ver0.4.2 修复目前反馈遇到的问题

chatgpt_tool_hub/apps/app.py

@@ -2,7 +2,7 @@
 
 from typing import List
 
-from chatgpt_tool_hub.chains.base import Chain
+from chatgpt_tool_hub.engine.tool_engine import ToolEngine
 from chatgpt_tool_hub.common.log import LOG
 from chatgpt_tool_hub.tools.base_tool import BaseTool
 
@@ -11,7 +11,7 @@ class App:
     _instance = None  # 存储单例实例
     init_flag = False  # 记录是否执行过初始化动作
 
-    engine: Chain = None
+    engine: ToolEngine = None
 
     # 当前已加载工具
     tools: set = set()
@@ -76,4 +76,4 @@ def _check_mandatory_tools(self, use_tools: list) -> bool:
         return True
 
     def get_tool_list(self) -> List[str]:
-        return list(self.tools)
+        return list(self.engine.get_tool_list())

chatgpt_tool_hub/apps/app_factory.py

@@ -1,4 +1,5 @@
 import logging
+import os
 
 from rich.console import Console
 
@@ -33,10 +34,16 @@ def init_env(self, **kwargs):
         else:
             self.default_tools_list = ["python", "terminal", "url-get", "meteo-weather"]
 
+        # set proxy
+        # _proxy = get_from_dict_or_env(kwargs, "proxy", "PROXY", "")
+        # if not _proxy:
+        #     os.environ["http_proxy"] = str(_proxy)
+        #     os.environ["https_proxy"] = str(_proxy)
+
         # dynamic loading tool
         dynamic_tool_loader()
 
-    def create_app(self, app_type: str = 'victorinox', tools_list: list = None, **kwargs) -> App:
+    def create_app(self, app_type: str = 'victorinox', tools_list: list = None, console=Console(quiet=True), **kwargs) -> App:
         tools_list = tools_list if tools_list else []
 
         self.init_env(**kwargs)
@@ -58,7 +65,7 @@ def create_app(self, app_type: str = 'victorinox', tools_list: list = None, **kw
             if "browser" in tools_list:
                 tools_list = list(filter(lambda tool: tool != "url-get", tools_list))
 
-            app = Victorinox(**build_model_params(kwargs))
+            app = Victorinox(console, **build_model_params(kwargs))
             app.create(tools_list, **kwargs)
             return app
         else:

chatgpt_tool_hub/bots/chat_bot/base.py

@@ -173,7 +173,7 @@ def _extract_tool_and_input(self, llm_output: str) -> Optional[Tuple[str, str]]:
 
         if action.lower() != "answer-user":
             self.console.print(f"√ 我在用 [bold cyan]{action}[/] 工具...")
-        # todo
+
         LOG.info(f"执行Tool: {action}中...")
         return action.strip(), action_input.strip()
 
@@ -223,7 +223,8 @@ def parse_reply_json(self, assistant_reply) -> dict:
                                      title=f"{self.ai_prefix.upper()}的内心独白",
                                      highlight=True, style='dim'))
             # it's useful for avoid splitting Panel
-            self.console.print("\n")
+            LOG.info(f"{self.ai_prefix.upper()}的内心独白: {thoughts_text}")
+
             return assistant_reply_json
         except json.decoder.JSONDecodeError as e:
             call_stack = traceback.format_exc()

chatgpt_tool_hub/bots/chat_bot/prompt.py

@@ -27,7 +27,7 @@
         "speak": "thoughts summary to say to {human_prefix}",
     }}}},
     "tool": {{{{
-        "name": "the tool to use, You must use one of the tools from the list: [{tool_names}]", 
+        "name": "the tool to use, You must use one of the tools from the list: [{tool_names}, answer-user]", 
         "input": "the input to the tool" 
     }}}}
 }}}}

chatgpt_tool_hub/chains/api/base.py

@@ -5,6 +5,7 @@
 
 from pydantic import BaseModel, Field, root_validator
 from rich.console import Console
+from rich.panel import Panel
 
