fix: fix the first time calling knowledge tool error#194
Merged
Conversation
Contributor
Reviewer's Guide为任务/worker 上下文引入了专用的异步数据库会话上下文，并在知识工具操作中使用它；同时重构了签到积分计算及其类型标注，并在一个文献相关测试中做了小的类型改进。 Sequence diagram for knowledge tool operations using get_task_db_sessionsequenceDiagram
actor Worker
participant KnowledgeToolOperations
participant DatabaseConnection
participant AsyncSession
participant FileRepository
Worker->>KnowledgeToolOperations: list_files(user_id, knowledge_set_id)
activate KnowledgeToolOperations
KnowledgeToolOperations->>DatabaseConnection: get_task_db_session()
activate DatabaseConnection
DatabaseConnection->>DatabaseConnection: pid = os.getpid()
alt engine_not_cached_for_pid
DatabaseConnection->>DatabaseConnection: create_async_engine(ASYNC_DATABASE_URL)
DatabaseConnection->>DatabaseConnection: async_sessionmaker(bind=task_engine)
DatabaseConnection->>DatabaseConnection: _worker_engines[pid] = sessionmaker
end
DatabaseConnection-->>KnowledgeToolOperations: AsyncSession context manager
deactivate DatabaseConnection
KnowledgeToolOperations->>AsyncSession: open session (async with)
activate AsyncSession
KnowledgeToolOperations->>FileRepository: FileRepository(db_session)
activate FileRepository
FileRepository-->>KnowledgeToolOperations: files
deactivate FileRepository
AsyncSession-->>KnowledgeToolOperations: close session (context exit)
deactivate AsyncSession
KnowledgeToolOperations-->>Worker: result dict
deactivate KnowledgeToolOperations
Class diagram for CheckInService and CheckInStatusclassDiagram
class CheckInStatus {
bool checked_in_today
int consecutive_days
int next_points
int total_check_ins
}
class CheckInService {
+calculate_points(consecutive_days int) int
+check_in(user_id str) tuple
+get_check_in_status(user_id str) CheckInStatus
}
File-Level Changes
Tips and commandsInteracting with Sourcery
自定义你的使用体验访问你的 dashboard 来：
获取帮助Original review guide in EnglishReviewer's GuideIntroduces a dedicated async DB session context for task/worker contexts and uses it in knowledge tool operations, refactors check-in points calculation and typing, and adds a small typing improvement in a literature test. Sequence diagram for knowledge tool operations using get_task_db_sessionsequenceDiagram
actor Worker
participant KnowledgeToolOperations
participant DatabaseConnection
participant AsyncSession
participant FileRepository
Worker->>KnowledgeToolOperations: list_files(user_id, knowledge_set_id)
activate KnowledgeToolOperations
KnowledgeToolOperations->>DatabaseConnection: get_task_db_session()
activate DatabaseConnection
DatabaseConnection->>DatabaseConnection: pid = os.getpid()
alt engine_not_cached_for_pid
DatabaseConnection->>DatabaseConnection: create_async_engine(ASYNC_DATABASE_URL)
DatabaseConnection->>DatabaseConnection: async_sessionmaker(bind=task_engine)
DatabaseConnection->>DatabaseConnection: _worker_engines[pid] = sessionmaker
end
DatabaseConnection-->>KnowledgeToolOperations: AsyncSession context manager
deactivate DatabaseConnection
KnowledgeToolOperations->>AsyncSession: open session (async with)
activate AsyncSession
KnowledgeToolOperations->>FileRepository: FileRepository(db_session)
activate FileRepository
FileRepository-->>KnowledgeToolOperations: files
deactivate FileRepository
AsyncSession-->>KnowledgeToolOperations: close session (context exit)
deactivate AsyncSession
KnowledgeToolOperations-->>Worker: result dict
deactivate KnowledgeToolOperations
Class diagram for CheckInService and CheckInStatusclassDiagram
class CheckInStatus {
bool checked_in_today
int consecutive_days
int next_points
int total_check_ins
}
class CheckInService {
+calculate_points(consecutive_days int) int
+check_in(user_id str) tuple
+get_check_in_status(user_id str) CheckInStatus
}
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}
Contributor
There was a problem hiding this comment.
