fix(lifecycle): Next-2.0.1 hotfix release#58
Merged
NickCharlie merged 7 commits intomainfrom Feb 24, 2026
Merged
Conversation
MySQL strict mode prohibits default values on TEXT/BLOB columns. Remove server_default from persona_content in PersonaBackup ORM.
Enable column-level auto-migration by default so missing columns (e.g. user_id on composite_psychological_states) are added via ALTER TABLE on existing databases. Skip DEFAULT clause for TEXT/BLOB/JSON columns in ALTER statements to avoid MySQL strict mode errors.
Lazy relative imports (from ....models) inside method bodies fail with ModuleNotFoundError after plugin hot-reload due to corrupted module hierarchy. Move all 65 lazy imports to module top level across 10 facade files.
Wrap every shutdown step with asyncio.wait_for to prevent the plugin from hanging on uninstall. Each service stop gets a 5s timeout, each lifecycle step gets 8s, and background task cancellation gets 3s per task. Server.stop() now runs thread.join in an executor to avoid blocking the event loop.
Root cause of 100% CPU during uninstall/reload: five background tasks with `while True` loops were created via asyncio.create_task() but never tracked or cancelled during shutdown. They kept running after the plugin was unloaded, spinning the event loop. - affection_manager: track _mood_task, cancel in _do_stop - enhanced_interaction: track _sync_task/_cleanup_task, cancel in _do_stop - intelligence_enhancement: track _knowledge_task/_recommend_task, cancel - plugin_lifecycle: register all create_task calls in background_tasks - plugin_lifecycle: clear SingletonABCMeta._instances on shutdown
The persona_backups table has persona_content as NOT NULL in existing MySQL databases. The INSERT was passing None because the field was never set, causing IntegrityError 1048.
Remove VIDEO_SCRIPT.md, update CHANGELOG with all fixes since 2.0.0
Reviewer's GuideHotfix release Next-2.0.1 focuses on making plugin/service shutdown robust (tracked background tasks, per-step timeouts, non-blocking WebUI stop), improving MySQL compatibility (TEXT/BLOB default handling, persona backup content), and fixing persona/jargon approval paths and SQLAlchemy facade imports for better hot-reload behavior. Sequence diagram for plugin shutdown and background task cancellationsequenceDiagram
actor Admin
participant PluginLifecycle as PluginLifecycle
participant PluginInstance as PluginInstance_p
participant GroupOrchestrator as GroupLearningOrchestrator
participant LearningScheduler as LearningScheduler
participant BackgroundTasks as background_tasks_set
participant FactoryManager as FactoryManager
participant V2Integration as V2LearningIntegration
participant TempPersonaUpdater as TemporaryPersonaUpdater
participant MessageCollector as MessageCollector
participant WebUIManager as WebUIManager
Admin->>PluginLifecycle: shutdown()
activate PluginLifecycle
PluginLifecycle->>PluginInstance: get _group_orchestrator
alt has_group_orchestrator
PluginLifecycle->>GroupOrchestrator: cancel_all() via _safe_step
activate GroupOrchestrator
GroupOrchestrator->>GroupOrchestrator: stop_learning() with timeout 3s
GroupOrchestrator->>GroupOrchestrator: cancel learning_tasks with per-task timeout 2s
deactivate GroupOrchestrator
end
alt has_learning_scheduler
PluginLifecycle->>LearningScheduler: stop() via _safe_step
end
PluginLifecycle->>BackgroundTasks: iterate over tasks
loop for_each_task
PluginLifecycle->>BackgroundTasks: task.cancel() if not done
PluginLifecycle->>BackgroundTasks: wait_for(shield(task), timeout=3s)
end
PluginLifecycle->>BackgroundTasks: clear()
alt has_factory_manager
PluginLifecycle->>FactoryManager: cleanup() via _safe_step
end
alt has_v2_integration
PluginLifecycle->>V2Integration: stop() via _safe_step
end
PluginLifecycle->>PluginLifecycle: clear SingletonABCMeta._instances
