@ynhcj/xiaoyi-channel 0.0.91-beta → 0.0.92-beta

package/dist/src/provider.js

@@ -9,6 +9,60 @@
 //   models.providers.xiaoyiprovider.models = [...]
 import { createHash } from "crypto";
 import { getCurrentSessionContext } from "./tools/session-manager.js";
+// ── Retry config ──────────────────────────────────────────────
+const RETRY_DELAYS_MS = [10_000, 20_000, 40_000, 60_000];
+const MAX_RETRY_ATTEMPTS = 8;
+/** Check if an errorMessage indicates a retryable provider error by type. */
+function isRetryableProviderError(message) {
+    if (!message)
+        return false;
+    const lower = message.toLowerCase();
+    if (lower.includes("server_error"))
+        return true;
+    if (lower.includes("rate_limit_error"))
+        return true;
+    return false;
+}
+/** Compute retry delay in ms for the given 1-based attempt. */
+function getRetryDelayMs(attempt) {
+    // attempt 1→10s, 2→20s, 3→40s, 4+→60s
+    if (attempt <= RETRY_DELAYS_MS.length)
+        return RETRY_DELAYS_MS[attempt - 1];
+    return RETRY_DELAYS_MS[RETRY_DELAYS_MS.length - 1]; // 60s cap
+}
+function sleep(ms) {
+    return new Promise((resolve) => setTimeout(resolve, ms));
+}
+/**
+ * Build a minimal EventStream-compatible object that replays a single
+ * done/error event. This avoids importing @mariozechner/pi-ai at runtime
+ * (the package is not available in the extension sandbox).
+ */
+function buildReplayStream(result) {
+    let settled = false;
+    const queued = [
+        result.stopReason === "error"
+            ? { type: "error", reason: "error", error: result }
+            : { type: "done", reason: result.stopReason, message: result },
+    ];
+    return {
+        result: () => Promise.resolve(result),
+        push: () => { },
+        end: () => { },
+        [Symbol.asyncIterator]: () => {
+            return {
+                next: async () => {
+                    if (settled || queued.length === 0) {
+                        settled = true;
+                        return { value: undefined, done: true };
+                    }
+                    settled = true;
+                    return { value: queued.shift(), done: false };
+                },
+            };
+        },
+    };
+}
 /**
  * Dynamic header keys injected via extraParams and forwarded to the HTTP request.
  * Correspond to the three fields written to .xiaoyiruntime:
@@ -73,10 +127,12 @@ export const xiaoyiProvider = {
     },
     /**
      * Wrap the stream function to inject dynamic headers into every
-     * HTTP request to the model provider.
+     * HTTP request to the model provider, and retry on retryable errors
+     * (server_error / rate_limit_error) with backoff: 10s, 20s, 40s, 60s (cap).
      *
-     * Reads the values injected by prepareExtraParams and adds them
-     * as HTTP headers on the outgoing request.
+     * The retry loop awaits stream.result() to detect errors before deciding
+     * whether to retry. This keeps the agent loop waiting (no timeout risk
+     * since the default agent timeout is 48 hours).
      */
     wrapStreamFn: (ctx) => {
         const underlying = ctx.streamFn;
@@ -135,21 +191,51 @@ export const xiaoyiProvider = {
             if (sessionCtx?.deviceType) {
                 const rawDevice = sessionCtx.deviceType;
                 const displayDevice = (rawDevice === "2in1") ? "鸿蒙PC" : rawDevice;
-                const deviceSection = `\n\n## Current User Device Context\nThe current user is using the following device: ${displayDevice}\nYou need to be aware of the user’s current device and provide guidance accordingly. If the response involves device-related tools or actions, you must tailor the reply based on the user’s current device, using device-specific references such as “saved to the Notes/Calendar on your {deviceType}.\n”`;
+                const deviceSection = `\n\n## Current User Device Context\nThe current user is using the following device: ${displayDevice}\nYou need to be aware of the user's current device and provide guidance accordingly. If the response involves device-related tools or actions, you must tailor the reply based on the user's current device, using device-specific references such as "saved to the Notes/Calendar on your {deviceType}.\n"`;
                 context.systemPrompt = (context.systemPrompt ?? "") + deviceSection;
             }
-            const stream = await underlying(model, context, {
+            // ── Retry loop ─────────────────────────────────────────
+            for (let attempt = 0; attempt < MAX_RETRY_ATTEMPTS; attempt++) {
+                const stream = await underlying(model, context, {
+                    ...options,
+                    headers: {
+                        ...options?.headers,
+                        ...dynamicHeaders,
+                    },
+                });
+                // Wait for the stream to settle (done or error) to inspect the result.
+                // stream.result() resolves to the final AssistantMessage (even on error).
+                const result = await stream.result();
+                // Check if this is a retryable error
+                if (result.stopReason === "error" && isRetryableProviderError(result.errorMessage)) {
+                    const delayMs = getRetryDelayMs(attempt);
+                    console.log(`[xiaoyiprovider] retryable error (attempt ${attempt + 1}/${MAX_RETRY_ATTEMPTS}): ` +
+                        `${result.errorMessage} — retrying in ${delayMs}ms`);
+                    await sleep(delayMs);
+                    continue;
+                }
+                // Success or non-retryable error — log and return
+                if (result.stopReason === "error") {
+                    console.log(`[xiaoyiprovider] non-retryable error: ${result.errorMessage}`);
+                }
+                else {
+                    console.log(`[xiaoyiprovider] stream completed, usage: input=${result.usage?.input} output=${result.usage?.output}`);
+                }
+                // The original stream has already been consumed by result().
+                // Build a replay stream that delivers the final result.
+                return buildReplayStream(result);
+            }
+            // All retries exhausted — return the last attempt's real error via a new stream
+            console.log(`[xiaoyiprovider] all ${MAX_RETRY_ATTEMPTS} retries exhausted, surfacing last error`);
+            const lastStream = await underlying(model, context, {
                 ...options,
                 headers: {
                     ...options?.headers,
                     ...dynamicHeaders,
                 },
             });
-            // 异步监听输出（不阻塞 stream 返回）
-            stream.result().then((result) => {
-                console.log(`[xiaoyiprovider] stream completed, usage: input=${result.usage?.input} output=${result.usage?.output}`);
-            }, (err) => console.log(`[xiaoyiprovider] stream error: ${JSON.stringify(err)}`));
-            return stream;
+            const lastResult = await lastStream.result();
+            return buildReplayStream(lastResult);
         };
     },
 };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ynhcj/xiaoyi-channel",
-  "version": "0.0.91-beta",
+  "version": "0.0.92-beta",
   "description": "OpenClaw Xiaoyi Channel plugin - Xiaoyi A2A protocol integration",
   "type": "module",
   "main": "./dist/index.js",