Hey - 我发现了 3 个问题，并给出了一些高层次的反馈：
- 在
get_task_db_session中，按 PID 作为 key 的全局_worker_engines字典不是并发安全的，所以在同一进程中并发的首次调用可能会创建多个 engine，并且有些不会被释放；建议在初始化时加保护（例如使用锁），或者使用setdefault并统一到单一的 engine 创建路径。 - 在
get_task_db_session中，存储在_worker_engines里的 engine 从未被释放，这会导致在长时间运行的 worker 生命周期中出现资源泄漏；你可能需要一个清理钩子（例如在 worker 关闭时）或一个显式的释放路径。 get_task_db_session中那大块被注释掉的代码让函数更难阅读；建议删除它，或者将这些备用实现的细节移到 docstring 或设计文档里。
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- The global `_worker_engines` dict keyed by PID in `get_task_db_session` is not concurrency-safe, so concurrent first-time calls in the same process could create multiple engines and leave some undisposed; consider protecting initialization (e.g., with a lock) or using `setdefault` plus a single engine creation path.
- In `get_task_db_session`, engines stored in `_worker_engines` are never disposed, which can lead to resource leakage over the lifetime of long-running workers; you may want a cleanup hook (e.g., on worker shutdown) or an explicit disposal path.
- The large block of commented-out code in `get_task_db_session` makes the function harder to read; consider removing it or moving the alternative implementation details into a docstring or design note.
## Individual Comments
### Comment 1
<location> `service/app/infra/database/connection.py:150-158` </location>
<code_context>
+ """
+ Get a database session suitable for Celery Worker / tool execution contexts.
+
+ Each call creates a new engine bound to the current event loop to avoid cross-process issues.
+
+ Usage:
</code_context>
<issue_to_address>
**suggestion:** The docstring no longer matches the implementation now that engines are reused per process.
Since the implementation now caches engines per worker process via `_worker_engines`, the docstring should describe that behavior (e.g., engines are cached per PID and sessions are created from those engines) instead of saying each call creates a new engine.
```suggestion
"""
Get a database session suitable for Celery Worker / tool execution contexts.
Engines are cached per worker process (PID) via `_worker_engines` to avoid cross-process
issues, and each call creates a new `AsyncSession` from the process-local sessionmaker.
Usage:
async with get_task_db_session() as db:
result = await db.execute(...)
"""
```
</issue_to_address>
### Comment 2
<location> `service/app/infra/database/connection.py:175-177` </location>
<code_context>
+ # Better version that reuses engines per process
+ pid = os.getpid()
+
+ if pid not in _worker_engines:
+ task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
+ _worker_engines[pid] = async_sessionmaker(
+ bind=task_engine,
+ class_=AsyncSession,
</code_context>
<issue_to_address>
**suggestion:** Engine lifecycle for worker sessions is implicit; consider making disposal/shutdown behavior explicit.
Because `_worker_engines` keeps one engine per PID for the process lifetime, this is fine for long‑lived workers but couples behavior to that assumption. If you have (or plan) a worker shutdown hook, consider wiring an explicit `dispose()` there, or wrapping engine creation/cleanup in a helper so it’s easy to adapt if workers become short‑lived or need manual teardown.
</issue_to_address>
### Comment 3
<location> `service/app/infra/database/connection.py:145` </location>
<code_context>
yield session
+
+
+_worker_engines: dict[int, async_sessionmaker[AsyncSession]] = {}
+
+
</code_context>
<issue_to_address>
**issue (complexity):** Consider replacing the PID→sessionmaker dict and commented-out code with a lazily-initialized per-process sessionmaker helper to simplify `get_task_db_session` while preserving per-process safety.
