jarvis-ai-assistant 0.1.225__tar.gz → 0.2.0__tar.gz

{jarvis_ai_assistant-0.1.225/src/jarvis_ai_assistant.egg-info → jarvis_ai_assistant-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: jarvis-ai-assistant
-Version: 0.1.225
+Version: 0.2.0
 Summary: Jarvis: An AI assistant that uses tools to interact with the system
 Home-page: https://github.com/skyfireitdiy/Jarvis
 Author: skyfire
@@ -508,9 +508,9 @@ ENV:
 #### Kimi
 ```yaml
 JARVIS_PLATFORM: kimi
-JARVIS_MODEL: kimi
+JARVIS_MODEL: k1.5
 JARVIS_THINKING_PLATFORM: kimi
-JARVIS_THINKING_MODEL: k1
+JARVIS_THINKING_MODEL: k1.5-thinking
 ENV:
   KIMI_API_KEY: <Kimi API KEY>
 ```
@@ -535,19 +535,55 @@ OPENAI_API_KEY: <OpenAI API Key>
 OPENAI_API_BASE: https://api.openai.com/v1
 ```
 
-### 2. 配置项说明
+### 2. 模型组配置 (高级)
+
+除了单独配置每个模型参数，您还可以定义和使用**模型组**来快速切换不同的模型组合。这对于需要在不同任务或平台间频繁切换的场景非常有用。
+
+**配置示例** (`~/.jarvis/config.yaml`):
+
+```yaml
+# 定义模型组
+JARVIS_MODEL_GROUPS:
+  - kimi:
+      JARVIS_PLATFORM: kimi
+      JARVIS_MODEL: k1.5
+      JARVIS_THINKING_PLATFORM: kimi
+      JARVIS_THINKING_MODEL: k1.5-thinking
+      JARVIS_MAX_TOKEN_COUNT: 8192
+  - ai8:
+      JARVIS_PLATFORM: ai8
+      JARVIS_MODEL: gemini-2.5-pro
+      # 如果不指定思考模型，将自动使用常规模型
+      # JARVIS_THINKING_PLATFORM: ai8
+      # JARVIS_THINKING_MODEL: gemini-2.5-pro
+
+# 选择要使用的模型组
+JARVIS_MODEL_GROUP: kimi
+```
+
+**配置优先级规则:**
+
+Jarvis 会按照以下顺序解析模型配置，序号越小优先级越高：
+
+1.  **独立配置**: 直接设置的 `JARVIS_PLATFORM`, `JARVIS_MODEL`, `JARVIS_THINKING_PLATFORM`, `JARVIS_THINKING_MODEL` 环境变量。这些配置会**覆盖**任何模型组中的设置。
+2.  **模型组配置**: 通过 `JARVIS_MODEL_GROUP` 选中的模型组配置。
+3.  **默认值**: 如果以上均未配置，则使用代码中定义的默认模型（如 `yuanbao` 和 `deep_seek_v3`）。
+
+### 3. 全部配置项说明
 | 变量名称 | 默认值 | 说明 |
 |----------|--------|------|
 | `ENV` | {} | 环境变量配置 |
-| `JARVIS_MAX_TOKEN_COUNT` | 960000 | 上下文窗口的最大token数量 |
-| `JARVIS_MAX_INPUT_TOKEN_COUNT` | 32000 | 输入的最大token数量 |
-| `JARVIS_PLATFORM` | yuanbao | 默认AI平台 |
-| `JARVIS_MODEL` | deep_seek_v3 | 默认模型 |
-| `JARVIS_THINKING_PLATFORM` | JARVIS_PLATFORM | 推理任务使用的平台 |
-| `JARVIS_THINKING_MODEL` | JARVIS_MODEL | 推理任务使用的模型 |
+| `JARVIS_MODEL_GROUPS` | `[]` | 预定义的模型配置组列表 |
+| `JARVIS_MODEL_GROUP` | `null` | 选择要激活的模型组名称 |
+| `JARVIS_MAX_TOKEN_COUNT` | 960000 | 上下文窗口的最大token数量 (可被模型组覆盖) |
+| `JARVIS_MAX_INPUT_TOKEN_COUNT` | 32000 | 输入的最大token数量 (可被模型组覆盖) |
+| `JARVIS_PLATFORM` | yuanbao | 默认AI平台 (可被模型组覆盖) |
+| `JARVIS_MODEL` | deep_seek_v3 | 默认模型 (可被模型组覆盖) |
+| `JARVIS_THINKING_PLATFORM` | JARVIS_PLATFORM | 推理任务使用的平台 (可被模型组覆盖) |
+| `JARVIS_THINKING_MODEL` | JARVIS_MODEL | 推理任务使用的模型 (可被模型组覆盖) |
 | `JARVIS_EXECUTE_TOOL_CONFIRM` | false | 执行工具前是否需要确认 |
 | `JARVIS_CONFIRM_BEFORE_APPLY_PATCH` | false | 应用补丁前是否需要确认 |
-| `JARVIS_MAX_BIG_CONTENT_SIZE` | 160000 | 最大大内容大小 |
+| `JARVIS_MAX_BIG_CONTENT_SIZE` | 160000 | 最大大内容大小 (可被模型组覆盖) |
 | `JARVIS_PRETTY_OUTPUT` | false | 是否启用PrettyOutput |
 | `JARVIS_GIT_COMMIT_PROMPT` | "" | 自定义git提交信息生成提示模板 |
 | `JARVIS_PRINT_PROMPT` | false | 是否打印提示 |

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/README.md

@@ -424,9 +424,9 @@ ENV:
 #### Kimi
 ```yaml
 JARVIS_PLATFORM: kimi
-JARVIS_MODEL: kimi
+JARVIS_MODEL: k1.5
 JARVIS_THINKING_PLATFORM: kimi
-JARVIS_THINKING_MODEL: k1
+JARVIS_THINKING_MODEL: k1.5-thinking
 ENV:
   KIMI_API_KEY: <Kimi API KEY>
 ```
@@ -451,19 +451,55 @@ OPENAI_API_KEY: <OpenAI API Key>
 OPENAI_API_BASE: https://api.openai.com/v1
 ```
 
-### 2. 配置项说明
+### 2. 模型组配置 (高级)
+
+除了单独配置每个模型参数，您还可以定义和使用**模型组**来快速切换不同的模型组合。这对于需要在不同任务或平台间频繁切换的场景非常有用。
+
+**配置示例** (`~/.jarvis/config.yaml`):
+
+```yaml
+# 定义模型组
+JARVIS_MODEL_GROUPS:
+  - kimi:
+      JARVIS_PLATFORM: kimi
+      JARVIS_MODEL: k1.5
+      JARVIS_THINKING_PLATFORM: kimi
+      JARVIS_THINKING_MODEL: k1.5-thinking
+      JARVIS_MAX_TOKEN_COUNT: 8192
+  - ai8:
+      JARVIS_PLATFORM: ai8
+      JARVIS_MODEL: gemini-2.5-pro
+      # 如果不指定思考模型，将自动使用常规模型
+      # JARVIS_THINKING_PLATFORM: ai8
+      # JARVIS_THINKING_MODEL: gemini-2.5-pro
+
+# 选择要使用的模型组
+JARVIS_MODEL_GROUP: kimi
+```
+
+**配置优先级规则:**
+
+Jarvis 会按照以下顺序解析模型配置，序号越小优先级越高：
+
+1.  **独立配置**: 直接设置的 `JARVIS_PLATFORM`, `JARVIS_MODEL`, `JARVIS_THINKING_PLATFORM`, `JARVIS_THINKING_MODEL` 环境变量。这些配置会**覆盖**任何模型组中的设置。
+2.  **模型组配置**: 通过 `JARVIS_MODEL_GROUP` 选中的模型组配置。
+3.  **默认值**: 如果以上均未配置，则使用代码中定义的默认模型（如 `yuanbao` 和 `deep_seek_v3`）。
+
+### 3. 全部配置项说明
 | 变量名称 | 默认值 | 说明 |
 |----------|--------|------|
 | `ENV` | {} | 环境变量配置 |
-| `JARVIS_MAX_TOKEN_COUNT` | 960000 | 上下文窗口的最大token数量 |
-| `JARVIS_MAX_INPUT_TOKEN_COUNT` | 32000 | 输入的最大token数量 |
-| `JARVIS_PLATFORM` | yuanbao | 默认AI平台 |
-| `JARVIS_MODEL` | deep_seek_v3 | 默认模型 |
-| `JARVIS_THINKING_PLATFORM` | JARVIS_PLATFORM | 推理任务使用的平台 |
-| `JARVIS_THINKING_MODEL` | JARVIS_MODEL | 推理任务使用的模型 |
+| `JARVIS_MODEL_GROUPS` | `[]` | 预定义的模型配置组列表 |
+| `JARVIS_MODEL_GROUP` | `null` | 选择要激活的模型组名称 |
+| `JARVIS_MAX_TOKEN_COUNT` | 960000 | 上下文窗口的最大token数量 (可被模型组覆盖) |
+| `JARVIS_MAX_INPUT_TOKEN_COUNT` | 32000 | 输入的最大token数量 (可被模型组覆盖) |
+| `JARVIS_PLATFORM` | yuanbao | 默认AI平台 (可被模型组覆盖) |
+| `JARVIS_MODEL` | deep_seek_v3 | 默认模型 (可被模型组覆盖) |
+| `JARVIS_THINKING_PLATFORM` | JARVIS_PLATFORM | 推理任务使用的平台 (可被模型组覆盖) |
+| `JARVIS_THINKING_MODEL` | JARVIS_MODEL | 推理任务使用的模型 (可被模型组覆盖) |
 | `JARVIS_EXECUTE_TOOL_CONFIRM` | false | 执行工具前是否需要确认 |
 | `JARVIS_CONFIRM_BEFORE_APPLY_PATCH` | false | 应用补丁前是否需要确认 |
-| `JARVIS_MAX_BIG_CONTENT_SIZE` | 160000 | 最大大内容大小 |
+| `JARVIS_MAX_BIG_CONTENT_SIZE` | 160000 | 最大大内容大小 (可被模型组覆盖) |
 | `JARVIS_PRETTY_OUTPUT` | false | 是否启用PrettyOutput |
 | `JARVIS_GIT_COMMIT_PROMPT` | "" | 自定义git提交信息生成提示模板 |
 | `JARVIS_PRINT_PROMPT` | false | 是否打印提示 |

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "jarvis-ai-assistant"
-version = "0.1.225"
+version = "0.2.0"
 description = "Jarvis: An AI assistant that uses tools to interact with the system"
 readme = "README.md"
 authors = [{ name = "skyfire", email = "skyfireitdiy@hotmail.com" }]

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/setup.py

@@ -3,7 +3,7 @@ from setuptools import setup, find_packages  # type: ignore
 
 setup(
     name="jarvis-ai-assistant",
-    version="0.1.225",
+    version="0.2.0",
     author="skyfire",
     author_email="skyfireitdiy@hotmail.com",
     description="An AI assistant that uses various tools to interact with the system",

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/__init__.py

@@ -1,4 +1,4 @@
 # -*- coding: utf-8 -*-
 """Jarvis AI Assistant"""
 
-__version__ = "0.1.225"
+__version__ = "0.2.0"

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_agent/__init__.py

@@ -102,6 +102,7 @@ class Agent:
         name: str = "Jarvis",
         description: str = "",
         llm_type: str = "normal",
+        model_group: Optional[str] = None,
         summary_prompt: Optional[str] = None,
         auto_complete: bool = False,
         output_handler: List[OutputHandlerProtocol] = [],
@@ -137,11 +138,11 @@ class Agent:
         self.description = description
         # 初始化平台和模型
         if llm_type == "thinking":
-            platform_name = get_thinking_platform_name()
-            model_name = get_thinking_model_name()
+            platform_name = get_thinking_platform_name(model_group)
+            model_name = get_thinking_model_name(model_group)
         else:  # 默认为 normal
-            platform_name = get_normal_platform_name()
-            model_name = get_normal_model_name()
+            platform_name = get_normal_platform_name(model_group)
+            model_name = get_normal_model_name(model_group)
 
         self.model = PlatformRegistry().create_platform(platform_name)
         if self.model is None:
@@ -153,6 +154,8 @@ class Agent:
         if model_name:
             self.model.set_model_name(model_name)
 
+        self.model.set_model_group(model_group)
+
         self.user_data: Dict[str, Any] = {}
 
         self.model.set_suppress_output(False)
@@ -197,7 +200,7 @@ class Agent:
             summary_prompt if summary_prompt else DEFAULT_SUMMARY_PROMPT
         )
 
-        self.max_token_count = get_max_token_count()
+        self.max_token_count = get_max_token_count(model_group)
         self.auto_complete = auto_complete
         welcome_message = f"{name} 初始化完成 - 使用 {self.model.name()} 模型"
 

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_agent/jarvis.py

@@ -125,6 +125,11 @@ def _parse_args() -> argparse.Namespace:
         type=str,
         help="Directly input task content from command line",
     )
+    parser.add_argument(
+        "--model_group",
+        type=str,
+        help="Model group to use, overriding config",
+    )
     parser.add_argument("-f", "--config", type=str, help="Path to custom config file")
     parser.add_argument(
         "--restore-session",
@@ -170,6 +175,7 @@ def _initialize_agent(args: argparse.Namespace) -> Agent:
     agent = Agent(
         system_prompt=origin_agent_system_prompt,
         llm_type=args.llm_type,
+        model_group=args.model_group,
         input_handler=[shell_input_handler, builtin_input_handler],
         output_handler=[ToolRegistry()],  # type: ignore
         need_summary=False,

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_agent/main.py

@@ -52,6 +52,11 @@ def main():
         choices=["normal", "thinking"],
         help="LLM type to use, overriding config",
     )
+    parser.add_argument(
+        "--model_group",
+        type=str,
+        help="Model group to use, overriding config",
+    )
     args = parser.parse_args()
 
     # Initialize environment
@@ -65,6 +70,8 @@ def main():
     # Override config with command-line arguments if provided
     if args.llm_type:
         config["llm_type"] = args.llm_type
+    if args.model_group:
+        config["model_group"] = args.model_group
 
     # Create and run agent
     try:

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_code_agent/code_agent.py

@@ -47,6 +47,7 @@ class CodeAgent:
     def __init__(
         self,
         llm_type: str = "normal",
+        model_group: Optional[str] = None,
         need_summary: bool = True,
     ):
         self.root_dir = os.getcwd()
@@ -120,6 +121,7 @@ class CodeAgent:
             auto_complete=False,
             output_handler=[tool_registry, EditFileHandler()],  # type: ignore
             llm_type=llm_type,
+            model_group=model_group,
             input_handler=[shell_input_handler, builtin_input_handler],
             need_summary=need_summary,
             use_methodology=False,  # 禁用方法论
@@ -411,6 +413,11 @@ def main() -> None:
         choices=["normal", "thinking"],
         help="LLM type to use",
     )
+    parser.add_argument(
+        "--model_group",
+        type=str,
+        help="Model group to use, overriding config",
+    )
     parser.add_argument(
         "-r", "--requirement", type=str, help="Requirement to process", default=None
     )
@@ -427,7 +434,11 @@ def main() -> None:
     PrettyOutput.print(f"当前目录: {git_dir}", OutputType.INFO)
 
     try:
-        agent = CodeAgent(llm_type=args.llm_type, need_summary=False)
+        agent = CodeAgent(
+            llm_type=args.llm_type,
+            model_group=args.model_group,
+            need_summary=False,
+        )
 
         # 尝试恢复会话
         if args.restore_session:

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_code_analysis/code_review.py

@@ -3,12 +3,13 @@ import os
 import re
 import subprocess
 import tempfile
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Optional
 
 from jarvis.jarvis_agent import Agent
 from jarvis.jarvis_code_analysis.checklists.loader import get_language_checklist
 from jarvis.jarvis_platform.registry import PlatformRegistry
 from jarvis.jarvis_tools.read_code import ReadCodeTool
+from jarvis.jarvis_utils.globals import get_agent, current_agent_name
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 from jarvis.jarvis_utils.tag import ct, ot
 from jarvis.jarvis_utils.utils import init_env, is_context_overflow
@@ -261,7 +262,9 @@ class CodeReviewTool:
         checklist = get_language_checklist(language)
         return checklist if checklist else ""
 
-    def execute(self, args: Dict[str, Any]) -> Dict[str, Any]:
+    def execute(
+        self, args: Dict[str, Any], agent: Optional["Agent"] = None
+    ) -> Dict[str, Any]:
         try:
             review_type = args.get("review_type", "current").strip()
             root_dir = args.get("root_dir", ".")
@@ -570,9 +573,17 @@ class CodeReviewTool:
 
                 tool_registry = ToolRegistry()
                 tool_registry.dont_use_tools(["code_review"])
+                
+                # Use the provided agent's model_group or get it from globals
+                calling_agent = agent or get_agent(current_agent_name)
+                model_group = None
+                if calling_agent and hasattr(calling_agent, "model") and calling_agent.model:
+                    model_group = calling_agent.model.model_group
+
                 agent = Agent(
                     system_prompt=system_prompt,
                     name="Code Review Agent",
+                    model_group=model_group,
                     summary_prompt=f"""<code_review_report>
 <overview>
 # 整体评估
@@ -675,7 +686,7 @@ class CodeReviewTool:
 
                 try:
                     # Check if content is too large
-                    is_large_content = is_context_overflow(diff_output)
+                    is_large_content = is_context_overflow(diff_output, model_group)
 
                     # Upload the file to the agent's model
                     if is_large_content:

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_data/config_schema.json

@@ -141,6 +141,47 @@
       "description": "思考操作模型名称",
       "default": "deep_seek"
     },
+    "JARVIS_MODEL_GROUP": {
+      "type": "string",
+      "description": "选择一个预定义的模型组"
+    },
+    "JARVIS_MODEL_GROUPS": {
+      "type": "array",
+      "description": "预定义的模型配置组",
+      "items": {
+        "type": "object",
+        "additionalProperties": {
+          "type": "object",
+          "properties": {
+            "JARVIS_PLATFORM": {
+              "type": "string"
+            },
+            "JARVIS_MODEL": {
+              "type": "string"
+            },
+            "JARVIS_THINKING_PLATFORM": {
+              "type": "string"
+            },
+            "JARVIS_THINKING_MODEL": {
+              "type": "string"
+            },
+            "JARVIS_MAX_TOKEN_COUNT": {
+              "type": "number"
+            },
+            "JARVIS_MAX_INPUT_TOKEN_COUNT": {
+              "type": "number"
+            },
+            "JARVIS_MAX_BIG_CONTENT_SIZE": {
+              "type": "number"
+            }
+          },
+          "required": [
+            "JARVIS_PLATFORM",
+            "JARVIS_MODEL"
+          ]
+        }
+      }
+    },
     "JARVIS_EXECUTE_TOOL_CONFIRM": {
       "type": "boolean",
       "description": "执行工具前是否需要确认",

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_git_utils/git_commiter.py

@@ -16,6 +16,7 @@ from jarvis.jarvis_utils.git_utils import (
     find_git_root_and_cd,
     has_uncommitted_changes,
 )
+from jarvis.jarvis_utils.globals import get_agent, current_agent_name
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 from jarvis.jarvis_utils.tag import ct, ot
 from jarvis.jarvis_utils.utils import init_env, is_context_overflow
@@ -162,11 +163,17 @@ commit信息
                 """
 
                 # 获取模型并尝试上传文件
-                platform = PlatformRegistry().get_normal_platform()
+                agent = get_agent(current_agent_name)
+                if agent:
+                    platform = agent.model
+                    model_group = agent.model.model_group
+                else:
+                    platform = PlatformRegistry().get_normal_platform()
+                    model_group = None
                 upload_success = False
 
                 # Check if content is too large
-                is_large_content = is_context_overflow(diff)
+                is_large_content = is_context_overflow(diff, model_group)
 
                 if is_large_content:
                     if not platform.support_upload_files():

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_mcp/sse_mcp_client.py

@@ -2,7 +2,7 @@
 import json
 import threading
 import time
-from typing import Any, Callable, Dict, List
+from typing import Any, Callable, Dict, List, Optional
 from urllib.parse import parse_qs, urlencode, urljoin
 
 import requests
@@ -46,10 +46,10 @@ class SSEMcpClient(McpClient):
         self.session.headers.update(extra_headers)
 
         # SSE相关属性
-        self.sse_response = None
-        self.sse_thread = None
-        self.messages_endpoint = None
-        self.session_id = None  # 从SSE连接获取的会话ID
+        self.sse_response: Optional[requests.Response] = None
+        self.sse_thread: Optional[threading.Thread] = None
+        self.messages_endpoint: Optional[str] = None
+        self.session_id: Optional[str] = None
         self.pending_requests = {}  # 存储等待响应的请求 {id: Event}
         self.request_results = {}  # 存储请求结果 {id: result}
         self.notification_handlers = {}
@@ -123,13 +123,15 @@ class SSEMcpClient(McpClient):
             self.sse_response = self.session.get(
                 sse_url, stream=True, headers=sse_headers, timeout=30
             )
-            self.sse_response.raise_for_status()
+            if self.sse_response:
+                self.sse_response.raise_for_status()
 
             # 启动事件处理线程
             self.sse_thread = threading.Thread(
                 target=self._process_sse_events, daemon=True
             )
-            self.sse_thread.start()
+            if self.sse_thread:
+                self.sse_thread.start()
 
         except Exception as e:
             PrettyOutput.print(f"SSE连接失败: {str(e)}", OutputType.ERROR)

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_mcp/stdio_mcp_client.py

@@ -2,7 +2,7 @@
 import json
 import os
 import subprocess
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Optional
 
 from jarvis.jarvis_mcp import McpClient
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
@@ -17,7 +17,7 @@ class StdioMcpClient(McpClient):
 
     def __init__(self, config: Dict[str, Any]):
         self.config = config
-        self.process = None
+        self.process: Optional[subprocess.Popen] = None
         self.protocol_version = "2025-03-26"  # MCP协议版本
         self._start_process()
         self._initialize()

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_platform/base.py

@@ -1,7 +1,10 @@
 # -*- coding: utf-8 -*-
 import re
 from abc import ABC, abstractmethod
-from typing import Generator, List, Tuple
+from types import TracebackType
+from typing import Generator, List, Optional, Tuple, Type
+
+from typing_extensions import Self
 
 from rich import box  # type: ignore
 from rich.live import Live  # type: ignore
@@ -28,9 +31,19 @@ class BasePlatform(ABC):
         self.suppress_output = True  # 添加输出控制标志
         self.web = False  # 添加web属性，默认false
         self._saved = False
-
-    def __del__(self):
-        """Destroy model"""
+        self.model_group: Optional[str] = None
+
+    def __enter__(self) -> Self:
+        """Enter context manager"""
+        return self
+
+    def __exit__(
+        self,
+        exc_type: Optional[Type[BaseException]],
+        exc_val: Optional[BaseException],
+        exc_tb: Optional[TracebackType],
+    ) -> None:
+        """Exit context manager"""
         if not self._saved:
             self.delete_chat()
 
@@ -64,9 +77,11 @@ class BasePlatform(ABC):
 
         input_token_count = get_context_token_count(message)
 
-        if input_token_count > get_max_input_token_count():
-            max_chunk_size = get_max_input_token_count() - 1024  # 留出一些余量
-            min_chunk_size = get_max_input_token_count() - 2048
+        if input_token_count > get_max_input_token_count(self.model_group):
+            max_chunk_size = (
+                get_max_input_token_count(self.model_group) - 1024
+            )  # 留出一些余量
+            min_chunk_size = get_max_input_token_count(self.model_group) - 2048
             inputs = split_text_into_chunks(message, max_chunk_size, min_chunk_size)
             print("📤 正在提交长上下文...")
             prefix_prompt = f"""
@@ -95,14 +110,10 @@ class BasePlatform(ABC):
                 ):
                     response += trunk
 
-                print(
-                    f"📤 提交第{submit_count}部分完成，当前进度：{length}/{len(message)}"
-                )
+                print(f"📤 提交第{submit_count}部分完成，当前进度：{length}/{len(message)}")
             print("✅ 提交完成")
             response += "\n" + while_true(
-                lambda: while_success(
-                    lambda: self._chat("内容已经全部提供完毕，请根据内容继续"), 5
-                ),
+                lambda: while_success(lambda: self._chat("内容已经全部提供完毕，请根据内容继续"), 5),
                 5,
             )
         else:
@@ -233,6 +244,10 @@ class BasePlatform(ABC):
         """Set whether to suppress output"""
         self.suppress_output = suppress
 
+    def set_model_group(self, model_group: Optional[str]):
+        """Set model group"""
+        self.model_group = model_group
+
     def set_web(self, web: bool):
         """Set web flag"""
         self.web = web

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_tools/generate_new_tool.py

@@ -12,6 +12,7 @@ class generate_new_tool:
     生成并注册新的Jarvis工具。该工具会在用户数据目录下创建新的工具文件，
     并自动注册到当前的工具注册表中。适用场景：1. 需要创建新的自定义工具；
     2. 扩展Jarvis功能；3. 自动化重复性操作；4. 封装特定领域的功能。
+    重要提示：在编写工具代码时，应尽量将工具执行的过程和结果打印出来，方便追踪工具的执行状态。
     """
 
     parameters = {

{jarvis_ai_assistant-0.1.225 → jarvis_ai_assistant-0.2.0}/src/jarvis/jarvis_tools/registry.py

@@ -17,7 +17,7 @@ from jarvis.jarvis_tools.base import Tool
 from jarvis.jarvis_utils.config import get_data_dir, get_tool_load_dirs
 from jarvis.jarvis_utils.output import OutputType, PrettyOutput
 from jarvis.jarvis_utils.tag import ct, ot
-from jarvis.jarvis_utils.utils import is_context_overflow
+from jarvis.jarvis_utils.utils import is_context_overflow, daily_check_git_updates
 
 tool_call_help = f"""
 <tool_system_guide>
@@ -124,7 +124,7 @@ class ToolRegistry(OutputHandlerProtocol):
         """加载工具"""
         tools = self.get_all_tools()
         if tools:
-            tools_prompt = "<tools_section>\n"
+            tools_prompt = f"{tool_call_help}\n<tools_section>\n"
             tools_prompt += "  <header>## 可用工具:</header>\n"
             tools_prompt += "  <tools_list>\n"
             for tool in tools:
@@ -287,16 +287,18 @@ class ToolRegistry(OutputHandlerProtocol):
 
     def _load_external_tools(self) -> None:
         """从jarvis_data/tools和配置的目录加载外部工具"""
-        tool_dirs = [Path(get_data_dir()) / "tools"] + [
-            Path(p) for p in get_tool_load_dirs()
-        ]
+        tool_dirs = [str(Path(get_data_dir()) / "tools")] + get_tool_load_dirs()
+
+        # --- 全局每日更新检查 ---
+        daily_check_git_updates(tool_dirs, "tools")
 
         for tool_dir in tool_dirs:
-            if not tool_dir.exists():
+            p_tool_dir = Path(tool_dir)
+            if not p_tool_dir.exists() or not p_tool_dir.is_dir():
                 continue
 
             # 遍历目录中的所有.py文件
-            for file_path in tool_dir.glob("*.py"):
+            for file_path in p_tool_dir.glob("*.py"):
                 # 跳过__init__.py
                 if file_path.name == "__init__.py":
                     continue
@@ -717,7 +719,10 @@ class ToolRegistry(OutputHandlerProtocol):
             )
 
             # 检查内容是否过大
-            is_large_content = is_context_overflow(output)
+            model_group = None
+            if agent_instance.model:
+                model_group = agent_instance.model.model_group
+            is_large_content = is_context_overflow(output, model_group)
 
             if is_large_content:
                 # 创建临时文件