alt has_temporary_persona_updater
PluginLifecycle->>TempPersonaUpdater: cleanup_temp_personas() via _safe_step
end
alt has_message_collector
PluginLifecycle->>MessageCollector: save_state() via _safe_step
end
alt has_webui_manager
PluginLifecycle->>WebUIManager: stop() via _safe_step
end
PluginLifecycle-->>Admin: shutdown_complete
deactivate PluginLifecycle
Sequence diagram for WebUI stop with lock timeout and non-blocking server thread joinsequenceDiagram
participant PluginLifecycle as PluginLifecycle
participant WebUIManager as WebUIManager
participant CleanupLock as _server_cleanup_lock
participant WebUIServer as WebUIServer
participant ServerThread as server_thread
participant EventLoop as asyncio_event_loop
PluginLifecycle->>WebUIManager: stop()
activate WebUIManager
WebUIManager->>CleanupLock: acquire(timeout=3s)
alt lock_acquired
WebUIManager->>WebUIServer: stop()
alt server_has_thread
WebUIServer->>EventLoop: run_in_executor(join_thread, 5s)
activate EventLoop
EventLoop->>ServerThread: join(timeout=5s)
alt join_completes_in_6s
ServerThread-->>EventLoop: joined
EventLoop-->>WebUIServer: ok
else join_timeout
EventLoop-->>WebUIServer: asyncio.TimeoutError
WebUIServer->>WebUIServer: log timeout warning
end
deactivate EventLoop
WebUIServer->>WebUIServer: set server_thread=None
else no_thread
WebUIServer->>WebUIServer: skip_thread_join
end
WebUIManager->>CleanupLock: release()
WebUIManager-->>PluginLifecycle: stop_completed
else lock_timeout
WebUIManager->>WebUIManager: log lock timeout warning
alt server_instance_exists
WebUIManager->>WebUIServer: stop() best_effort
end
WebUIManager->>WebUIManager: set _server_instance=None
WebUIManager-->>PluginLifecycle: stop_completed_without_lock
end
deactivate WebUIManager
Updated class diagram for background services, service registry, and PersonaBackupclassDiagram
class ServiceRegistry{
+Dict~str,Any~ _services
+int _SERVICE_STOP_TIMEOUT
+bool start_all_services()
+bool stop_all_services()
+Dict~str,str~ get_service_status()
}
class PluginLifecycle{
-Any _plugin
-Any _webui_manager
-float _STEP_TIMEOUT
-float _TASK_CANCEL_TIMEOUT
+async shutdown() None
+async _safe_step(label str, coro Any, timeout float) None
}
class IntelligenceEnhancementService{
-Any _logger
-Task _knowledge_task
-Task _recommend_task
+async _do_start() bool
+async _do_stop() bool
+async _periodic_knowledge_update() None
+async _periodic_recommendation_refresh() None
}
class EnhancedInteractionService{
-Any _logger
-Task _sync_task
-Task _cleanup_task
-float memory_sync_interval
-float context_retention_time
+async _do_start() bool
+async _do_stop() bool
+async _periodic_memory_sync() None
+async _periodic_context_cleanup() None
}
class AffectionManager{
-Any _logger
-Any config
-Task _mood_task
+async _do_start() bool
+async _do_stop() bool
+async _daily_mood_updater() None
}
class DatabaseEngine{
+Any engine
+async create_tables(enable_auto_migration bool) None
-_get_existing_columns(sync_conn Any) Dict~str,Any~
}
class SQLAlchemyDatabaseManager{
+Any engine
+async start() bool
-_init_facades() None
}
class PersonaBackup{
+Integer id
+String persona_id
+Text persona_name
+Text persona_description
+Text persona_traits
+Text persona_memory
+Text persona_goals
+Text persona_constraints
+Text imitation_dialogues
+Text backup_reason
+Float backup_time
+Text persona_content
+DateTime created_at
}
class PersonaFacade{
+async backup_persona(backup_data Dict~str,Any~) bool
+async get_persona_update_history(limit int) List~Dict~str,Any~~
}
class WebUIServer{
-Thread server_thread
-Any _thread_loop
+async stop() None
}
class WebUIManager{
+async stop() None
}
ServiceRegistry --> PluginLifecycle : manages
PluginLifecycle --> WebUIManager : uses
WebUIManager --> WebUIServer : controls
SQLAlchemyDatabaseManager --> DatabaseEngine : owns
PersonaFacade --> PersonaBackup : creates
IntelligenceEnhancementService ..> ServiceRegistry : registered_as_service
EnhancedInteractionService ..> ServiceRegistry : registered_as_service
AffectionManager ..> ServiceRegistry : registered_as_service
IntelligenceEnhancementService --> PluginLifecycle : background_tasks_cancelled_on_shutdown
EnhancedInteractionService --> PluginLifecycle : background_tasks_cancelled_on_shutdown
AffectionManager --> PluginLifecycle : background_tasks_cancelled_on_shutdown
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hey - I've found 1 issue, and left some high level feedback:
- The periodic background loops that previously caught and logged generic exceptions now only handle
asyncio.CancelledError, so any unexpected error will terminate the task permanently; consider keeping an internal try/except (with logging) inside the loop to avoid silent degradation over time. - The various shutdown timeouts (
_STEP_TIMEOUT,_TASK_CANCEL_TIMEOUT,_SERVICE_STOP_TIMEOUT, WebUI lock/stop timeouts, etc.) are currently hardcoded; consider centralizing or making them configurable so they can be tuned for different environments without code changes.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- The periodic background loops that previously caught and logged generic exceptions now only handle `asyncio.CancelledError`, so any unexpected error will terminate the task permanently; consider keeping an internal try/except (with logging) inside the loop to avoid silent degradation over time.
- The various shutdown timeouts (`_STEP_TIMEOUT`, `_TASK_CANCEL_TIMEOUT`, `_SERVICE_STOP_TIMEOUT`, WebUI lock/stop timeouts, etc.) are currently hardcoded; consider centralizing or making them configurable so they can be tuned for different environments without code changes.
## Individual Comments
### Comment 1
<location path="core/database/engine.py" line_range="290-292" />
<code_context>
- default = f" DEFAULT {col.server_default.arg!r}"
- elif col.default is not None and col.default.is_scalar:
- default = f" DEFAULT {col.default.arg!r}"
+ # MySQL 不允许 TEXT/BLOB 列有 DEFAULT 值
+ is_text_type = col_type.upper() in ("TEXT", "BLOB", "MEDIUMTEXT", "LONGTEXT", "JSON")
+ if not is_text_type:
+ if col.server_default is not None:
+ default = f" DEFAULT {col.server_default.arg!r}"
</code_context>
<issue_to_address>
**suggestion (bug_risk):** Type comparison for TEXT/BLOB defaults may miss dialect-specific variants, causing unexpected defaults to be emitted.
`is_text_type` depends on `col_type.upper()` matching one of a few exact strings. Dialects/configs may produce variants like `TINYTEXT`, `LONGBLOB`, or `TEXT CHARACTER SET utf8mb4`, which would bypass this check and allow invalid defaults again. To harden this, consider prefix/suffix checks (e.g. `startswith("TEXT")` / `endswith("BLOB")`) or, preferably, use SQLAlchemy’s type classes instead of string comparisons.
Suggested implementation:
```python
col_type = col.type.compile(self.engine.dialect)
nullable = "NULL" if col.nullable else "NOT NULL"
default = ""
# MySQL 不允许 TEXT/BLOB/JSON 列有 DEFAULT 值
# 使用 SQLAlchemy 类型而不是字符串比较，以兼容不同方言生成的类型名称
from sqlalchemy import Text, LargeBinary
from sqlalchemy.dialects.mysql import JSON as MySQLJSON
is_text_blob_or_json_type = isinstance(
col.type,
(Text, LargeBinary, MySQLJSON),
)
if not is_text_blob_or_json_type:
if col.server_default is not None:
default = f" DEFAULT {col.server_default.arg!r}"
elif col.default is not None and col.default.is_scalar:
default = f" DEFAULT {col.default.arg!r}"
alter_statements.append(
```
To follow best practices and avoid importing inside the function/method body, you should:
1. Move the `from sqlalchemy import Text, LargeBinary` and `from sqlalchemy.dialects.mysql import JSON as MySQLJSON` imports to the top of `core/database/engine.py`, alongside the other imports.
2. After moving the imports, remove the two import lines from this block, leaving only the `is_text_blob_or_json_type = isinstance(...)` check.
If `core/database/engine.py` is intended to be dialect-agnostic and used with non-MySQL engines, you may additionally want to guard this logic with a dialect check (e.g. only apply the `is_text_blob_or_json_type` restriction when `self.engine.dialect.name == "mysql"` or `"mariadb"`).
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
Comment on lines
+290
to
+292
| # MySQL 不允许 TEXT/BLOB 列有 DEFAULT 值 | ||
| is_text_type = col_type.upper() in ("TEXT", "BLOB", "MEDIUMTEXT", "LONGTEXT", "JSON") | ||
| if not is_text_type: |
There was a problem hiding this comment.
suggestion (bug_risk): Type comparison for TEXT/BLOB defaults may miss dialect-specific variants, causing unexpected defaults to be emitted.
is_text_type depends on col_type.upper() matching one of a few exact strings. Dialects/configs may produce variants like TINYTEXT, LONGBLOB, or TEXT CHARACTER SET utf8mb4, which would bypass this check and allow invalid defaults again. To harden this, consider prefix/suffix checks (e.g. startswith("TEXT") / endswith("BLOB")) or, preferably, use SQLAlchemy’s type classes instead of string comparisons.
Suggested implementation:
col_type = col.type.compile(self.engine.dialect)
nullable = "NULL" if col.nullable else "NOT NULL"
default = ""
# MySQL 不允许 TEXT/BLOB/JSON 列有 DEFAULT 值
# 使用 SQLAlchemy 类型而不是字符串比较，以兼容不同方言生成的类型名称
from sqlalchemy import Text, LargeBinary
from sqlalchemy.dialects.mysql import JSON as MySQLJSON
is_text_blob_or_json_type = isinstance(
col.type,
(Text, LargeBinary, MySQLJSON),
)
if not is_text_blob_or_json_type:
if col.server_default is not None:
default = f" DEFAULT {col.server_default.arg!r}"
elif col.default is not None and col.default.is_scalar:
default = f" DEFAULT {col.default.arg!r}"
alter_statements.append(To follow best practices and avoid importing inside the function/method body, you should:
- Move the
from sqlalchemy import Text, LargeBinaryandfrom sqlalchemy.dialects.mysql import JSON as MySQLJSONimports to the top ofcore/database/engine.py, alongside the other imports. - After moving the imports, remove the two import lines from this block, leaving only the
is_text_blob_or_json_type = isinstance(...)check.
If core/database/engine.py is intended to be dialect-agnostic and used with non-MySQL engines, you may additionally want to guard this logic with a dialect check (e.g. only apply the is_text_blob_or_json_type restriction when self.engine.dialect.name == "mysql" or "mariadb").
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概述
Next-2.0.1 热修复版本，修复插件卸载/重载卡死、MySQL 兼容性及人格审批流程问题。
Summary
Hotfix release for Next-2.0.1, resolving plugin uninstall/reload hang, MySQL compatibility, and persona approval issues.
修复内容
插件生命周期
Server.stop()使用线程池执行器避免阻塞事件循环SingletonABCMeta._instances防止重载后单例残留MySQL 兼容性
persona_content列 NOT NULL 约束导致 INSERT 失败人格审批
PersonaWebManager路由，修复跨线程调用卡死save_or_update_jargon参数顺序和类型错误测试计划
Summary by Sourcery
Stabilize plugin lifecycle shutdown and improve MySQL compatibility and persona/jargon data handling for the Next-2.0.1 hotfix release.
Bug Fixes:
Enhancements:
Documentation:
Chores: