)}\n`);
         this.lastLineKey = null;
       } else if (payload.type === "request_begin") {
-        // The previous request ended in timeout/malformed -> this request is a retry
+        // The previous request ended in a retryable status -> this request is a retry
         // carrying [turn_retried]: printed when the retry **actually starts** (when
         // retries are exhausted, there's no retry after the last failure, only an abort
         // explaining why).
@@ -720,7 +720,10 @@ export class StreamRenderer {
             this.hasUsage = true;
           }
         }
-        if (p.status === "timeout" || p.status === "malformed") {
+        // Every status the engine reconnects on, `failed` included — only `auth` is terminal.
+        // Leaving `failed` out would print nothing for a retry that is really happening, and
+        // reset the counter mid-ladder so a mixed run renumbers back to retry #1.
+        if (p.status === "failed" || p.status === "timeout" || p.status === "malformed") {
           this.pendingRetry = p.status;
         } else {
           this.pendingRetry = null;