fix: cherry-pick 5 high-priority fixes from upstream PRs

- fix(runner): prevent action data loss from partial line reads (666ghj#460)
  Track safe_position; only advance past newline-terminated lines to
  avoid permanent data loss when the writer is mid-flush.

- fix(build): restrict Python to <3.13 to avoid tiktoken failure (666ghj#453)
  tiktoken wheels are not available for Python 3.13+.

- fix(security): default FLASK_DEBUG to False (666ghj#445)
  Debug mode was on by default, exposing the Werkzeug debugger in
  production deployments on 0.0.0.0.

- fix(polling): stop frontend polling when backend fails (666ghj#448)
  Step2 and Step3 now detect failed status and error-count limits,
  stopping the polling interval instead of looping indefinitely.

- fix(simulation): stop Step2 hanging after prepare failures (666ghj#336)
  Reset state flags on each prepare run; raise ValueError instead of
  returning silently so the task manager marks the task as failed.

Co-Authored-By: koyouko <koyouko@users.noreply.github.com>

Author: koyouko

backend/app/config.py

@@ -22,7 +22,7 @@ class Config:
 
     # ── Flask ────────────────────────────────────────────────────────────────
     SECRET_KEY = os.environ.get('SECRET_KEY', 'mirofish-secret-key')
-    DEBUG = os.environ.get('FLASK_DEBUG', 'True').lower() == 'true'
+    DEBUG = os.environ.get('FLASK_DEBUG', 'False').lower() == 'true'
 
     # Disable ASCII escaping so non-ASCII characters display correctly in JSON
     JSON_AS_ASCII = False

backend/app/services/simulation_manager.py

@@ -59,6 +59,7 @@ class SimulationState:
     entity_types: List[str] = field(default_factory=list)
 
     # 配置生成信息
+    profiles_generated: bool = False
     config_generated: bool = False
     config_reasoning: str = ""
 
@@ -86,6 +87,7 @@ def to_dict(self) -> Dict[str, Any]:
             "entities_count": self.entities_count,
             "profiles_count": self.profiles_count,
             "entity_types": self.entity_types,
+            "profiles_generated": self.profiles_generated,
             "config_generated": self.config_generated,
             "config_reasoning": self.config_reasoning,
             "current_round": self.current_round,
@@ -106,6 +108,7 @@ def to_simple_dict(self) -> Dict[str, Any]:
             "entities_count": self.entities_count,
             "profiles_count": self.profiles_count,
             "entity_types": self.entity_types,
+            "profiles_generated": self.profiles_generated,
             "config_generated": self.config_generated,
             "error": self.error,
         }
@@ -177,6 +180,7 @@ def _load_simulation_state(self, simulation_id: str) -> Optional[SimulationState
             entities_count=data.get("entities_count", 0),
             profiles_count=data.get("profiles_count", 0),
             entity_types=data.get("entity_types", []),
+            profiles_generated=data.get("profiles_generated", False),
             config_generated=data.get("config_generated", False),
             config_reasoning=data.get("config_reasoning", ""),
             current_round=data.get("current_round", 0),
@@ -264,6 +268,10 @@ def prepare_simulation(
 
         try:
             state.status = SimulationStatus.PREPARING
+            state.error = None
+            state.profiles_generated = False
+            state.config_generated = False
+            state.config_reasoning = ""
             self._save_simulation_state(state)
 
             sim_dir = self._get_simulation_dir(simulation_id)
@@ -298,7 +306,7 @@ def prepare_simulation(
                 state.status = SimulationStatus.FAILED
                 state.error = "没有找到符合条件的实体，请检查图谱是否正确构建"
                 self._save_simulation_state(state)
-                return state
+                raise ValueError(state.error)
 
             # ========== 阶段2: 生成Agent Profile ==========
             total_entities = len(filtered.entities)
@@ -346,7 +354,9 @@ def profile_progress(current, total, msg):
             )
 
             state.profiles_count = len(profiles)
-            
+            state.profiles_generated = len(profiles) > 0
+            self._save_simulation_state(state)
+
             # 保存Profile文件（注意：Twitter使用CSV格式，Reddit使用JSON格式）
             # Reddit 已经在生成过程中实时保存了，这里再保存一次确保完整性
             if progress_callback:

backend/app/services/simulation_runner.py

@@ -608,7 +608,17 @@ def _read_action_log(
         try:
             with open(log_path, 'r', encoding='utf-8') as f:
                 f.seek(position)
-                for line in f:
+                # Only advance safe_position after reading a complete
+                # newline-terminated line. Partial lines (writer mid-flush)
+                # are left unread for the next poll cycle to prevent data loss.
+                safe_position = position
+                while True:
+                    line = f.readline()
+                    if not line:
+                        break  # EOF
+                    if not line.endswith('\n'):
+                        break  # incomplete line — retry next poll
+                    safe_position = f.tell()
                     line = line.strip()
                     if line:
                         try:
@@ -684,7 +694,7 @@ def _read_action_log(
 
                         except json.JSONDecodeError:
                             pass
-                return f.tell()
+                return safe_position
         except Exception as e:
             logger.warning(f"读取动作日志失败: {log_path}, error={e}")
             return position

backend/pyproject.toml

@@ -2,7 +2,7 @@
 name = "mirofish-backend"
 version = "0.1.0"
 description = "MiroFish - 简洁通用的群体智能引擎，预测万物"
-requires-python = ">=3.11"
+requires-python = ">=3.11,<3.13"
 license = { text = "AGPL-3.0" }
 authors = [
     { name = "MiroFish Team" }

frontend/src/components/Step2EnvSetup.vue

@@ -952,7 +952,11 @@ const fetchProfilesRealtime = async () => {
 }
 
 // 配置轮询
+let configErrorCount = 0
+const MAX_CONFIG_ERRORS = 10
+
 const startConfigPolling = () => {
+  configErrorCount = 0
   configTimer = setInterval(fetchConfigRealtime, 2000)
 }
 
@@ -971,7 +975,17 @@ const fetchConfigRealtime = async () => {
     
     if (res.success && res.data) {
       const data = res.data
-      
+      configErrorCount = 0
+
+      // Stop polling if backend reported a failure
+      if (data.failed || data.generation_stage === 'failed') {
+        const errorMsg = data.error || 'Unknown error'
+        addLog(`❌ Config generation failed: ${errorMsg}`)
+        stopConfigPolling()
+        emit('update-status', 'failed')
+        return
+      }
+
       // 输出配置生成阶段日志（避免重复）
       if (data.generation_stage && data.generation_stage !== lastLoggedConfigStage) {
         lastLoggedConfigStage = data.generation_stage
@@ -1016,6 +1030,12 @@ const fetchConfigRealtime = async () => {
     }
   } catch (err) {
     console.warn('获取 Config 失败:', err)
+    configErrorCount++
+    if (configErrorCount >= MAX_CONFIG_ERRORS) {
+      addLog(`❌ Config generation failed: Too many consecutive polling errors`)
+      stopConfigPolling()
+      emit('update-status', 'failed')
+    }
   }
 }
 

frontend/src/components/Step3Simulation.vue

@@ -465,8 +465,11 @@ const handleStopSimulation = async () => {
 // 轮询状态
 let statusTimer = null
 let detailTimer = null
+let statusErrorCount = 0
+const MAX_STATUS_ERRORS = 10
 
 const startStatusPolling = () => {
+  statusErrorCount = 0
   statusTimer = setInterval(fetchRunStatus, 2000)
 }
 
@@ -497,9 +500,18 @@ const fetchRunStatus = async () => {
     
     if (res.success && res.data) {
       const data = res.data
-      
+      statusErrorCount = 0
       runStatus.value = data
-      
+
+      // Stop polling if backend reported a failure
+      if (data.runner_status === 'failed') {
+        const errorMsg = data.error || 'Unknown error'
+        addLog(`❌ Simulation failed: ${errorMsg}`)
+        stopPolling()
+        emit('update-status', 'failed')
+        return
+      }
+
       // 分别检测各平台的轮次变化并输出日志
       if (data.twitter_current_round > prevTwitterRound.value) {
         addLog(`[Plaza] R${data.twitter_current_round}/${data.total_rounds} | T:${data.twitter_simulated_hours || 0}h | A:${data.twitter_actions_count}`)
@@ -530,6 +542,12 @@ const fetchRunStatus = async () => {
     }
   } catch (err) {
     console.warn('获取运行状态失败:', err)
+    statusErrorCount++
+    if (statusErrorCount >= MAX_STATUS_ERRORS) {
+      addLog(`❌ Simulation failed: Too many consecutive polling errors`)
+      stopPolling()
+      emit('update-status', 'failed')
+    }
   }
 }