laoxong/AstrBot
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
4,368 Commits 55 Branches 213 Tags
dev
Commit Graph

18 Commits

Author SHA1 Message Date
zenfun e85eef05b8 fix: stabilize tool injection for LLM prefix cache hits 
Two changes to make the tool schema sent to the LLM deterministic:
1. ToolSet.normalize() — sort tools by name before serialization.
   Called at the end of build_main_agent() after all injection passes.
   Eliminates ordering drift from plugin load order, MCP reconnection,
   and persona tool list differences.
2. Always inject full sandbox tool set — ComputerToolProvider now
   returns get_default_sandbox_tools() unconditionally, regardless of
   sandbox boot state. Browser tools are always in the schema even if
   the sandbox profile lacks browser capability. The executor rejects
   calls to unavailable browser tools with a descriptive error instead
   of silently omitting them from the schema.
   This eliminates the pre-boot/post-boot tool set jump that caused
   prefix cache misses on the second request of a conversation.
 2026-03-12 02:43:19 +08:00
advent259141 ff4412a627 refactor: Centralize and decouple computer-use tool injection logic into a new ComputerToolProvider and associated tool modules. 2026-03-10 22:00:23 +08:00
Soulter cda49c3a9a fix: remove additionalProperties from tool schema properties (#5253) 
fixes: #5217
 2026-02-20 16:13:20 +08:00
Dt8333 7dd95d8a59 chore: auto ann fix by ruff (#4903) 
* chore: auto fix by ruff

* refactor: 统一修正返回类型注解为 None/bool 以匹配实现

* refactor: 将 _get_next_page 改为异步并移除多余的请求错误抛出

* refactor: 将 get_client 的返回类型改为 object

* style: 为 LarkMessageEvent 的相关方法添加返回类型注解 None

---------

Co-authored-by: Soulter <37870767+Soulter@users.noreply.github.com>
 2026-02-09 00:22:24 +08:00
boushi1111 d69592aaa8 fix: TypeError when MCP schema type is a list (#4867) 
* Fix TypeError when MCP schema type is a list

Fixes crash in Gemini native tools with VRChat MCP.

* Refactor: avoid modifying schema in place per feedback

* Fix formatting and cleanup comments
 2026-02-05 22:51:29 +08:00
Soulter 4ea865f017 feat: add cron job management tools and dashboard integration 
- Implemented proactive cron job tools in InternalAgentSubStage for scheduling tasks.
- Created SendMessageToUserTool for sending messages to users based on cron job triggers.
- Added CreateActiveCronTool, DeleteCronJobTool, and ListCronJobsTool for cron job management.
- Introduced CronRoute for handling cron job API requests in the dashboard.
- Developed CronJobPage.vue for managing cron jobs in the dashboard UI.
- Updated SubAgentPage.vue to include persona selection for subagents.
 2026-01-31 17:08:37 +08:00
vmoranv f92f0a3e5d feat(core): supports anthropic-skills-like tool call mode (#4681) 
* feat(core): change llmtool to claude skills like func call

* feat: refactor tool execution logic in ToolLoopAgentRunner for improved clarity and efficiency

* feat(core): 添加工具调用模式配置选项

新增 tool_schema_mode 配置项，支持两种工具调用模式：
- skills_like：先发送工具名称和描述，再查询参数（两阶段）
- full：一次性发送完整工具模式

更新了默认配置、配置元数据定义以及代理子阶段处理逻辑，
添加了完整的工具调用提示语句，并在仪表板中提供了国际化支持。

* feat: 优化工具集获取逻辑，添加轻量和参数工具集返回方法

* refactor(runner): 重构工具模式处理逻辑到ToolLoopAgentRunner

- 将工具集激活逻辑提取到新的_build_active_tool_set方法中
- 实现工具模式配置功能，支持full和light模式的动态切换
- 移除InternalAgentSubStage中的工具模式应用逻辑，统一在runner中处理
- 添加_tool_schema_full_set和_tool_schema_param_set实例变量来管理工具集状态
- 修改工具查询逻辑以使用新的工具集管理方式

* fix: update default tool_schema_mode to 'full' in InternalAgentSubStage

* refactor: rename TOOL_CALL_PROMPT_FULL to TOOL_CALL_PROMPT_SKILLS_LIKE_MODE and update prompt logic

---------

Co-authored-by: Soulter <905617992@qq.com>
 2026-01-28 22:49:34 +08:00
Soulter 22bd8d6824 feat: support anthropic skills (#4715) 
* feat: support anthropic skills

closes: #4687

* chore: ruff

* feat: implement skills management and selection in persona configuration

* feat: enhance skills management with local environment tools and permissions
 2026-01-28 01:48:57 +08:00
Dt8333 f624971613 chore: fix bunches of type checking errors (#3213) 
* chore(core.utils): 🚨 修正错误Lint

* chore(core.provider): 🚨 修复基类错误Lint

* chore(core.utils): 补全session_get()的重载

* chore(core.provider): 🚨 修正实现错误Lint

* chore(core.platform): 🚨 修正platform基类和webchat的错误Lint

* chore(core.platform): 修正错误实现Lint

* fix(core.provider): 修复循环调用和错误assert

* chore(core.platform): 修复部分实现Lint

* chore(core.provider): 补充Dify.text_chat_stream的参数类型

* chore(core.pipeline): 🚨 修复错误Lint

* fix(core.slack): 补充遗漏导入

* chore(core.utils): 修复错误的session_get声明

* chore(core.platform): 移除Lark adapter import中的wildcard

* chore(core.db): 修复声明和部分逻辑

* chore(core.db): 添加typings，使faiss参数能被正确识别。

* chore(core): 修复声明

* chore(core): 修改声明

* chore: 补充faiss声明

* chore(dashboard): 修改实现，减少报错

* chore(package): 修改部分声明与实现，减少报错

* chore(core): 添加Handler的overload，以去除部分assert同时通过类型检查

* chore(core.pipeline): 修改Pipeline Scheduler的execute，将判断属性改为判断类型，通过静态类型检查

* chore(core.config): 添加类型标注，通过类型检查

* chore(core.message): 为File._download_file添加检查，通过类型检查

* fix: 将断言改为条件判断以实现优雅关闭的容错性

* refactor: 移除 discord 客户端中的 assert，改用 if None 判断并抛出异常

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: DiscordPlatformAdapter 对 self.client.user 为 None 做日志并返回，移除断言

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 增强 Lark 相关空值/异常检查并完善日志输出

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 将断言替换为条件检查并加入日志与错误处理

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* chore: 移除LLM生成的无用注释

* refactor: 使用 File.get_file 替换下载逻辑并移除 assert，提供默认 filename

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: Slack Socket 未初始化抛出运行时异常，图片 URL 判空改为非空判断

* refactor: 将 WeChatPadProAdapter 的断言改为空值判断并添加日志

* refactor: 使用 isinstance 替代断言实现类型判断，便于静态检查

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 去除cast，直接使用字段与字典访问，修正端口解析

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 使用 match-case 重构 ProviderManager 加载并通过类型检查抛出 TypeError

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: group_name_display 时若 group 对象为空则记录错误并返回

* fix: 将 _get_current_persona_id 的 assert 替换成 if guard 并返回 None

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 优化插件目录存在性检查及图片URL非空验证，更新JSON排序配置

* fix: 将 datetime_str 的 assert 替换为显式检查并抛出异常

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 移除 cast，改为运行时检查并在找不到调度器时跳过

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 移除 cast，改用 isinstance 检查 FaissVecDB 并警告

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 删除 typing.cast 导入，并在获取文件绝对路径前校验 file_

* refactor: 移除 typing.cast，简化内容安全检查调用

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 将 PlatformMetadata.id 设为必填并在注册时传入 id，移除 cast

* refactor: 移除 cast，改用 HasInitialize 与 isinstance 进行初始化

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 为 ProviderManager.initialize 增加ID类型判断，避免 None 导致 get 失败

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 为 OTTSProvider 与 AzureNativeProvider 引入 _client 与 client 属性改进上下文管理

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 为 Whisper 自托管源添加模型未初始化校验并直接调用 transcribe

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 移除未使用的 cast 导入并简化 platform_name 赋值

* refactor: 引入 cast 并对 id 使用 cast(str, ...) 提升类型安全

* fix: 将 _id_to_sid 返回改为 str，空值返回空串；对 id 与 message_id 使用 cast

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 重构 Discord 处理逻辑：强制 类型转换、优先斜杠指令并优化提及判断

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* fix: 统一对 id 获取执行 cast，并在微信消息解析失败时抛错

* Revert "fix: 去除cast，直接使用字段与字典访问，修正端口解析"

This reverts commit 1cbfdf9d1b.

* fix: 百炼 Rerank 会话关闭时返回空结果；初始化 request.prompt 避免空值拼接

* fix: 统一处理搜索结果链接为字符串，新增 _get_url 助手并适配 Bing/Sogo

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>

* refactor: 调整 call_handler 泛型、Discord 通道注解及 FishAudioTTS API 请求类型

* refactor: 使用 col(...) 替代列引用并对结果进行 CursorResult 强转

* chore: ruff format

---------

Co-authored-by: aider (openai/gemini-3-pro-high) <aider@aider.chat>
Co-authored-by: Soulter <905617992@qq.com>
 2025-12-09 14:13:47 +08:00
Soulter fd9cb703db refactor: update ToolSet initialization to use Pydantic Field and clean up deprecated methods in Context 2025-11-16 12:13:11 +08:00
Soulter 17422ba9c3 feat: introduce messages field in agent RunContext 2025-11-15 21:15:20 +08:00
Soulter 91136bb9f7 fix: llm tool register error (#3493) 2025-11-05 14:27:37 +08:00
Soulter 50144ddcae refactor: revise LLM message schema and fix the reload logic when using dataclass-based LLM Tool registration (#3234) 
* refactor: llm message schema

* feat: implement MCPTool and local LLM tools with enhanced context handling

* refactor: reorganize imports and enhance docstrings for clarity

* refactor: enhance ContentPart validation and add message pair handling in ConversationManager

* chore: ruff format

* refactor: remove debug print statement from payloads in ProviderOpenAIOfficial

* Update astrbot/core/agent/tool.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update astrbot/core/agent/message.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update astrbot/core/agent/message.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update astrbot/core/agent/tool.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update astrbot/core/pipeline/process_stage/method/llm_request.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update astrbot/core/agent/message.py

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* refactor: enhance documentation and import mcp in tool.py; update call method return type

* fix: 修复以数据类的方式注册 tool 时的插件重载机制问题

* refactor: change role attributes to use Literal types for message segments

* fix: add support for 'decorator_handler' method in call_local_llm_tool

* fix: handle None prompt in text_chat method and ensure context is properly formatted

---------

Co-authored-by: LIghtJUNction <lightjunction.me@gmail.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
 2025-11-02 18:12:20 +08:00
LIghtJUNction 0b7fc29ac4 style: add ruff lint module of isort and pyupgrade, and some ruff check fix (#3214) 
Co-authored-by: Dt8333 <25431943+Dt8333@users.noreply.github.com>
Co-authored-by: Soulter <905617992@qq.com>
 2025-11-01 13:26:19 +08:00
Soulter 80a86f5b1b fix: 修复 astrbot.core.star 等包下的 type checking error (#2787) 
* fix: 修复 astrbot.core.star 等包下的 type checking error

* refactor: improve type checking and annotations

* chore: ruff format
 2025-09-21 18:10:04 +08:00
Soulter 69aaf09ac8 chore: 修复当自动更新 webchat title 时，history 被重置的问题 2025-08-24 00:23:08 +08:00
Soulter 563972fd29 fix: bugfixes 2025-08-22 17:41:06 +08:00
Soulter 1df49d1d6f refactor: 重构 Function Tool 管理并初步引入 Multi Agent 及 Agent Handsoff 机制 (#2454) 
* stage

* refactor: 重构 Function Tool 管理并引入 multi agent handsoff 机制

- Updated `star_request.py` to use the global `call_handler` instead of context-specific calls.
- Modified `entities.py` to remove the dependency on `FunctionToolManager` and streamline the function tool handling.
- Refactored `func_tool_manager.py` to simplify the `FunctionTool` class and its methods, removing deprecated code and enhancing clarity.
- Adjusted `provider.py` to align with the new function tool structure, removing unnecessary type unions.
- Enhanced `star_handler.py` to support agent registration and tool association, introducing `RegisteringAgent` for better encapsulation.
- Updated `star_manager.py` to handle tool registration for agents, ensuring proper binding of handlers.
- Revised `main.py` in the web searcher package to utilize the new agent registration system for web search tools.

* chore: websearch

* perf: 减少嵌套

* chore: 移除未使用的 mcp 导入
 2025-08-17 10:57:25 +08:00