From c92799119405babd15b63624086f7783c462ee5c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 15:41:18 +0800
Subject: [PATCH 01/21] update get_choice_function_arguments.

---
 metagpt/provider/base_llm.py | 48 +++++++++++++++++++++++++++++++++++-
 1 file changed, 47 insertions(+), 1 deletion(-)

diff --git a/metagpt/provider/base_llm.py b/metagpt/provider/base_llm.py
index dbef15fa1..c482aaf35 100644
--- a/metagpt/provider/base_llm.py
+++ b/metagpt/provider/base_llm.py
@@ -6,10 +6,14 @@
 @File    : base_llm.py
 @Desc    : mashenquan, 2023/8/22. + try catch
 """
+import re
 import json
 from abc import ABC, abstractmethod
 from typing import Optional
 
+from metagpt.logs import logger
+from metagpt.utils.common import CodeParser
+
 
 class BaseLLM(ABC):
     """LLM API abstract class, requiring all inheritors to provide a series of standard capabilities"""
@@ -118,6 +122,30 @@ class BaseLLM(ABC):
         """
         return rsp.get("choices")[0]["message"]["tool_calls"][0]["function"]
 
+    def _parse_arguments(self, arguments: str) -> dict:
+        """parse arguments in openai function call"""
+        if 'langugae' not in arguments and 'code' not in arguments:
+            logger.warning(f"Not found `code`, `language`, We assume it is pure code:\n {arguments}\n. ")
+            return {'language': 'python', 'code': arguments}
+
+        # 匹配language
+        language_pattern = re.compile(r'[\"\']?language[\"\']?\s*:\s*["\']([^"\']+?)["\']', re.DOTALL)
+        language_match = language_pattern.search(arguments)
+        language_value = language_match.group(1) if language_match else None
+
+        # 匹配code
+        code_pattern = r'(["\']{3}|["])([\s\S]*?)\1'
+        try:
+            code_value = re.findall(code_pattern, arguments)[-1][-1]
+        except Exception as e:
+            logger.error(f"{e}, when re.findall({code_pattern}, {arguments})")
+            code_value = None
+
+        if code_value is None:
+            raise ValueError(f"Parse code error for {arguments}")
+        # arguments只有code的情况
+        return {'language': language_value, 'code': code_value}
+
     def get_choice_function_arguments(self, rsp: dict) -> dict:
         """Required to provide the first function arguments of choice.
 
@@ -125,7 +153,25 @@ class BaseLLM(ABC):
         :return dict: return the first function arguments of choice, for example,
             {'language': 'python', 'code': "print('Hello, World!')"}
         """
-        return json.loads(self.get_choice_function(rsp)["arguments"], strict=False)
+        try:
+            arguments: str = self.get_choice_function(rsp)["arguments"]
+            return json.loads(arguments, strict=False)
+        except json.decoder.JSONDecodeError as e:
+            logger.debug(f"Got JSONDecodeError for {arguments}, we will use RegExp to parse code, \n {e}")
+            return self._parse_arguments(arguments)
+        except KeyError as e:
+            if 'tool_calls' in e.args:
+                txt_rsp = self.get_choice_text(rsp)
+                # find code
+                code = CodeParser.parse_code(None, txt_rsp, lang='python')
+                if code != txt_rsp:
+                    return {'language': 'python', 'code': code}
+                # no code
+                return {'language': 'markdown', 'code': txt_rsp}
+            raise e
+        except Exception as e:
+            logger.error(f"Got error `{e}` for parsing\n {rsp}\n")
+            return {}
 
     def messages_to_prompt(self, messages: list[dict]):
         """[{"role": "user", "content": msg}] to user: <msg> etc."""

From bb356fbc02a666d970799798a8f7d921252ec703 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 15:49:07 +0800
Subject: [PATCH 02/21] update truncate.

---
 metagpt/actions/execute_code.py | 34 +++++++++++++++++++++++----------
 1 file changed, 24 insertions(+), 10 deletions(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index c75711e75..458dc0898 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -212,26 +212,40 @@ class ExecutePyCode(ExecuteCode, Action):
             cell_index = len(self.nb.cells) - 1
             success, error_message = await self.run_cell(self.nb.cells[-1], cell_index)
 
-            if success:
-                outputs = self.parse_outputs(self.nb.cells[-1].outputs)
-                return truncate(remove_escape_and_color_codes(outputs)), True
-            else:
-                return error_message, False
+            if not success:
+                return truncate(remove_escape_and_color_codes(error_message), is_success=success)
+
+            # code success
+            outputs = self.parse_outputs(self.nb.cells[-1].outputs)
+            return truncate(remove_escape_and_color_codes(outputs), is_success=success)
         else:
             # TODO: markdown
             raise NotImplementedError(f"Not support this code type : {language}, Only support code!")
 
 
-def truncate(result: str, keep_len: int = 2000) -> str:
-    desc = f"Truncated to show only the last {keep_len} characters\n"
+def truncate(result: str, keep_len: int = 2000, is_success: bool = True) -> str | bool:
+    desc = f"Executed code {'successfully' if is_success else 'failed, please reflect the cause of bug and then debug'}"
+    if is_success:
+        desc += f"Truncated to show only {keep_len} characters\n"
+    else:
+        desc += "Show complete information for you."
+
     if result.startswith(desc):
         result = result[len(desc) :]
 
     if len(result) > keep_len:
-        result = result[-keep_len:]
-        return desc + result
+        result = result[-keep_len:] if not is_success else result
+        if not result:
+            result = 'No output about your code. Only when importing packages it is normal case. Recap and go ahead.'
+            return result, False
 
-    return result
+        if result.strip().startswith("<coroutine object"):
+            result = "Executed code failed, you need use key word 'await' to run a async code."
+            return result, False
+
+        return desc + result[:keep_len+500], is_success
+
+    return result, is_success
 
 
 def remove_escape_and_color_codes(input_str):

From ec07513caf4ddc995d2630c0b90a5930313f65ce Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 15:51:36 +0800
Subject: [PATCH 03/21] support for markdown.