 from chatgpt_tool_hub.chains.api.prompt import API_RESPONSE_PROMPT, API_URL_PROMPT
 from chatgpt_tool_hub.chains.base import Chain
@@ -69,12 +70,21 @@ def _call(self, inputs: Dict[str, str]) -> Dict[str, str]:
         api_url = self.api_request_chain.predict(
             question=question, api_docs=self.api_docs
         )
-        LOG.debug(f"[API] generate url: {str(api_url)}")
+        self.console.print(Panel(f"{str(api_url)}",
+                                 title=f"[bright_magenta]URL 构造[/]",
+                                 highlight=True))
+
+        LOG.info(f"URL 构造: {str(api_url)}")
         self.callback_manager.on_text(
             api_url, color="green", end="\n", verbose=self.verbose
         )
         api_response = self.requests_wrapper.get(api_url)
-        LOG.debug(f"[API] response: {str(api_response)}")
+
+        self.console.print(Panel(f"{repr(api_response)}",
+                                 title=f"[bright_magenta]API 响应[/]",
+                                 highlight=True))
+
+        LOG.info(f"API 响应: {repr(api_response)}")
         self.callback_manager.on_text(
             api_response, color="yellow", end="\n", verbose=self.verbose
         )

chatgpt_tool_hub/common/log.py

@@ -6,13 +6,12 @@
 
 def _get_logger(level: int = LOGGING_LEVEL):
     logger = logging.getLogger("tool")
+    logger.setLevel(level)
 
     ch = logging.StreamHandler(sys.stdout)
-    ch.setLevel(level)
     ch.setFormatter(logging.Formatter(LOGGING_FMT, datefmt=LOGGING_DATEFMT))
 
     fh = logging.FileHandler(f'{os.getcwd()}/tool.log', encoding='utf-8')
-    fh.setLevel(logging.DEBUG)
     fh.setFormatter(logging.Formatter(LOGGING_FMT, datefmt=LOGGING_DATEFMT))
 
     logger.addHandler(ch)

chatgpt_tool_hub/database/chat_memory.py

@@ -38,8 +38,8 @@ def save_context(self, inputs: Dict[str, Any], outputs: Dict[str, str]) -> None:
             output_key = list(outputs.keys())[0]
         else:
             output_key = self.output_key
-        self.chat_memory.add_user_message(inputs[prompt_input_key])
-        self.chat_memory.add_ai_message(outputs[output_key])
+        self.chat_memory.add_user_message(repr(inputs[prompt_input_key]))
+        self.chat_memory.add_ai_message(repr(outputs[output_key]))
 
     def clear(self) -> None:
         """Clear memory contents."""

chatgpt_tool_hub/engine/bot.py

@@ -95,7 +95,7 @@ def _crop_full_input(self, inputs: str) -> str:
             try:
                 os.remove(file_path)
             except Exception as e:
-                LOG.info(f"remove {file_path} failed... error_info: {repr(e)}")
+                LOG.debug(f"remove {file_path} failed... error_info: {repr(e)}")
 
         return _input
 

chatgpt_tool_hub/engine/tool_engine.py

@@ -130,8 +130,8 @@ def _take_next_step(
             self.console.print(Panel(f"{output.tool_input}",
                                      title=f"我将给工具 [bright_magenta]{output.tool}[/] 发送如下信息",
                                      highlight=True))
-            self.console.print("\n")
-
+
+            LOG.info(f"我将给工具发送如下信息: \n{output.tool_input}")
             tool = name_to_tool_map[output.tool]
             return_direct = tool.return_direct
             # color = color_mapping[output.tool]
@@ -146,8 +146,8 @@ def _take_next_step(
             )
         else:
             self.console.print(f"× 该工具 [bright_magenta]{output.tool}[/] 无效")
-            self.console.print("\n")
-
+
+            LOG.info(f"该工具 {output.tool} 无效")
             observation = InvalidTool().run(
                 output.tool,
                 verbose=self.verbose,
@@ -159,8 +159,8 @@ def _take_next_step(
         self.console.print(Panel(observation + "\n",
                                  title=f"工具 [bright_magenta]{output.tool}[/] 返回内容",
                                  highlight=True, style='dim'))
-        self.console.print("\n")
-
+
+        LOG.info(f"工具 {output.tool} 返回内容: {observation}")
         return output, observation
 
     def _call(self, inputs: Dict[str, str]) -> Dict[str, Any]:
@@ -321,3 +321,9 @@ async def _areturn(
         if self.return_intermediate_steps:
             final_output["intermediate_steps"] = intermediate_steps
         return final_output
+
+    def get_tool_list(self) -> List[str]:
+        _tool_list = []
+        for tool in self.tools:
+            _tool_list.extend(tool.get_tool_list())
+        return _tool_list

chatgpt_tool_hub/tools/arxiv_search/api_prompt.py

@@ -1,5 +1,5 @@
 ARXIV_PROMPT = """
-You are a converter from natural language to json to search research papers in arxiv api.
+You are a converter from natural language to json to search research papers by arxiv api.
 
 You should only respond in JSON format as described below.
 Response Format: 
@@ -10,7 +10,7 @@
     "sort_order": "descending"
 }}
 json note:
-search_query: an arXiv query string. I will show you how to generate a search_query
+search_query: an arXiv query string. I will teach you how to generate a search_query below.
 max_results: int, range: 1~20
 sort_by: The sort criterion for results: "relevance", "lastUpdatedDate", or "submittedDate"
 sort_order: The sort order for results: "descending" or "ascending"

chatgpt_tool_hub/tools/arxiv_search/tool.py

@@ -1,16 +1,20 @@
+import os
+import tempfile
 from typing import Any
 
 from rich.console import Console
 
 from chatgpt_tool_hub.chains import LLMChain
 from chatgpt_tool_hub.common.log import LOG
+from chatgpt_tool_hub.common.utils import get_from_dict_or_env
 from chatgpt_tool_hub.models import build_model_params
 from chatgpt_tool_hub.models.model_factory import ModelFactory
 from chatgpt_tool_hub.prompts import PromptTemplate
 from chatgpt_tool_hub.tools.all_tool_list import main_tool_register
 from chatgpt_tool_hub.tools.arxiv_search.api_prompt import ARXIV_PROMPT
 from chatgpt_tool_hub.tools.arxiv_search.wrapper import ArxivAPIWrapper
 from chatgpt_tool_hub.tools.base_tool import BaseTool
+from chatgpt_tool_hub.tools.summary import SummaryTool
 
 default_tool_name = "arxiv"
 
@@ -32,6 +36,7 @@ def __init__(self, console: Console = Console(), **tool_kwargs: Any):
         super().__init__(console=console, return_direct=True)
 
         self.api_wrapper = ArxivAPIWrapper(**tool_kwargs)
+        self.arxiv_summary = get_from_dict_or_env(tool_kwargs, "arxiv_summary", "ARXIV_SUMMARY", True)
 
         llm = ModelFactory().create_llm_model(**build_model_params(tool_kwargs))
         prompt = PromptTemplate(
@@ -46,7 +51,23 @@ def _run(self, query: str) -> str:
         _llm_response = self.bot.run(query)
         LOG.info(f"[arxiv]: search_query: {_llm_response}")
 
-        return self.api_wrapper.run(_llm_response)
+        _api_response = self.api_wrapper.run(_llm_response)
+
+        if not self.arxiv_summary:
+            return _api_response
+
+        temp_file = tempfile.mkstemp()
+        file_path = temp_file[1]
+        with open(file_path, "w") as f:
+            f.write(_api_response + "\n")
+
+        _input = SummaryTool(max_segment_length=2400).run(f"{str(file_path)}, 0")
+        try:
+            os.remove(file_path)
+        except Exception as e:
+            LOG.debug(f"remove {file_path} failed... error_info: {repr(e)}")
+
+        return _input
 
     async def _arun(self, query: str) -> str:
         """Use the Arxiv tool asynchronously."""

chatgpt_tool_hub/tools/base_tool.py

@@ -122,3 +122,6 @@ async def arun(
                 observation, verbose=verbose, color=color, **kwargs
             )
         return observation
+
+    def get_tool_list(self):
+        return [self.name]

chatgpt_tool_hub/tools/meteo/tool.py

@@ -24,7 +24,7 @@ class MeteoWeatherTool(BaseTool):
     def __init__(self, console: Console = Console(), **tool_kwargs):
         super().__init__(console=console, return_direct=False)
         llm = ModelFactory().create_llm_model(**build_model_params(tool_kwargs))
-        self.api_chain = APIChain.from_llm_and_api_docs(llm, OPEN_METEO_DOCS)
+        self.api_chain = APIChain.from_llm_and_api_docs(llm, OPEN_METEO_DOCS, console=console)
 
     def _run(self, query: str) -> str:
         """Use the tool."""

chatgpt_tool_hub/tools/news/morning_news/tool.py

@@ -57,7 +57,7 @@ def _run(self, query: str) -> str:
             _news_content = "\n".join(_return_data.get("news"))
             _weiyu = _return_data.get('weiyu')
             _image_url = _return_data.get('image')
-            return f"\n今日日期：{_date}\n今日早报：{_news_content}\n今日微语：{_weiyu}\nURL: {_image_url}"
+            return f"\n今日日期：{_date}\n\n今日早报：{_news_content}\n\n今日微语：{_weiyu}\n\nURL: {_image_url}"
         else:
             return f"[{default_tool_name}] api error, error_info: {_response_json.get('msg')}"
 

chatgpt_tool_hub/tools/news/news_api/docs_prompts.py → chatgpt_tool_hub/tools/news/news_api/docs_prompt.py

@@ -7,7 +7,6 @@
 
 This endpoint is great for retrieving headlines for use with news tickers or similar.
 Request parameters
-
     country | The 2-letter ISO 3166-1 code of the country you want to get headlines for. Possible options: ae ar at au be bg br ca ch cn co cu cz de eg fr gb gr hk hu id ie il in it jp kr lt lv ma mx my ng nl no nz ph pl pt ro rs ru sa se sg si sk th tr tw ua us ve za. Note: you can't mix this param with the sources param.
     category | The category you want to get headlines for. Possible options: business entertainment general health science sports technology. Note: you can't mix this param with the sources param.
     sources | A comma-seperated string of identifiers for the news sources or blogs you want headlines from. Use the /top-headlines/sources endpoint to locate these programmatically or look at the sources index. Note: you can't mix this param with the country or category params.
@@ -28,7 +27,11 @@
     publishedAt | string | The date and time that the article was published, in UTC (+000)
     content | string | The unformatted content of the article, where available. This is truncated to 200 chars.
 
-Use page size: 2
-The endpoint and prefix of the document are absolutely correct, and that other URL prefixes should not be fabricated.
-the url must not contain apiKey param
-"""
+Pay attention: 
+1. page size: 5
+2. You should only use the parameters described in this document 
+to construct the API URL, and should not make up parameters.
+3. The endpoint and top headlines of the document are absolutely correct, 
+any other URL prefixes should not be fabricated.
+
+URL: """

chatgpt_tool_hub/tools/news/news_api/tool.py

@@ -8,7 +8,7 @@
 from chatgpt_tool_hub.models.model_factory import ModelFactory
 from chatgpt_tool_hub.tools.base_tool import BaseTool
 from chatgpt_tool_hub.tools.news import news_tool_register
-from chatgpt_tool_hub.tools.news.news_api.docs_prompts import NEWS_DOCS
+from chatgpt_tool_hub.tools.news.news_api.docs_prompt import NEWS_DOCS
 
 default_tool_name = "news-api"
 
@@ -29,7 +29,7 @@ def __init__(self, console: Console = Console(), **tool_kwargs: Any):
         llm = ModelFactory().create_llm_model(**build_model_params(tool_kwargs))
 
         self.api_chain = APIChain.from_llm_and_api_docs(
-            llm, NEWS_DOCS, headers={"X-Api-Key": news_api_key}
+            llm, NEWS_DOCS, console=console, headers={"X-Api-Key": news_api_key}
         )
 
     def _run(self, query: str) -> str:

chatgpt_tool_hub/tools/news/tool.py

@@ -17,9 +17,9 @@
 class NewsTool(BaseTool):
     name: str = default_tool_name
     description: str = (
-        "Useful when you want to get information about current news stories, "
-        "such as financial news, daily morning reports and any other news. "
-        "The input should be a description of your needs in natural language."
+        "当你想要获取实时新闻资讯时可以使用该工具，你能获取任何与新闻有关的信息。"
+        "该工具包含了金融、早报和news-api三个子工具，访问这些工具前你需要先访问本工具。"
+        "工具输入：你目前了解到的所有信息的总结 和 用户想获取的新闻内容。"
     )
     engine: ToolEngine = Any
 
@@ -30,6 +30,7 @@ def __init__(self, console: Console = Console(), **tool_kwargs: Any):
         tools = load_tools(news_tool_register.get_registered_tool_names(), news_tool_register, console, **tool_kwargs)
         llm = ModelFactory().create_llm_model(**build_model_params(tool_kwargs))
 
+        # subtool 思考深度暂时固定为2
         self.engine = init_tool_engine(tools, llm, bot="qa-bot", verbose=True, console=self.console,
                                        max_iterations=2, early_stopping_method="force")
 
@@ -47,5 +48,8 @@ async def _arun(self, query: str) -> str:
         """Use the tool asynchronously."""
         raise NotImplementedError("NewsTool does not support async")
 
+    def get_tool_list(self):
+        return news_tool_register.get_registered_tool_names()
+
 
 main_tool_register.register_tool(default_tool_name, lambda console, kwargs: NewsTool(console, **kwargs), [])

chatgpt_tool_hub/tools/summary/tool.py

@@ -151,11 +151,6 @@ def _run(self, tool_input: str) -> str:
             map_text_list = asyncio.run(self._acall(self.map_bot, _clip_text_list))
             map_text = _clipper.seperator.join(map_text_list)
 
-            self.console.print(Panel(f"{map_text}",
-                                     title=f"=[bright_magenta]Summary tool[/] 总结",
-                                     highlight=True))
-            self.console.print("\n")
-
             LOG.debug(f"[summary] round:{ctn}, map_list: {map_text}")
             # reduce
             _clip_summary_list = _clipper.clip(map_text, self.message_num)
@@ -164,6 +159,16 @@ def _run(self, tool_input: str) -> str:
             reduce_text = _clipper.seperator.join(reduce_text_list)
             LOG.debug(f"[summary] round:{ctn}, reduce_list: {reduce_text}")
             _text = reduce_text
+
+            self.console.print(Panel(f"{_text}",
+                                     title=f"[bright_magenta]Summary tool[/] 第{ctn}轮总结",
+                                     highlight=True))
+
+        if ctn > 2:
+            self.console.print(Panel(f"{_text}",
+                                     title=f"[bright_magenta]Summary tool[/] 最终总结",
+                                     highlight=True))
+
         return _text
 
     async def _arun(self, file_path: str) -> str:

chatgpt_tool_hub/tools/web_requests/__init__.py

@@ -37,7 +37,7 @@ def filter_text(html: str) -> str:
     try:
         os.remove(file_path)
     except Exception as e:
-        LOG.info(f"remove {file_path} failed... error_info: {repr(e)}")
+        LOG.debug(f"remove {file_path} failed... error_info: {repr(e)}")
 
     return _summary.encode('utf-8').decode()