You can keep the per-process safety without the PID→sessionmaker dict or large commented block by collapsing this into a single lazily-initialized sessionmaker that is refreshed when the PID changes.
### 1. Replace `_worker_engines` with per-process singleton
```python
_task_sessionmaker: async_sessionmaker[AsyncSession] | None = None
_task_session_pid: int | None = None
def _get_task_sessionmaker() -> async_sessionmaker[AsyncSession]:
global _task_sessionmaker, _task_session_pid
pid = os.getpid()
if _task_sessionmaker is None or _task_session_pid != pid:
# New process (or first call): create a fresh engine + sessionmaker
task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
_task_sessionmaker = async_sessionmaker(
bind=task_engine,
class_=AsyncSession,
expire_on_commit=False,
)
_task_session_pid = pid
return _task_sessionmaker
```
Then simplify the contextmanager:
```python
@asynccontextmanager
async def get_task_db_session() -> AsyncGenerator[AsyncSession, None]:
SessionLocal = _get_task_sessionmaker()
async with SessionLocal() as session:
yield session
```
This preserves:
- Per-process isolation (PID change ⇒ new engine/sessionmaker)
- Engine reuse within a worker process
while avoiding the dict keyed by PID and extra indirection.
### 2. Remove the large commented-out alternative
Instead of the full commented version, keep at most a short rationale:
```python
# Note: we create the engine lazily per-process (based on PID) to avoid
# sharing engines across forked workers.
```
All prior versions are still visible in git history, so the full commented implementation isn’t needed here.
</issue_to_address>帮我变得更有用！请对每条评论点👍或👎，我会根据这些反馈改进后续的代码审查。
Original comment in English
Hey - I've found 3 issues, and left some high level feedback:
- The global
_worker_enginesdict keyed by PID inget_task_db_sessionis not concurrency-safe, so concurrent first-time calls in the same process could create multiple engines and leave some undisposed; consider protecting initialization (e.g., with a lock) or usingsetdefaultplus a single engine creation path. - In
get_task_db_session, engines stored in_worker_enginesare never disposed, which can lead to resource leakage over the lifetime of long-running workers; you may want a cleanup hook (e.g., on worker shutdown) or an explicit disposal path. - The large block of commented-out code in
get_task_db_sessionmakes the function harder to read; consider removing it or moving the alternative implementation details into a docstring or design note.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- The global `_worker_engines` dict keyed by PID in `get_task_db_session` is not concurrency-safe, so concurrent first-time calls in the same process could create multiple engines and leave some undisposed; consider protecting initialization (e.g., with a lock) or using `setdefault` plus a single engine creation path.
- In `get_task_db_session`, engines stored in `_worker_engines` are never disposed, which can lead to resource leakage over the lifetime of long-running workers; you may want a cleanup hook (e.g., on worker shutdown) or an explicit disposal path.
- The large block of commented-out code in `get_task_db_session` makes the function harder to read; consider removing it or moving the alternative implementation details into a docstring or design note.
## Individual Comments
### Comment 1
<location> `service/app/infra/database/connection.py:150-158` </location>
<code_context>
+ """
+ Get a database session suitable for Celery Worker / tool execution contexts.
+
+ Each call creates a new engine bound to the current event loop to avoid cross-process issues.
+
+ Usage:
</code_context>
<issue_to_address>
**suggestion:** The docstring no longer matches the implementation now that engines are reused per process.
Since the implementation now caches engines per worker process via `_worker_engines`, the docstring should describe that behavior (e.g., engines are cached per PID and sessions are created from those engines) instead of saying each call creates a new engine.
```suggestion
"""
Get a database session suitable for Celery Worker / tool execution contexts.
Engines are cached per worker process (PID) via `_worker_engines` to avoid cross-process
issues, and each call creates a new `AsyncSession` from the process-local sessionmaker.
Usage:
async with get_task_db_session() as db:
result = await db.execute(...)