---
 metagpt/actions/execute_code.py | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index 458dc0898..1a97e49d6 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -15,7 +15,7 @@ import nbformat
 from nbclient import NotebookClient
 from nbclient.exceptions import CellTimeoutError, DeadKernelError
 from nbformat import NotebookNode
-from nbformat.v4 import new_code_cell, new_output
+from nbformat.v4 import new_code_cell, new_output, new_markdown_cell
 from rich.console import Console
 from rich.syntax import Syntax
 
@@ -91,6 +91,9 @@ class ExecutePyCode(ExecuteCode, Action):
     def add_code_cell(self, code):
         self.nb.cells.append(new_code_cell(source=code))
 
+    def add_markdown_cell(self, markdown):
+        self.nb.cells.append(new_markdown_cell(source=markdown))
+
     def _display(self, code, language: str = "python"):
         if language == "python":
             code = Syntax(code, "python", theme="paraiso-dark", line_numbers=True)
@@ -219,8 +222,9 @@ class ExecutePyCode(ExecuteCode, Action):
             outputs = self.parse_outputs(self.nb.cells[-1].outputs)
             return truncate(remove_escape_and_color_codes(outputs), is_success=success)
         else:
-            # TODO: markdown
-            raise NotImplementedError(f"Not support this code type : {language}, Only support code!")
+            # markdown
+            self.add_markdown_cell(code)
+            return code, True
 
 
 def truncate(result: str, keep_len: int = 2000, is_success: bool = True) -> str | bool:

From b69f2be165a2c76343f5f7b9b03c18321aa5d588 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 16:51:25 +0800
Subject: [PATCH 04/21] delete type.

---
 metagpt/actions/execute_code.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index 1a97e49d6..fb0ecd893 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -227,7 +227,7 @@ class ExecutePyCode(ExecuteCode, Action):
             return code, True
 
 
-def truncate(result: str, keep_len: int = 2000, is_success: bool = True) -> str | bool:
+def truncate(result: str, keep_len: int = 2000, is_success: bool = True):
     desc = f"Executed code {'successfully' if is_success else 'failed, please reflect the cause of bug and then debug'}"
     if is_success:
         desc += f"Truncated to show only {keep_len} characters\n"

From 4f93c5fad3f03fd0302e3a93760216fc9ca58ffa Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 16:59:49 +0800
Subject: [PATCH 05/21] add only_code arg for WriteCodeByGenerate.

---
 metagpt/actions/write_analysis_code.py | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/metagpt/actions/write_analysis_code.py b/metagpt/actions/write_analysis_code.py
index 04cad34a5..76d47ba28 100644
--- a/metagpt/actions/write_analysis_code.py
+++ b/metagpt/actions/write_analysis_code.py
@@ -88,8 +88,14 @@ class WriteCodeByGenerate(BaseWriteAnalysisCode):
     ) -> str:
         # context.append(Message(content=self.REUSE_CODE_INSTRUCTION, role="user"))
         prompt = self.process_msg(context, system_msg)
+        is_only_code = kwargs.pop("only_code", True)
+
         code_content = await self.llm.aask_code(prompt, **kwargs)
-        return code_content["code"]
+        if is_only_code:
+            return code_content["code"]
+        else:
+            return code_content
+
 
 
 class WriteCodeWithTools(BaseWriteAnalysisCode):

From 00f7f93234d0c19286aca3d16233367be2d5fd2a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 18:09:56 +0800
Subject: [PATCH 06/21] add scrape_web.

---
 metagpt/tools/__init__.py                     |  6 ++++++
 .../tools/functions/schemas/scrape_web.yml    | 21 +++++++++++++++++++
 2 files changed, 27 insertions(+)
 create mode 100644 metagpt/tools/functions/schemas/scrape_web.yml

diff --git a/metagpt/tools/__init__.py b/metagpt/tools/__init__.py
index 41c8708b2..c24dc6fce 100644
--- a/metagpt/tools/__init__.py
+++ b/metagpt/tools/__init__.py
@@ -76,6 +76,12 @@ TOOL_TYPE_MAPPINGS = {
         desc="Related to text2image, image2image using stable diffusion model.",
         usage_prompt="",
     ),
+    "scrape_web": ToolType(
+        name="scrape_web",
+        module="metagpt.tools.scrape_web",
+        desc="Scrape data from web page.",
+        usage_prompt="",
+    ),
     "other": ToolType(
         name="other",
         module="",
diff --git a/metagpt/tools/functions/schemas/scrape_web.yml b/metagpt/tools/functions/schemas/scrape_web.yml
new file mode 100644
index 000000000..ecca3fbed
--- /dev/null
+++ b/metagpt/tools/functions/schemas/scrape_web.yml
@@ -0,0 +1,21 @@
+scrape_web:
+  type: async funciton
+  description: "Scrape and save the HTML structure and inner text content of a web page using Playwright."
+  parameters:
+    properties:
+      url:
+        type: str
+        description: "web url"
+      \*url:
+        type: Non-Keyword Arguments
+        description: "other web urls, you can assagin sub url link to it."
+    required:
+      - url
+  returns:
+    inner_text:
+      type: str
+      description: The inner text content of the web page.
+    html:
+      type: str
+      description: The html structure of the web page.
+

From 75628caf4d68b7519c63a84c0203326ea05ace5c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 18:10:57 +0800
Subject: [PATCH 07/21] add scrape_web.py

---
 .../functions/libs/scrape_web/__init__.py     |  1 +
 .../functions/libs/scrape_web/scrape_web.py   | 26 +++++++++++++++++++
 2 files changed, 27 insertions(+)
 create mode 100644 metagpt/tools/functions/libs/scrape_web/__init__.py
 create mode 100644 metagpt/tools/functions/libs/scrape_web/scrape_web.py

diff --git a/metagpt/tools/functions/libs/scrape_web/__init__.py b/metagpt/tools/functions/libs/scrape_web/__init__.py
new file mode 100644
index 000000000..d5cd1524b
--- /dev/null
+++ b/metagpt/tools/functions/libs/scrape_web/__init__.py
@@ -0,0 +1 @@
+from metagpt.tools.functions.libs.scrape_web.scrape_web import scrape_web
diff --git a/metagpt/tools/functions/libs/scrape_web/scrape_web.py b/metagpt/tools/functions/libs/scrape_web/scrape_web.py
new file mode 100644
index 000000000..5cd984f4d
--- /dev/null
+++ b/metagpt/tools/functions/libs/scrape_web/scrape_web.py
@@ -0,0 +1,26 @@
+import asyncio
+
+from metagpt.tools.web_browser_engine_playwright import PlaywrightWrapper
+
+
+async def scrape_web(url, *urls):
+    """
+    Scrape and save the HTML structure and inner text content of a web page using Playwright.
+
+    Args:
+        url (str): The main URL to fetch inner text from.
+        *urls (str): Additional URLs to fetch inner text from.
+
+    Returns:
+        (dict): The inner text content and html structure of the web page, key are : 'inner_text', 'html'.
+
+    Raises:
+        Any exceptions that may occur during the Playwright operation.
+    """
+    # Create a PlaywrightWrapper instance for the Chromium browser
+    web = await PlaywrightWrapper("chromium").run(url, *urls)
+
+    # Return the inner text content of the web page
+    return {"inner_text": web.inner_text, "html": web.html}
+
+# 需要改三个地方: yaml, 对应路径下init, MetaGPT/metagpt/prompts/ml_engineer.py中ML_MODULE_MAP

From a7e088845e508464281daed1301ce32c0acc0797 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 18:11:22 +0800
Subject: [PATCH 08/21] update scrape_web docstring.

---
 metagpt/tools/functions/libs/scrape_web/scrape_web.py | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/metagpt/tools/functions/libs/scrape_web/scrape_web.py b/metagpt/tools/functions/libs/scrape_web/scrape_web.py
index 5cd984f4d..e68ce0e64 100644
--- a/metagpt/tools/functions/libs/scrape_web/scrape_web.py
+++ b/metagpt/tools/functions/libs/scrape_web/scrape_web.py
@@ -13,9 +13,6 @@ async def scrape_web(url, *urls):
 
     Returns:
         (dict): The inner text content and html structure of the web page, key are : 'inner_text', 'html'.
-
-    Raises:
-        Any exceptions that may occur during the Playwright operation.
     """
     # Create a PlaywrightWrapper instance for the Chromium browser
     web = await PlaywrightWrapper("chromium").run(url, *urls)

From 559a1604ad1e273d9de50fd466b1d0ac2a045d56 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 18:26:32 +0800
Subject: [PATCH 09/21] restore.

---
 metagpt/provider/base_llm.py | 48 +-----------------------------------
 1 file changed, 1 insertion(+), 47 deletions(-)

diff --git a/metagpt/provider/base_llm.py b/metagpt/provider/base_llm.py
index c482aaf35..dbef15fa1 100644
--- a/metagpt/provider/base_llm.py
+++ b/metagpt/provider/base_llm.py
@@ -6,14 +6,10 @@
 @File    : base_llm.py
 @Desc    : mashenquan, 2023/8/22. + try catch
 """
-import re
 import json
 from abc import ABC, abstractmethod
 from typing import Optional
 
-from metagpt.logs import logger
-from metagpt.utils.common import CodeParser
-
 
 class BaseLLM(ABC):
     """LLM API abstract class, requiring all inheritors to provide a series of standard capabilities"""
@@ -122,30 +118,6 @@ class BaseLLM(ABC):
         """
         return rsp.get("choices")[0]["message"]["tool_calls"][0]["function"]
 
-    def _parse_arguments(self, arguments: str) -> dict:
-        """parse arguments in openai function call"""
-        if 'langugae' not in arguments and 'code' not in arguments:
-            logger.warning(f"Not found `code`, `language`, We assume it is pure code:\n {arguments}\n. ")
-            return {'language': 'python', 'code': arguments}
-
-        # 匹配language
-        language_pattern = re.compile(r'[\"\']?language[\"\']?\s*:\s*["\']([^"\']+?)["\']', re.DOTALL)
-        language_match = language_pattern.search(arguments)
-        language_value = language_match.group(1) if language_match else None
-
-        # 匹配code
-        code_pattern = r'(["\']{3}|["])([\s\S]*?)\1'
-        try:
-            code_value = re.findall(code_pattern, arguments)[-1][-1]
-        except Exception as e:
-            logger.error(f"{e}, when re.findall({code_pattern}, {arguments})")
-            code_value = None
-
-        if code_value is None:
-            raise ValueError(f"Parse code error for {arguments}")
-        # arguments只有code的情况
-        return {'language': language_value, 'code': code_value}
-
     def get_choice_function_arguments(self, rsp: dict) -> dict:
         """Required to provide the first function arguments of choice.
 
@@ -153,25 +125,7 @@ class BaseLLM(ABC):
         :return dict: return the first function arguments of choice, for example,
             {'language': 'python', 'code': "print('Hello, World!')"}
         """
-        try:
-            arguments: str = self.get_choice_function(rsp)["arguments"]
-            return json.loads(arguments, strict=False)
-        except json.decoder.JSONDecodeError as e:
-            logger.debug(f"Got JSONDecodeError for {arguments}, we will use RegExp to parse code, \n {e}")
-            return self._parse_arguments(arguments)
-        except KeyError as e:
-            if 'tool_calls' in e.args:
-                txt_rsp = self.get_choice_text(rsp)
-                # find code
-                code = CodeParser.parse_code(None, txt_rsp, lang='python')
-                if code != txt_rsp:
-                    return {'language': 'python', 'code': code}
-                # no code
-                return {'language': 'markdown', 'code': txt_rsp}
-            raise e
-        except Exception as e:
-            logger.error(f"Got error `{e}` for parsing\n {rsp}\n")
-            return {}
+        return json.loads(self.get_choice_function(rsp)["arguments"], strict=False)
 
     def messages_to_prompt(self, messages: list[dict]):
         """[{"role": "user", "content": msg}] to user: <msg> etc."""

From b430e2c88fe6db7104faf9dc44cad639f95965c9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 19:02:37 +0800
Subject: [PATCH 10/21] update scrape_web module.

---
 metagpt/tools/__init__.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/metagpt/tools/__init__.py b/metagpt/tools/__init__.py
index c24dc6fce..2f8941fdb 100644
--- a/metagpt/tools/__init__.py
+++ b/metagpt/tools/__init__.py
@@ -78,7 +78,7 @@ TOOL_TYPE_MAPPINGS = {
     ),
     "scrape_web": ToolType(
         name="scrape_web",
-        module="metagpt.tools.scrape_web",
+        module=str(TOOL_LIBS_PATH / "scrape_web"),
         desc="Scrape data from web page.",
         usage_prompt="",
     ),

From d1666c3307289edd0fcb53c8ba881574ee0dca19 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 21:17:01 +0800
Subject: [PATCH 11/21] update get_choice_function_arguments.

---
 metagpt/provider/openai_api.py | 71 +++++++++++++++++++++++-----------
 1 file changed, 48 insertions(+), 23 deletions(-)

diff --git a/metagpt/provider/openai_api.py b/metagpt/provider/openai_api.py
index 747e36480..66d215eda 100644
--- a/metagpt/provider/openai_api.py
+++ b/metagpt/provider/openai_api.py
@@ -9,6 +9,7 @@
 @Modified By: mashenquan, 2023/12/1. Fix bug: Unclosed connection caused by openai 0.x.
 """
 
+import re
 import json
 from typing import AsyncIterator, Union
 
@@ -37,6 +38,7 @@ from metagpt.utils.token_counter import (
     count_string_tokens,
     get_max_completion_tokens,
 )
+from metagpt.utils.common import CodeParser
 
 
 def log_and_reraise(retry_state):
@@ -147,10 +149,7 @@ class OpenAILLM(BaseLLM):
     def _func_configs(self, messages: list[dict], timeout=3, **kwargs) -> dict:
         """Note: Keep kwargs consistent with https://platform.openai.com/docs/api-reference/chat/create"""
         if "tools" not in kwargs:
-            configs = {
-                "tools": [{"type": "function", "function": GENERAL_FUNCTION_SCHEMA}],
-                "tool_choice": GENERAL_TOOL_CHOICE,
-            }
+            configs = {"tools": [{"type": "function", "function": GENERAL_FUNCTION_SCHEMA}]}
             kwargs.update(configs)
 
         return self._cons_kwargs(messages=messages, timeout=timeout, **kwargs)
@@ -161,23 +160,7 @@ class OpenAILLM(BaseLLM):
         self._update_costs(rsp.usage)
         return rsp
 
-    def _process_message(self, messages: Union[str, Message, list[dict], list[Message], list[str]]) -> list[dict]:
-        """convert messages to list[dict]."""
-        if isinstance(messages, list):
-            messages = [Message(content=msg) if isinstance(msg, str) else msg for msg in messages]
-            return [msg if isinstance(msg, dict) else msg.to_dict() for msg in messages]
-
-        if isinstance(messages, Message):
-            messages = [messages.to_dict()]
-        elif isinstance(messages, str):
-            messages = [{"role": "user", "content": messages}]
-        else:
-            raise ValueError(
-                f"Only support messages type are: str, Message, list[dict], but got {type(messages).__name__}!"
-            )
-        return messages
-
-    async def aask_code(self, messages: Union[str, Message, list[dict]], **kwargs) -> dict:
+    async def aask_code(self, messages: list[dict], **kwargs) -> dict:
         """Use function of tools to ask a code.
         Note: Keep kwargs consistent with https://platform.openai.com/docs/api-reference/chat/create
 
@@ -187,18 +170,60 @@ class OpenAILLM(BaseLLM):
         >>> rsp = await llm.aask_code(msg)
         # -> {'language': 'python', 'code': "print('Hello, World!')"}
         """
-        messages = self._process_message(messages)
         rsp = await self._achat_completion_function(messages, **kwargs)
         return self.get_choice_function_arguments(rsp)
 
+    def _parse_arguments(self, arguments: str) -> dict:
+        """parse arguments in openai function call"""
+        if 'langugae' not in arguments and 'code' not in arguments:
+            logger.warning(f"Not found `code`, `language`, We assume it is pure code:\n {arguments}\n. ")
+            return {'language': 'python', 'code': arguments}
+
+        # 匹配language
+        language_pattern = re.compile(r'[\"\']?language[\"\']?\s*:\s*["\']([^"\']+?)["\']', re.DOTALL)
+        language_match = language_pattern.search(arguments)
+        language_value = language_match.group(1) if language_match else None
+
+        # 匹配code
+        code_pattern = r'(["\'`]{3}|["\'`])([\s\S]*?)\1'
+        try:
+            code_value = re.findall(code_pattern, arguments)[-1][-1]
+        except Exception as e:
+            logger.error(f"{e}, when re.findall({code_pattern}, {arguments})")
+            code_value = None
+
+        if code_value is None:
+            raise ValueError(f"Parse code error for {arguments}")
+        # arguments只有code的情况
+        return {'language': language_value, 'code': code_value}
+
     @handle_exception
     def get_choice_function_arguments(self, rsp: ChatCompletion) -> dict:
         """Required to provide the first function arguments of choice.
 
+        :param dict rsp: same as in self.get_choice_function(rsp)
         :return dict: return the first function arguments of choice, for example,
             {'language': 'python', 'code': "print('Hello, World!')"}
         """
-        return json.loads(rsp.choices[0].message.tool_calls[0].function.arguments)
+        message = rsp.choices[0].message
+        if (
+            message.tool_calls is not None and
+            message.tool_calls[0].function is not None and
+            message.tool_calls[0].function.arguments is not None
+        ):
+            # reponse is code
+            try:
+                return json.loads(message.tool_calls[0].function.arguments, strict=False)
+            except json.decoder.JSONDecodeError as e:
+                logger.debug(f"Got JSONDecodeError for {message.tool_calls[0].function.arguments},\
+                    we will use RegExp to parse code, \n {e}")
+                return {'language': 'python', 'code': self._parse_arguments(message.tool_calls[0].function.arguments)}
+        elif message.tool_calls is None and message.content is not None:
+            # reponse is message
+            return {'language': 'markdown', 'code': self.get_choice_text(rsp)}
+        else:
+            logger.error(f"Failed to parse \n {rsp}\n")
+            raise Exception(f"Failed to parse \n {rsp}\n")
 
     def get_choice_text(self, rsp: ChatCompletion) -> str:
         """Required to provide the first text of choice"""

From f9b1cce654e36ac764acd7db0b7d5c74404dc877 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Mon, 15 Jan 2024 22:21:56 +0800
Subject: [PATCH 12/21] update code-intepreter by auto aask.

---
 metagpt/actions/write_analysis_code.py |  2 +-
 metagpt/provider/openai_api.py         | 23 +++++++++++++++++++++++
 metagpt/roles/code_interpreter.py      |  9 ++++++---
 3 files changed, 30 insertions(+), 4 deletions(-)

diff --git a/metagpt/actions/write_analysis_code.py b/metagpt/actions/write_analysis_code.py
index 76d47ba28..bceb100b1 100644
--- a/metagpt/actions/write_analysis_code.py
+++ b/metagpt/actions/write_analysis_code.py
@@ -88,7 +88,7 @@ class WriteCodeByGenerate(BaseWriteAnalysisCode):
     ) -> str:
         # context.append(Message(content=self.REUSE_CODE_INSTRUCTION, role="user"))
         prompt = self.process_msg(context, system_msg)
-        is_only_code = kwargs.pop("only_code", True)
+        is_only_code = kwargs.pop("only_code", False)
 
         code_content = await self.llm.aask_code(prompt, **kwargs)
         if is_only_code:
diff --git a/metagpt/provider/openai_api.py b/metagpt/provider/openai_api.py
index 66d215eda..7bdb4bfbe 100644
--- a/metagpt/provider/openai_api.py
+++ b/metagpt/provider/openai_api.py
@@ -154,7 +154,30 @@ class OpenAILLM(BaseLLM):
 
         return self._cons_kwargs(messages=messages, timeout=timeout, **kwargs)
 
+    def _process_message(self, messages: Union[str, Message, list[dict], list[Message], list[str]]) -> list[dict]:
+        """convert messages to list[dict]."""
+        # 全部转成list
+        if not isinstance(messages, list):
+            messages = [messages]
+
+        # 转成list[dict]
+        processed_messages = []
+        for msg in messages:
+            if isinstance(msg, str):
+                processed_messages.append({"role": "user", "content": msg})
+            elif isinstance(msg, dict):
+                assert set(msg.keys()) == set(['role', 'content'])
+                processed_messages.append(msg)
+            elif isinstance(msg, Message):
+                processed_messages.append(msg.to_dict())
+            else:
+                raise ValueError(
+                    f"Only support message type are: str, Message, dict, but got {type(messages).__name__}!"
+                )
+        return processed_messages
+
     async def _achat_completion_function(self, messages: list[dict], timeout=3, **chat_configs) -> ChatCompletion:
+        messages = self._process_message(messages)
         kwargs = self._func_configs(messages=messages, timeout=timeout, **chat_configs)
         rsp: ChatCompletion = await self.aclient.chat.completions.create(**kwargs)
         self._update_costs(rsp.usage)
diff --git a/metagpt/roles/code_interpreter.py b/metagpt/roles/code_interpreter.py
index 164c7cb12..afd51a575 100644
--- a/metagpt/roles/code_interpreter.py
+++ b/metagpt/roles/code_interpreter.py
@@ -52,7 +52,7 @@ class CodeInterpreter(Role):
 
     async def _act_on_task(self, current_task: Task) -> TaskResult:
         code, result, is_success = await self._write_and_exec_code()
-        task_result = TaskResult(code=code, result=result, is_success=is_success)
+        task_result = TaskResult(code=code['code'], result=result, is_success=is_success)
         return task_result
 
     async def _write_and_exec_code(self, max_retry: int = 3):
@@ -63,10 +63,10 @@ class CodeInterpreter(Role):
             ### write code ###
             code, cause_by = await self._write_code()
 
-            self.working_memory.add(Message(content=code, role="assistant", cause_by=cause_by))
+            self.working_memory.add(Message(content=code['code'], role="assistant", cause_by=cause_by))
 
             ### execute code ###
-            result, success = await self.execute_code.run(code)
+            result, success = await self.execute_code.run(**code)
             print(result)
 
             self.working_memory.add(Message(content=result, role="user", cause_by=ExecutePyCode))
@@ -91,6 +91,9 @@ class CodeInterpreter(Role):
 
         context = self.planner.get_useful_memories()
         code = await todo.run(context=context, plan=self.planner.plan, temperature=0.0)
+        # 暂时在这里转换 WriteCodeWithTools 的输出
+        if isinstance(code, str):
+            code = {'code': code, 'language': 'python'}
 
         return code, todo
 

From 29fd7117ef5cf187506e53727437881068118113 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Tue, 16 Jan 2024 11:57:08 +0800
Subject: [PATCH 13/21] update module.

---
 metagpt/tools/__init__.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/metagpt/tools/__init__.py b/metagpt/tools/__init__.py
index 2f8941fdb..73de03156 100644
--- a/metagpt/tools/__init__.py
+++ b/metagpt/tools/__init__.py
@@ -78,7 +78,7 @@ TOOL_TYPE_MAPPINGS = {
     ),
     "scrape_web": ToolType(
         name="scrape_web",
-        module=str(TOOL_LIBS_PATH / "scrape_web"),
+        module="metagpt.tools.functions.libs.scrape_web.scrape_web",
         desc="Scrape data from web page.",
         usage_prompt="",
     ),

From eef77d1628bf48657ecf307ba69ab21f21e2d71e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Tue, 16 Jan 2024 12:29:52 +0800
Subject: [PATCH 14/21] display markdown.

---
 metagpt/actions/execute_code.py | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index fb0ecd893..9fadd0acd 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -18,6 +18,8 @@ from nbformat import NotebookNode
 from nbformat.v4 import new_code_cell, new_output, new_markdown_cell
 from rich.console import Console
 from rich.syntax import Syntax
+from rich.markdown import Markdown
+
 
 from metagpt.actions import Action
 from metagpt.logs import logger
@@ -97,8 +99,12 @@ class ExecutePyCode(ExecuteCode, Action):
     def _display(self, code, language: str = "python"):
         if language == "python":
             code = Syntax(code, "python", theme="paraiso-dark", line_numbers=True)
-            self.console.print("\n")
             self.console.print(code)
+        elif language == "markdown":
+            code = Markdown(code, inline_code_theme="paraiso-dark")
+            self.console.print(code)
+        else:
+            raise ValueError(f"Only support for python, markdown, but got {language}")
 
     def add_output_to_cell(self, cell, output):
         if "outputs" not in cell:
@@ -221,10 +227,12 @@ class ExecutePyCode(ExecuteCode, Action):
             # code success
             outputs = self.parse_outputs(self.nb.cells[-1].outputs)
             return truncate(remove_escape_and_color_codes(outputs), is_success=success)
-        else:
+        elif language == 'markdown':
             # markdown
             self.add_markdown_cell(code)
             return code, True
+        else:
+            raise ValueError(f"Only support for language: python, markdown, but got {language}, ")
 
 
 def truncate(result: str, keep_len: int = 2000, is_success: bool = True):

From 95ce190f32a88d59455bb8bb982b64dd3a5018c0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Tue, 16 Jan 2024 14:30:07 +0800
Subject: [PATCH 15/21] feature: display markdown content.

---
 metagpt/actions/execute_code.py | 36 ++++++++++++++++++++++++++++++---
 1 file changed, 33 insertions(+), 3 deletions(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index 9fadd0acd..6d9135ec3 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -19,7 +19,10 @@ from nbformat.v4 import new_code_cell, new_output, new_markdown_cell
 from rich.console import Console
 from rich.syntax import Syntax
 from rich.markdown import Markdown
-
+from rich.panel import Panel
+from rich.box import MINIMAL
+from rich.live import Live
+from rich.console import Group
 
 from metagpt.actions import Action
 from metagpt.logs import logger
@@ -101,8 +104,7 @@ class ExecutePyCode(ExecuteCode, Action):
             code = Syntax(code, "python", theme="paraiso-dark", line_numbers=True)
             self.console.print(code)
         elif language == "markdown":
-            code = Markdown(code, inline_code_theme="paraiso-dark")
-            self.console.print(code)
+            _display_markdown(code)
         else:
             raise ValueError(f"Only support for python, markdown, but got {language}")
 
@@ -265,3 +267,31 @@ def remove_escape_and_color_codes(input_str):
     pattern = re.compile(r"\x1b\[[0-9;]*[mK]")
     result = pattern.sub("", input_str)
     return result
+
+
+def _display_markdown(content: str):
+    # 使用正则表达式逐个匹配代码块
+    matches = re.finditer(r'```(.+?)```', content, re.DOTALL)
+    start_index = 0
+    content_panels = []
+    # 逐个打印匹配到的文本和代码
+    for match in matches:
+        text_content = content[start_index:match.start()].strip()
+        code_content = match.group(0).strip()[3:-3]           # Remove triple backticks
+
+        if text_content:
+            content_panels.append(Panel(Markdown(text_content), box=MINIMAL))
+
+        if code_content:
+            content_panels.append(Panel(Markdown(f"```{code_content}"), box=MINIMAL))
+        start_index = match.end()
+
+    # 打印剩余文本（如果有）
+    remaining_text = content[start_index:].strip()
+    if remaining_text:
+        content_panels.append(Panel(Markdown(remaining_text), box=MINIMAL))
+
+    # 在Live模式中显示所有Panel
+    with Live(auto_refresh=False, console=Console(), vertical_overflow="visible") as live:
+        live.update(Group(*content_panels))
+        live.refresh()

From 43558c208ebd1dbf6bf980f0a271abaed4557a7d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Tue, 16 Jan 2024 15:03:12 +0800
Subject: [PATCH 16/21] doc: add ipywidgets.

---
 requirements.txt | 1 +
 1 file changed, 1 insertion(+)

diff --git a/requirements.txt b/requirements.txt
index 7ef6d884e..016c2f5d5 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -65,3 +65,4 @@ networkx~=3.2.1
 google-generativeai==0.3.2
 # playwright==1.40.0  # playwright extras require
 anytree
+ipywidgets==8.1.1

From ff10c9bdda72cce908da673c29a5980105129797 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Wed, 17 Jan 2024 18:10:30 +0800
Subject: [PATCH 17/21] change name: _display_markdown -> display_markdown.

---
 metagpt/actions/execute_code.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/metagpt/actions/execute_code.py b/metagpt/actions/execute_code.py
index 6d9135ec3..5b6cba57d 100644
--- a/metagpt/actions/execute_code.py
+++ b/metagpt/actions/execute_code.py
@@ -104,7 +104,7 @@ class ExecutePyCode(ExecuteCode, Action):
             code = Syntax(code, "python", theme="paraiso-dark", line_numbers=True)
             self.console.print(code)
         elif language == "markdown":
-            _display_markdown(code)
+            display_markdown(code)
         else:
             raise ValueError(f"Only support for python, markdown, but got {language}")
 
@@ -269,7 +269,7 @@ def remove_escape_and_color_codes(input_str):
     return result
 
 
-def _display_markdown(content: str):
+def display_markdown(content: str):
     # 使用正则表达式逐个匹配代码块
     matches = re.finditer(r'```(.+?)```', content, re.DOTALL)
     start_index = 0

From 20f31fa027b32181cbcddce8fc7b24cdb2bb0a0f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Wed, 17 Jan 2024 18:17:52 +0800
Subject: [PATCH 18/21] pre-commit.

---
 metagpt/provider/openai_api.py | 29 +++++++++++++++--------------
 1 file changed, 15 insertions(+), 14 deletions(-)

diff --git a/metagpt/provider/openai_api.py b/metagpt/provider/openai_api.py
index 7bdb4bfbe..3edd89835 100644
--- a/metagpt/provider/openai_api.py
+++ b/metagpt/provider/openai_api.py
@@ -9,8 +9,8 @@
 @Modified By: mashenquan, 2023/12/1. Fix bug: Unclosed connection caused by openai 0.x.
 """
 
-import re
 import json
+import re
 from typing import AsyncIterator, Union
 
 from openai import APIConnectionError, AsyncOpenAI, AsyncStream
@@ -28,7 +28,7 @@ from tenacity import (
 from metagpt.config import CONFIG, Config, LLMProviderEnum
 from metagpt.logs import log_llm_stream, logger
 from metagpt.provider.base_llm import BaseLLM
-from metagpt.provider.constant import GENERAL_FUNCTION_SCHEMA, GENERAL_TOOL_CHOICE
+from metagpt.provider.constant import GENERAL_FUNCTION_SCHEMA
 from metagpt.provider.llm_provider_registry import register_provider
 from metagpt.schema import Message
 from metagpt.utils.cost_manager import Costs
@@ -38,7 +38,6 @@ from metagpt.utils.token_counter import (
     count_string_tokens,
     get_max_completion_tokens,
 )
-from metagpt.utils.common import CodeParser
 
 
 def log_and_reraise(retry_state):
@@ -166,7 +165,7 @@ class OpenAILLM(BaseLLM):
             if isinstance(msg, str):
                 processed_messages.append({"role": "user", "content": msg})
             elif isinstance(msg, dict):
-                assert set(msg.keys()) == set(['role', 'content'])
+                assert set(msg.keys()) == set(["role", "content"])
                 processed_messages.append(msg)
             elif isinstance(msg, Message):
                 processed_messages.append(msg.to_dict())
@@ -198,9 +197,9 @@ class OpenAILLM(BaseLLM):
 
     def _parse_arguments(self, arguments: str) -> dict:
         """parse arguments in openai function call"""
-        if 'langugae' not in arguments and 'code' not in arguments:
+        if "langugae" not in arguments and "code" not in arguments:
             logger.warning(f"Not found `code`, `language`, We assume it is pure code:\n {arguments}\n. ")
-            return {'language': 'python', 'code': arguments}
+            return {"language": "python", "code": arguments}
 
         # 匹配language
         language_pattern = re.compile(r'[\"\']?language[\"\']?\s*:\s*["\']([^"\']+?)["\']', re.DOTALL)
@@ -218,7 +217,7 @@ class OpenAILLM(BaseLLM):
         if code_value is None:
             raise ValueError(f"Parse code error for {arguments}")
         # arguments只有code的情况
-        return {'language': language_value, 'code': code_value}
+        return {"language": language_value, "code": code_value}
 
     @handle_exception
     def get_choice_function_arguments(self, rsp: ChatCompletion) -> dict:
@@ -230,20 +229,22 @@ class OpenAILLM(BaseLLM):
         """
         message = rsp.choices[0].message
         if (
-            message.tool_calls is not None and
-            message.tool_calls[0].function is not None and
-            message.tool_calls[0].function.arguments is not None
+            message.tool_calls is not None
+            and message.tool_calls[0].function is not None
+            and message.tool_calls[0].function.arguments is not None
         ):
             # reponse is code
             try:
                 return json.loads(message.tool_calls[0].function.arguments, strict=False)
             except json.decoder.JSONDecodeError as e:
-                logger.debug(f"Got JSONDecodeError for {message.tool_calls[0].function.arguments},\
-                    we will use RegExp to parse code, \n {e}")
-                return {'language': 'python', 'code': self._parse_arguments(message.tool_calls[0].function.arguments)}
+                logger.debug(
+                    f"Got JSONDecodeError for {message.tool_calls[0].function.arguments},\
+                    we will use RegExp to parse code, \n {e}"
+                )
+                return {"language": "python", "code": self._parse_arguments(message.tool_calls[0].function.arguments)}
         elif message.tool_calls is None and message.content is not None:
             # reponse is message
-            return {'language': 'markdown', 'code': self.get_choice_text(rsp)}
+            return {"language": "markdown", "code": self.get_choice_text(rsp)}
         else:
             logger.error(f"Failed to parse \n {rsp}\n")
             raise Exception(f"Failed to parse \n {rsp}\n")

From 10129c6ecf431b9f40b1dd2a349eb2cc0de5c024 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Thu, 18 Jan 2024 12:07:31 +0800
Subject: [PATCH 19/21] update scrape_web.

---
 metagpt/tools/__init__.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/metagpt/tools/__init__.py b/metagpt/tools/__init__.py
index 95872940f..222edf312 100644
--- a/metagpt/tools/__init__.py
+++ b/metagpt/tools/__init__.py
@@ -16,7 +16,7 @@ from metagpt.prompts.tool_type import (
     FEATURE_ENGINEERING_PROMPT,
     MODEL_TRAIN_PROMPT,
     MODEL_EVALUATE_PROMPT,
-    VISION_PROMPT
+    VISION_PROMPT,
 )
 
 
@@ -81,7 +81,8 @@ TOOL_TYPE_MAPPINGS = {
         name="scrape_web",
         module="metagpt.tools.functions.libs.scrape_web.scrape_web",
         desc="Scrape data from web page.",
-        usage_prompt=""）,
+        usage_prompt="",
+    ),
     "vision": ToolType(
         name="vision",
         module=str(TOOL_LIBS_PATH / "vision"),

From f3612f8123941abfa5e8aae221ba5c1ab69512a0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Thu, 18 Jan 2024 12:10:37 +0800
Subject: [PATCH 20/21] add only_code arg for WriteCodeByGenerate.

---
 metagpt/roles/ml_engineer_simple.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/metagpt/roles/ml_engineer_simple.py b/metagpt/roles/ml_engineer_simple.py
index 3f10af8d0..9ff1c9880 100644
--- a/metagpt/roles/ml_engineer_simple.py
+++ b/metagpt/roles/ml_engineer_simple.py
@@ -75,7 +75,7 @@ class MLEngineerSimple(Role):
             context = self.get_useful_memories()
             print(f"memories数量：{len(context)}")
             # print("===\n" +str(context) + "\n===")
-            code = await WriteCodeByGenerate().run(context=context, temperature=0.0)
+            code = await WriteCodeByGenerate().run(context=context, temperature=0.0, only_code=True)
             cause_by = WriteCodeByGenerate
             self.working_memory.add(Message(content=code, role="assistant", cause_by=cause_by))
 

From d78db8994c6cb6c05f40c30891987358dcafd242 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=88=98=E6=A3=92=E6=A3=92?= <liubangbang@fuzhi.ai>
Date: Thu, 18 Jan 2024 20:57:43 +0800
Subject: [PATCH 21/21] delete arg only_code.

---
 metagpt/actions/write_analysis_code.py | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)

diff --git a/metagpt/actions/write_analysis_code.py b/metagpt/actions/write_analysis_code.py
index bceb100b1..9104fdf82 100644
--- a/metagpt/actions/write_analysis_code.py
+++ b/metagpt/actions/write_analysis_code.py
@@ -85,17 +85,11 @@ class WriteCodeByGenerate(BaseWriteAnalysisCode):
         plan: Plan = None,
         system_msg: str = None,
         **kwargs,
-    ) -> str:
+    ) -> dict:
         # context.append(Message(content=self.REUSE_CODE_INSTRUCTION, role="user"))
         prompt = self.process_msg(context, system_msg)
-        is_only_code = kwargs.pop("only_code", False)
-
         code_content = await self.llm.aask_code(prompt, **kwargs)
-        if is_only_code:
-            return code_content["code"]
-        else:
-            return code_content
-
+        return code_content
 
 
 class WriteCodeWithTools(BaseWriteAnalysisCode):