Improve Zep retry handling and frontend API proxy

backend/app/services/zep_tools.py

@@ -10,10 +10,12 @@
 
 import time
 import json
+import re
 from typing import Dict, Any, List, Optional
 from dataclasses import dataclass, field
 
 from zep_cloud.client import Zep
+from zep_cloud.core.api_error import ApiError
 
 from ..config import Config
 from ..utils.logger import get_logger
@@ -23,6 +25,55 @@
 
 logger = get_logger('mirofish.zep_tools')
 
+_MAX_RATE_LIMIT_DELAY = 60.0
+
+
+def _is_rate_limit_error(error: Exception) -> bool:
+    status_code = getattr(error, "status_code", None)
+    if status_code == 429:
+        return True
+
+    body = getattr(error, "body", None)
+    if isinstance(body, str) and "rate limit" in body.lower():
+        return True
+
+    text = str(error).lower()
+    return "status_code: 429" in text or "rate limit" in text or "too many requests" in text
+
+
+def _parse_retry_after(error: Exception) -> float | None:
+    headers = getattr(error, "headers", None)
+    if isinstance(headers, dict):
+        retry_after = headers.get("retry-after") or headers.get("Retry-After")
+        if retry_after:
+            try:
+                return max(float(retry_after), 0.0)
+            except (TypeError, ValueError):
+                pass
+
+        reset = headers.get("x-ratelimit-reset") or headers.get("X-RateLimit-Reset")
+        if reset:
+            try:
+                reset_seconds = float(reset)
+                if reset_seconds > 1_000_000_000:
+                    wait_seconds = reset_seconds - time.time()
+                else:
+                    wait_seconds = reset_seconds
+                if wait_seconds > 0:
+                    return wait_seconds
+            except (TypeError, ValueError):
+                pass
+
+    text = str(error)
+    match = re.search(r"retry-after['\"]?\s*:\s*['\"]?(\d+(?:\.\d+)?)", text, re.IGNORECASE)
+    if match:
+        try:
+            return max(float(match.group(1)), 0.0)
+        except (TypeError, ValueError):
+            pass
+
+    return None
+
 
 @dataclass
 class SearchResult:
@@ -448,14 +499,32 @@ def _call_with_retry(self, func, operation_name: str, max_retries: int = None):
         for attempt in range(max_retries):
             try:
                 return func()
-            except Exception as e:
+            except (Exception, ApiError) as e:
                 last_exception = e
                 if attempt < max_retries - 1:
-                    logger.warning(
-                        t("console.zepRetryAttempt", operation=operation_name, attempt=attempt + 1, error=str(e)[:100], delay=f"{delay:.1f}")
-                    )
+                    if _is_rate_limit_error(e):
+                        retry_after = _parse_retry_after(e)
+                        if retry_after is not None:
+                            delay = min(max(retry_after, self.RETRY_DELAY), _MAX_RATE_LIMIT_DELAY)
+                        else:
+                            delay = min(max(delay, self.RETRY_DELAY), _MAX_RATE_LIMIT_DELAY)
+                    else:
+                        delay = min(delay, _MAX_RATE_LIMIT_DELAY)
+
+                    if _is_rate_limit_error(e):
+                        logger.warning(
+                            f"Zep {operation_name} rate limit hit (attempt {attempt + 1}/{max_retries}); "
+                            f"retrying in {delay:.1f}s..."
+                        )
+                    else:
+                        logger.warning(
+                            t("console.zepRetryAttempt", operation=operation_name, attempt=attempt + 1, error=str(e)[:100], delay=f"{delay:.1f}")
+                        )
                     time.sleep(delay)
-                    delay *= 2
+                    if _is_rate_limit_error(e):
+                        delay = min(delay * 1.25, _MAX_RATE_LIMIT_DELAY)
+                    else:
+                        delay = min(delay * 2, _MAX_RATE_LIMIT_DELAY)
                 else:
                     logger.error(t("console.zepAllRetriesFailed", operation=operation_name, retries=max_retries, error=str(e)))
 

backend/app/utils/zep_paging.py

@@ -7,10 +7,12 @@
 from __future__ import annotations
 
 import time
+import re
 from collections.abc import Callable
 from typing import Any
 
 from zep_cloud import InternalServerError
+from zep_cloud.core.api_error import ApiError
 from zep_cloud.client import Zep
 
 from .logger import get_logger
@@ -19,8 +21,57 @@
 
 _DEFAULT_PAGE_SIZE = 100
 _MAX_NODES = 2000
-_DEFAULT_MAX_RETRIES = 3
+_DEFAULT_MAX_RETRIES = 5
 _DEFAULT_RETRY_DELAY = 2.0  # seconds, doubles each retry
+_MAX_RETRY_DELAY = 60.0
+
+
+def _is_rate_limit_error(error: Exception) -> bool:
+    status_code = getattr(error, "status_code", None)
+    if status_code == 429:
+        return True
+
+    body = getattr(error, "body", None)
+    if isinstance(body, str) and "rate limit" in body.lower():
+        return True
+
+    text = str(error).lower()
+    return "status_code: 429" in text or "rate limit" in text or "too many requests" in text
+
+
+def _parse_retry_after(error: Exception) -> float | None:
+    headers = getattr(error, "headers", None)
+    if isinstance(headers, dict):
+        retry_after = headers.get("retry-after") or headers.get("Retry-After")
+        if retry_after:
+            try:
+                return max(float(retry_after), 0.0)
+            except (TypeError, ValueError):
+                pass
+
+        reset = headers.get("x-ratelimit-reset") or headers.get("X-RateLimit-Reset")
+        if reset:
+            try:
+                reset_seconds = float(reset)
+                if reset_seconds > 1_000_000_000:
+                    # Some providers return a unix timestamp instead of a delta.
+                    wait_seconds = reset_seconds - time.time()
+                else:
+                    wait_seconds = reset_seconds
+                if wait_seconds > 0:
+                    return wait_seconds
+            except (TypeError, ValueError):
+                pass
+
+    text = str(error)
+    match = re.search(r"retry-after['\"]?\s*:\s*['\"]?(\d+(?:\.\d+)?)", text, re.IGNORECASE)
+    if match:
+        try:
+            return max(float(match.group(1)), 0.0)
+        except (TypeError, ValueError):
+            pass
+
+    return None
 
 
 def _fetch_page_with_retry(
@@ -41,14 +92,33 @@ def _fetch_page_with_retry(
     for attempt in range(max_retries):
         try:
             return api_call(*args, **kwargs)
-        except (ConnectionError, TimeoutError, OSError, InternalServerError) as e:
+        except (ConnectionError, TimeoutError, OSError, InternalServerError, ApiError) as e:
             last_exception = e
             if attempt < max_retries - 1:
-                logger.warning(
-                    f"Zep {page_description} attempt {attempt + 1} failed: {str(e)[:100]}, retrying in {delay:.1f}s..."
-                )
+                if _is_rate_limit_error(e):
+                    retry_after = _parse_retry_after(e)
+                    if retry_after is not None:
+                        delay = min(max(retry_after, retry_delay), _MAX_RETRY_DELAY)
+                    else:
+                        delay = min(max(delay, retry_delay), _MAX_RETRY_DELAY)
+                else:
+                    delay = min(delay, _MAX_RETRY_DELAY)
+
+                if _is_rate_limit_error(e):
+                    logger.warning(
+                        f"Zep {page_description} rate limit hit (attempt {attempt + 1}/{max_retries}); "
+                        f"retrying in {delay:.1f}s..."
+                    )
+                else:
+                    logger.warning(
+                        f"Zep {page_description} attempt {attempt + 1} failed: {str(e)[:100]}, retrying in {delay:.1f}s..."
+                    )
                 time.sleep(delay)
-                delay *= 2
+                if _is_rate_limit_error(e):
+                    # Respect server-advised retry delays; keep the same delay or back off slightly.
+                    delay = min(delay * 1.25, _MAX_RETRY_DELAY)
+                else:
+                    delay = min(delay * 2, _MAX_RETRY_DELAY)
             else:
                 logger.error(f"Zep {page_description} failed after {max_retries} attempts: {str(e)}")
 

frontend/src/api/graph.js

@@ -8,12 +8,9 @@ import service, { requestWithRetry } from './index'
 export function generateOntology(formData) {
   return requestWithRetry(() => 
     service({
-      url: '/api/graph/ontology/generate',
+      url: '/graph/ontology/generate',
       method: 'post',
-      data: formData,
-      headers: {
-        'Content-Type': 'multipart/form-data'
-      }
+      data: formData
     })
   )
 }
@@ -26,7 +23,7 @@ export function generateOntology(formData) {
 export function buildGraph(data) {
   return requestWithRetry(() =>
     service({
-      url: '/api/graph/build',
+      url: '/graph/build',
       method: 'post',
       data
     })
@@ -40,7 +37,7 @@ export function buildGraph(data) {
  */
 export function getTaskStatus(taskId) {
   return service({
-    url: `/api/graph/task/${taskId}`,
+    url: `/graph/task/${taskId}`,
     method: 'get'
   })
 }
@@ -52,7 +49,7 @@ export function getTaskStatus(taskId) {
  */
 export function getGraphData(graphId) {
   return service({
-    url: `/api/graph/data/${graphId}`,
+    url: `/graph/data/${graphId}`,
     method: 'get'
   })
 }
@@ -64,7 +61,7 @@ export function getGraphData(graphId) {
  */
 export function getProject(projectId) {
   return service({
-    url: `/api/graph/project/${projectId}`,
+    url: `/graph/project/${projectId}`,
     method: 'get'
   })
 }

frontend/src/api/index.js

@@ -3,17 +3,19 @@ import i18n from '../i18n'
 
 // 创建axios实例
 const service = axios.create({
-  baseURL: import.meta.env.VITE_API_BASE_URL || 'http://localhost:5001',
-  timeout: 300000, // 5分钟超时（本体生成可能需要较长时间）
-  headers: {
-    'Content-Type': 'application/json'
-  }
+  // Default to the Vite proxy so the frontend works both on localhost
+  // and when accessed from another machine via the dev server host.
+  baseURL: import.meta.env.VITE_API_BASE_URL || '/api',
+  timeout: 300000 // 5分钟超时（本体生成可能需要较长时间）
 })
 
 // 请求拦截器
 service.interceptors.request.use(
   config => {
     config.headers['Accept-Language'] = i18n.global.locale.value
+    if (config.data instanceof FormData) {
+      delete config.headers['Content-Type']
+    }
     return config
   },
   error => {

frontend/src/api/report.js

@@ -5,15 +5,15 @@ import service, { requestWithRetry } from './index'
  * @param {Object} data - { simulation_id, force_regenerate? }
  */
 export const generateReport = (data) => {
-  return requestWithRetry(() => service.post('/api/report/generate', data), 3, 1000)
+  return requestWithRetry(() => service.post('/report/generate', data), 3, 1000)
 }
 
 /**
  * 获取报告生成状态
  * @param {string} reportId
  */
 export const getReportStatus = (reportId) => {
-  return service.get(`/api/report/generate/status`, { params: { report_id: reportId } })
+  return service.get(`/report/generate/status`, { params: { report_id: reportId } })
 }
 
 /**
@@ -22,7 +22,7 @@ export const getReportStatus = (reportId) => {
  * @param {number} fromLine - 从第几行开始获取
  */
 export const getAgentLog = (reportId, fromLine = 0) => {
-  return service.get(`/api/report/${reportId}/agent-log`, { params: { from_line: fromLine } })
+  return service.get(`/report/${reportId}/agent-log`, { params: { from_line: fromLine } })
 }
 
 /**
@@ -31,21 +31,21 @@ export const getAgentLog = (reportId, fromLine = 0) => {
  * @param {number} fromLine - 从第几行开始获取
  */
 export const getConsoleLog = (reportId, fromLine = 0) => {
-  return service.get(`/api/report/${reportId}/console-log`, { params: { from_line: fromLine } })
+  return service.get(`/report/${reportId}/console-log`, { params: { from_line: fromLine } })
 }
 
 /**
  * 获取报告详情
  * @param {string} reportId
  */
 export const getReport = (reportId) => {
-  return service.get(`/api/report/${reportId}`)
+  return service.get(`/report/${reportId}`)
 }
 
 /**
  * 与 Report Agent 对话
  * @param {Object} data - { simulation_id, message, chat_history? }
  */
 export const chatWithReport = (data) => {
-  return requestWithRetry(() => service.post('/api/report/chat', data), 3, 1000)
+  return requestWithRetry(() => service.post('/report/chat', data), 3, 1000)
 }