"""
```
</issue_to_address>
### Comment 2
<location> `service/app/infra/database/connection.py:175-177` </location>
<code_context>
+ # Better version that reuses engines per process
+ pid = os.getpid()
+
+ if pid not in _worker_engines:
+ task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
+ _worker_engines[pid] = async_sessionmaker(
+ bind=task_engine,
+ class_=AsyncSession,
</code_context>
<issue_to_address>
**suggestion:** Engine lifecycle for worker sessions is implicit; consider making disposal/shutdown behavior explicit.
Because `_worker_engines` keeps one engine per PID for the process lifetime, this is fine for long‑lived workers but couples behavior to that assumption. If you have (or plan) a worker shutdown hook, consider wiring an explicit `dispose()` there, or wrapping engine creation/cleanup in a helper so it’s easy to adapt if workers become short‑lived or need manual teardown.
</issue_to_address>
### Comment 3
<location> `service/app/infra/database/connection.py:145` </location>
<code_context>
yield session
+
+
+_worker_engines: dict[int, async_sessionmaker[AsyncSession]] = {}
+
+
</code_context>
<issue_to_address>
**issue (complexity):** Consider replacing the PID→sessionmaker dict and commented-out code with a lazily-initialized per-process sessionmaker helper to simplify `get_task_db_session` while preserving per-process safety.
You can keep the per-process safety without the PID→sessionmaker dict or large commented block by collapsing this into a single lazily-initialized sessionmaker that is refreshed when the PID changes.
### 1. Replace `_worker_engines` with per-process singleton
```python
_task_sessionmaker: async_sessionmaker[AsyncSession] | None = None
_task_session_pid: int | None = None
def _get_task_sessionmaker() -> async_sessionmaker[AsyncSession]:
global _task_sessionmaker, _task_session_pid
pid = os.getpid()
if _task_sessionmaker is None or _task_session_pid != pid:
# New process (or first call): create a fresh engine + sessionmaker
task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
_task_sessionmaker = async_sessionmaker(
bind=task_engine,
class_=AsyncSession,
expire_on_commit=False,
)
_task_session_pid = pid
return _task_sessionmaker
```
Then simplify the contextmanager:
```python
@asynccontextmanager
async def get_task_db_session() -> AsyncGenerator[AsyncSession, None]:
SessionLocal = _get_task_sessionmaker()
async with SessionLocal() as session:
yield session
```
This preserves:
- Per-process isolation (PID change ⇒ new engine/sessionmaker)
- Engine reuse within a worker process
while avoiding the dict keyed by PID and extra indirection.
### 2. Remove the large commented-out alternative
Instead of the full commented version, keep at most a short rationale:
```python
# Note: we create the engine lazily per-process (based on PID) to avoid
# sharing engines across forked workers.
```
All prior versions are still visible in git history, so the full commented implementation isn’t needed here.
</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
+150
to
+158
| """ | ||
| Get a database session suitable for Celery Worker / tool execution contexts. | ||
|
|
||
| Each call creates a new engine bound to the current event loop to avoid cross-process issues. | ||
|
|
||
| Usage: | ||
| async with get_task_db_session() as db: | ||
| result = await db.execute(...) | ||
| """ |
Contributor
There was a problem hiding this comment.
suggestion: 现在 engine 已经在每个进程内复用，docstring 与实现不再匹配。
由于当前实现是通过 _worker_engines 在每个 worker 进程中缓存 engine，docstring 应该描述这种行为（例如：engine 以 PID 为维度进行缓存，并从这些 engine 创建 session），而不是说每次调用都会创建一个新的 engine。
Suggested change
| """ | |
| Get a database session suitable for Celery Worker / tool execution contexts. | |
| Each call creates a new engine bound to the current event loop to avoid cross-process issues. | |
| Usage: | |
| async with get_task_db_session() as db: | |
| result = await db.execute(...) | |
| """ | |
| """ | |
| Get a database session suitable for Celery Worker / tool execution contexts. | |
| Engines are cached per worker process (PID) via `_worker_engines` to avoid cross-process | |
| issues, and each call creates a new `AsyncSession` from the process-local sessionmaker. | |
| Usage: | |
| async with get_task_db_session() as db: | |
| result = await db.execute(...) | |
| """ |
Original comment in English
suggestion: The docstring no longer matches the implementation now that engines are reused per process.
Since the implementation now caches engines per worker process via _worker_engines, the docstring should describe that behavior (e.g., engines are cached per PID and sessions are created from those engines) instead of saying each call creates a new engine.
Suggested change
| """ | |
| Get a database session suitable for Celery Worker / tool execution contexts. | |
| Each call creates a new engine bound to the current event loop to avoid cross-process issues. | |
| Usage: | |
| async with get_task_db_session() as db: | |
| result = await db.execute(...) | |
| """ | |
| """ | |
| Get a database session suitable for Celery Worker / tool execution contexts. | |
| Engines are cached per worker process (PID) via `_worker_engines` to avoid cross-process | |
| issues, and each call creates a new `AsyncSession` from the process-local sessionmaker. | |
| Usage: | |
| async with get_task_db_session() as db: | |
| result = await db.execute(...) | |
| """ |
Comment on lines
+175
to
+177
| if pid not in _worker_engines: | ||
| task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True) | ||
| _worker_engines[pid] = async_sessionmaker( |
Contributor
There was a problem hiding this comment.
suggestion: 当前 worker session 的 engine 生命周期是隐式的；建议将释放/关闭行为显式化。
由于 _worker_engines 在一个进程的整个生命周期内为每个 PID 保留一个 engine，这对长生命周期的 worker 来说没问题，但实现依赖于这种假设。如果你已经有（或计划有）worker 的关闭钩子，可以考虑在那儿显式调用 dispose()，或者把 engine 的创建/清理封装到一个辅助函数里，这样当 worker 变成短生命周期或需要手动拆卸时更容易适配。
Original comment in English
suggestion: Engine lifecycle for worker sessions is implicit; consider making disposal/shutdown behavior explicit.
Because _worker_engines keeps one engine per PID for the process lifetime, this is fine for long‑lived workers but couples behavior to that assumption. If you have (or plan) a worker shutdown hook, consider wiring an explicit dispose() there, or wrapping engine creation/cleanup in a helper so it’s easy to adapt if workers become short‑lived or need manual teardown.
| yield session | ||
|
|
||
|
|
||
| _worker_engines: dict[int, async_sessionmaker[AsyncSession]] = {} |
Contributor
There was a problem hiding this comment.
issue (complexity): 建议用一个延迟初始化的、按进程的 sessionmaker helper 来替代当前的 PID→sessionmaker 字典和被注释掉的代码，在保持进程级安全的同时简化 get_task_db_session。
你可以在不使用 PID→sessionmaker 字典和大块注释代码的情况下保持进程级安全，只需要把逻辑折叠为一个在 PID 变化时刷新的延迟初始化 sessionmaker。
1. 用按进程的单例替换 _worker_engines
_task_sessionmaker: async_sessionmaker[AsyncSession] | None = None
_task_session_pid: int | None = None
def _get_task_sessionmaker() -> async_sessionmaker[AsyncSession]:
global _task_sessionmaker, _task_session_pid
pid = os.getpid()
if _task_sessionmaker is None or _task_session_pid != pid:
# New process (or first call): create a fresh engine + sessionmaker
task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
_task_sessionmaker = async_sessionmaker(
bind=task_engine,
class_=AsyncSession,
expire_on_commit=False,
)
_task_session_pid = pid
return _task_sessionmaker然后可以把 contextmanager 简化为：
@asynccontextmanager
async def get_task_db_session() -> AsyncGenerator[AsyncSession, None]:
SessionLocal = _get_task_sessionmaker()
async with SessionLocal() as session:
yield session这样既能保留：
- 进程级隔离（PID 变化 ⇒ 新的 engine/sessionmaker）
- 在同一 worker 进程内重用 engine
又能避免使用以 PID 为 key 的字典和额外的间接层。
2. 移除那块大型注释的备用实现
与其保留完整的被注释版本，不如最多留一段简短的设计理由说明：
# Note: we create the engine lazily per-process (based on PID) to avoid
# sharing engines across forked workers.之前的所有版本都可以在 git 历史里找到，因此没必要在这里保留完整的注释实现。
Original comment in English
issue (complexity): Consider replacing the PID→sessionmaker dict and commented-out code with a lazily-initialized per-process sessionmaker helper to simplify get_task_db_session while preserving per-process safety.
You can keep the per-process safety without the PID→sessionmaker dict or large commented block by collapsing this into a single lazily-initialized sessionmaker that is refreshed when the PID changes.
1. Replace _worker_engines with per-process singleton
_task_sessionmaker: async_sessionmaker[AsyncSession] | None = None
_task_session_pid: int | None = None
def _get_task_sessionmaker() -> async_sessionmaker[AsyncSession]:
global _task_sessionmaker, _task_session_pid
pid = os.getpid()
if _task_sessionmaker is None or _task_session_pid != pid:
# New process (or first call): create a fresh engine + sessionmaker
task_engine = create_async_engine(ASYNC_DATABASE_URL, echo=False, future=True)
_task_sessionmaker = async_sessionmaker(
bind=task_engine,
class_=AsyncSession,
expire_on_commit=False,
)
_task_session_pid = pid
return _task_sessionmakerThen simplify the contextmanager:
@asynccontextmanager
async def get_task_db_session() -> AsyncGenerator[AsyncSession, None]:
SessionLocal = _get_task_sessionmaker()
async with SessionLocal() as session:
yield sessionThis preserves:
- Per-process isolation (PID change ⇒ new engine/sessionmaker)
- Engine reuse within a worker process
while avoiding the dict keyed by PID and extra indirection.
2. Remove the large commented-out alternative
Instead of the full commented version, keep at most a short rationale:
# Note: we create the engine lazily per-process (based on PID) to avoid
# sharing engines across forked workers.All prior versions are still visible in git history, so the full commented implementation isn’t needed here.
Mile-Away
added a commit
that referenced
this pull request
Jan 22, 2026
* Feature/literature mcp (#192) * feat: literature-MCP 完整功能 * refactor: improve boolean parsing and logging in literature search functions * feat: enhance literature search functionality with improved query validation and detailed results formatting * refactor: rename oa_url to access_url in LiteratureWork model and related tests * feat: remove test-build workflow and update README for development setup * feat: tool cost system and PPTX image handling fixes (#193) * fix: prompt, factory * feat: enhanced ppt generation with image slides mode - Add image_slides mode for PPTX with full-bleed AI-generated images - Add ImageBlock.image_id field for referencing generated images - Add ImageSlideSpec for image-only slides - Add ImageFetcher service for fetching images from various sources - Reorganize knowledge module from single file to module structure - Move document utilities from app/mcp/ to app/tools/utils/documents/ - Resolve image_ids to storage URLs in async layer (operations.py) - Fix type errors and move tests to proper location Co-Authored-By: Claude <noreply@anthropic.com> * feat: implement the tool cost --------- Co-authored-by: Claude <noreply@anthropic.com> * fix: fix the first time calling knowledge tool error (#194) * fix: fix the wrong cache for second call of agent tools (#195) * feat: several improvements (#196) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * feat: multiple UI improvements and fixes (#198) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * fix: restore previous active topic when clicking agent Instead of always jumping to the latest topic, now tracks and restores the previously active topic for each agent when switching between them. Co-Authored-By: Claude <noreply@anthropic.com> * feat: add context menu to FocusedView agents and download button to lightbox - Add right-click context menu (edit/delete) to compact AgentListItem variant - Render context menu via portal to escape overflow:hidden containers - Add edit/delete handlers to FocusedView with AgentSettingsModal and ConfirmationModal - Add download button to image lightbox with smart filename detection Co-Authored-By: Claude <noreply@anthropic.com> * feat: add web_fetch tool bundled with web_search - Add web_fetch tool using Trafilatura for content extraction - Bundle web_fetch with web_search in frontend toolConfig - Group WEB_SEARCH_TOOLS for unified toggle behavior - Only load web_fetch when web_search is available (SearXNG enabled) - Update tool capabilities mapping for web_fetch Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> --------- Co-authored-by: Meng Junxing <junxingmeng@gmail.com> Co-authored-by: Harvey <q-query@outlook.com> Co-authored-by: Claude <noreply@anthropic.com>
Mile-Away
added a commit
that referenced
this pull request
Jan 25, 2026
* Feature/literature mcp (#192) * feat: literature-MCP 完整功能 * refactor: improve boolean parsing and logging in literature search functions * feat: enhance literature search functionality with improved query validation and detailed results formatting * refactor: rename oa_url to access_url in LiteratureWork model and related tests * feat: remove test-build workflow and update README for development setup * feat: tool cost system and PPTX image handling fixes (#193) * fix: prompt, factory * feat: enhanced ppt generation with image slides mode - Add image_slides mode for PPTX with full-bleed AI-generated images - Add ImageBlock.image_id field for referencing generated images - Add ImageSlideSpec for image-only slides - Add ImageFetcher service for fetching images from various sources - Reorganize knowledge module from single file to module structure - Move document utilities from app/mcp/ to app/tools/utils/documents/ - Resolve image_ids to storage URLs in async layer (operations.py) - Fix type errors and move tests to proper location Co-Authored-By: Claude <noreply@anthropic.com> * feat: implement the tool cost --------- Co-authored-by: Claude <noreply@anthropic.com> * fix: fix the first time calling knowledge tool error (#194) * fix: fix the wrong cache for second call of agent tools (#195) * feat: several improvements (#196) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * feat: multiple UI improvements and fixes (#198) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * fix: restore previous active topic when clicking agent Instead of always jumping to the latest topic, now tracks and restores the previously active topic for each agent when switching between them. Co-Authored-By: Claude <noreply@anthropic.com> * feat: add context menu to FocusedView agents and download button to lightbox - Add right-click context menu (edit/delete) to compact AgentListItem variant - Render context menu via portal to escape overflow:hidden containers - Add edit/delete handlers to FocusedView with AgentSettingsModal and ConfirmationModal - Add download button to image lightbox with smart filename detection Co-Authored-By: Claude <noreply@anthropic.com> * feat: add web_fetch tool bundled with web_search - Add web_fetch tool using Trafilatura for content extraction - Bundle web_fetch with web_search in frontend toolConfig - Group WEB_SEARCH_TOOLS for unified toggle behavior - Only load web_fetch when web_search is available (SearXNG enabled) - Update tool capabilities mapping for web_fetch Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> * feat: fix the fork issue and implement the locked fork --------- Co-authored-by: Meng Junxing <junxingmeng@gmail.com> Co-authored-by: Harvey <q-query@outlook.com> Co-authored-by: Claude <noreply@anthropic.com>
Mile-Away
added a commit
that referenced
this pull request
Jan 26, 2026
* Feature/better agent community (#200) * Feature/literature mcp (#192) * feat: literature-MCP 完整功能 * refactor: improve boolean parsing and logging in literature search functions * feat: enhance literature search functionality with improved query validation and detailed results formatting * refactor: rename oa_url to access_url in LiteratureWork model and related tests * feat: remove test-build workflow and update README for development setup * feat: tool cost system and PPTX image handling fixes (#193) * fix: prompt, factory * feat: enhanced ppt generation with image slides mode - Add image_slides mode for PPTX with full-bleed AI-generated images - Add ImageBlock.image_id field for referencing generated images - Add ImageSlideSpec for image-only slides - Add ImageFetcher service for fetching images from various sources - Reorganize knowledge module from single file to module structure - Move document utilities from app/mcp/ to app/tools/utils/documents/ - Resolve image_ids to storage URLs in async layer (operations.py) - Fix type errors and move tests to proper location Co-Authored-By: Claude <noreply@anthropic.com> * feat: implement the tool cost --------- Co-authored-by: Claude <noreply@anthropic.com> * fix: fix the first time calling knowledge tool error (#194) * fix: fix the wrong cache for second call of agent tools (#195) * feat: several improvements (#196) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * feat: multiple UI improvements and fixes (#198) * fix: jump to latest topic when click agent * feat: allow more than one image for generate image * feat: allow user directly edit mcp in the chat-toolbar * feat: improve the frontend perf * fix: restore previous active topic when clicking agent Instead of always jumping to the latest topic, now tracks and restores the previously active topic for each agent when switching between them. Co-Authored-By: Claude <noreply@anthropic.com> * feat: add context menu to FocusedView agents and download button to lightbox - Add right-click context menu (edit/delete) to compact AgentListItem variant - Render context menu via portal to escape overflow:hidden containers - Add edit/delete handlers to FocusedView with AgentSettingsModal and ConfirmationModal - Add download button to image lightbox with smart filename detection Co-Authored-By: Claude <noreply@anthropic.com> * feat: add web_fetch tool bundled with web_search - Add web_fetch tool using Trafilatura for content extraction - Bundle web_fetch with web_search in frontend toolConfig - Group WEB_SEARCH_TOOLS for unified toggle behavior - Only load web_fetch when web_search is available (SearXNG enabled) - Update tool capabilities mapping for web_fetch Co-Authored-By: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com> * feat: fix the fork issue and implement the locked fork --------- Co-authored-by: Meng Junxing <junxingmeng@gmail.com> Co-authored-by: Harvey <q-query@outlook.com> Co-authored-by: Claude <noreply@anthropic.com> * fix: prevent forked agents from being republished to marketplace (#201) * fix: prevent forked agents from being republished to marketplace Forked agents were able to be republished, which could expose the original agent's configuration. This fix adds validation at both API and UI levels: - Backend: Add validation in publish endpoint to reject agents with original_source_id set (HTTP 400) - Frontend: Hide publish button for forked agents in AgentSettingsModal and WorkflowEditor - Types: Add original_source_id and source_version fields to Agent interface Co-Authored-By: Claude <noreply@anthropic.com> * refactor: address code review feedback for fork detection - Extract `isForked` helper variable to avoid duplication - Use explicit nullish check (`!= null`) to match backend `is not None` semantic - Replace implicit empty div spacer with dynamic justify-* class in WorkflowEditor Co-Authored-By: Claude <noreply@anthropic.com> * feat: add justfile for better command * feat: improve AGENTS.md and fix backend fix --------- Co-authored-by: Claude <noreply@anthropic.com> --------- Co-authored-by: xinquiry(SII) <100398322+xinquiry@users.noreply.github.com> Co-authored-by: Meng Junxing <junxingmeng@gmail.com> Co-authored-by: Claude <noreply@anthropic.com>
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
2 participants
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.
变更内容
简要描述本次 PR 的主要变更内容。
相关 Issue
请关联相关 Issue（如有）：#编号
检查清单
默认已勾选，如不满足，请检查。
其他说明
如有特殊说明或注意事项，请补充。
Summary by Sourcery
为知识工具操作引入专用的任务数据库会话上下文，并简化签到状态/积分的处理。
New Features:
Bug Fixes:
Enhancements:
Tests:
Original summary in English
Summary by Sourcery
Introduce a dedicated task database session context for knowledge tool operations and streamline check-in status/points handling.
New Features:
Bug Fixes:
Enhancements:
Tests: