agentica 0.2.1__tar.gz → 0.2.2__tar.gz

agentica-0.2.2/MANIFEST.in

@@ -0,0 +1 @@
+exclude tests/*

{agentica-0.2.1 → agentica-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: agentica
-Version: 0.2.1
+Version: 0.2.2
 Summary: LLM agents
 Home-page: https://github.com/shibing624/agentica
 Author: XuMing
@@ -53,36 +53,50 @@ Requires-Dist: rich
 
 **Agentica**: 快速打造智能、能反思、协作的AI Agent。
 
-## Overview
+## 📖 Introduction
 
-#### LLM Agent
+**Agentica** can build AI agent, which have the components of planning, memory, and tool use. 
+
+**Agentica** 可以构建AI Agent，包括规划、记忆和工具使用、执行等组件。
+
+#### Agent Components
 <img src="https://github.com/shibing624/agentica/blob/main/docs/llm_agentv2.png" width="800" />
 
 - **规划（Planning）**：任务拆解、生成计划、反思
 - **记忆（Memory）**：短期记忆（prompt实现）、长期记忆（RAG实现）
 - **工具使用（Tool use）**：function call能力，调用外部API，以获取外部信息，包括当前日期、日历、代码执行能力、对专用信息源的访问等
 
-#### Agentica Architecture  
+#### Agentica Workflow
+
+**Agentica** can also build multi-agent systems and workflows.
+
+**Agentica** 还可以构建多Agent系统和工作流。
+
 <img src="https://github.com/shibing624/agentica/blob/main/docs/agent_arch.png" width="800" />
 
 - **Planner**：负责让LLM生成一个多步计划来完成复杂任务，生成相互依赖的“链式计划”，定义每一步所依赖的上一步的输出
 - **Worker**：接受“链式计划”，循环遍历计划中的每个子任务，并调用工具完成任务，可以自动反思纠错以完成任务
 - **Solver**：求解器将所有这些输出整合为最终答案
 
+## 🔥 News
+[2024/12/25] v0.2.0版本: 支持了多模态模型，输入可以是文本、图片、音频、视频，升级Assistant为Agent，Workflow支持拆解并实现复杂任务，详见[Release-v0.2.0](https://github.com/shibing624/agentica/releases/tag/0.2.0)
+
+[2024/07/02] v0.1.0版本：实现了基于LLM的Assistant，可以快速用function call搭建大语言模型助手，详见[Release-v0.1.0](https://github.com/shibing624/agentica/releases/tag/0.1.0)
+
 
-## Features
+## 😊 Features
 `Agentica`是一个用于构建Agent的工具，具有以下功能：
 
-- **Agent编排**：通过简单代码快速编排Agent，支持 Reflection(反思）、Plan and Solve(计划并执行)、RAG、Agent、Multi-Agent、Multi-Role、Workflow等功能
+- **Agent编排**：通过简单代码快速编排Agent，支持 Reflection(反思）、Plan and Solve(计划并执行)、RAG、Agent、Multi-Agent、Team、Workflow等功能
 - **自定义prompt**：Agent支持自定义prompt和多种工具调用（tool_calls）
 - **LLM集成**：支持OpenAI、Azure、Deepseek、Moonshot、Claude、Ollama、Together等多方大模型厂商的API
 - **记忆功能**：包括短期记忆和长期记忆功能
-- **Multi-Agent协作**：支持多Agent和多角色的团队协作。
+- **Multi-Agent协作**：支持多Agent和任务委托（Team）的团队协作。
 - **Workflow工作流**：拆解复杂任务为多个Agent，基于工作流自动化串行逐步完成任务，如投资研究、新闻文章撰写和技术教程创建
 - **自我进化Agent**：具有反思和增强记忆能力的自我进化Agent
 - **Web UI**：兼容ChatPilot，可以基于Web页面交互，支持主流的open-webui、streamlit、gradio等前端交互框架
 
-## Installation
+## 💾 Install
 
 ```bash
 pip install -U agentica
@@ -96,7 +110,7 @@ cd agentica
 pip install .
 ```
 
-## Getting Started
+## 🚀 Getting Started
 
 #### Run the example
 ```shell
@@ -127,7 +141,7 @@ print(r)
 ```
 
 
-## Web UI
+## ▶️ Web UI
 
 [shibing624/ChatPilot](https://github.com/shibing624/ChatPilot) 兼容`agentica`，可以通过Web UI进行交互。
 
@@ -146,7 +160,7 @@ bash start.sh
 ```
 
 
-## Examples
+## 😀 Examples
 
 
 | 示例                                                                                                                                                    | 描述                                                                                                                                |
@@ -230,7 +244,7 @@ streamlit run 34_llm_os_demo.py
 
 <img alt="LLM OS" src="https://github.com/shibing624/agentica/blob/main/docs/llm_os_snap.png" width="800" />
 
-## Contact
+## ☎️ Contact
 
 - Issue(建议)
   ：[![GitHub issues](https://img.shields.io/github/issues/shibing624/agentica.svg)](https://github.com/shibing624/agentica/issues)
@@ -239,7 +253,7 @@ streamlit run 34_llm_os_demo.py
 
 <img src="https://github.com/shibing624/agentica/blob/main/docs/wechat.jpeg" width="200" />
 
-## Citation
+## 😇 Citation
 
 如果你在研究中使用了`agentica`，请按如下格式引用：
 
@@ -260,10 +274,10 @@ BibTeX:
 }
 ```
 
-## License
+## ⚠️ License
 
 授权协议为 [The Apache License 2.0](/LICENSE)，可免费用做商业用途。请在产品说明中附加`agentica`的链接和授权协议。
-## Contribute
+## 😍 Contribute
 
 项目代码还很粗糙，如果大家对代码有所改进，欢迎提交回本项目，在提交之前，注意以下两点：
 
@@ -272,11 +286,11 @@ BibTeX:
 
 之后即可提交PR。
 
-## Acknowledgements 
+## 💕 Acknowledgements
 
-- [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain)
-- [https://github.com/simonmesmith/agentflow](https://github.com/simonmesmith/agentflow)
-- [https://github.com/phidatahq/phidata](https://github.com/phidatahq/phidata)
+- [langchain-ai/langchain](https://github.com/langchain-ai/langchain)
+- [simonmesmith/agentflow](https://github.com/simonmesmith/agentflow)
+- [phidatahq/phidata](https://github.com/phidatahq/phidata)
 
 
 Thanks for their great work!

{agentica-0.2.1 → agentica-0.2.2}/README.md

@@ -22,36 +22,50 @@
 
 **Agentica**: 快速打造智能、能反思、协作的AI Agent。
 
-## Overview
+## 📖 Introduction
 
-#### LLM Agent
+**Agentica** can build AI agent, which have the components of planning, memory, and tool use. 
+
+**Agentica** 可以构建AI Agent，包括规划、记忆和工具使用、执行等组件。
+
+#### Agent Components
 <img src="https://github.com/shibing624/agentica/blob/main/docs/llm_agentv2.png" width="800" />
 
 - **规划（Planning）**：任务拆解、生成计划、反思
 - **记忆（Memory）**：短期记忆（prompt实现）、长期记忆（RAG实现）
 - **工具使用（Tool use）**：function call能力，调用外部API，以获取外部信息，包括当前日期、日历、代码执行能力、对专用信息源的访问等
 
-#### Agentica Architecture  
+#### Agentica Workflow
+
+**Agentica** can also build multi-agent systems and workflows.
+
+**Agentica** 还可以构建多Agent系统和工作流。
+
 <img src="https://github.com/shibing624/agentica/blob/main/docs/agent_arch.png" width="800" />
 
 - **Planner**：负责让LLM生成一个多步计划来完成复杂任务，生成相互依赖的“链式计划”，定义每一步所依赖的上一步的输出
 - **Worker**：接受“链式计划”，循环遍历计划中的每个子任务，并调用工具完成任务，可以自动反思纠错以完成任务
 - **Solver**：求解器将所有这些输出整合为最终答案
 
+## 🔥 News
+[2024/12/25] v0.2.0版本: 支持了多模态模型，输入可以是文本、图片、音频、视频，升级Assistant为Agent，Workflow支持拆解并实现复杂任务，详见[Release-v0.2.0](https://github.com/shibing624/agentica/releases/tag/0.2.0)
+
+[2024/07/02] v0.1.0版本：实现了基于LLM的Assistant，可以快速用function call搭建大语言模型助手，详见[Release-v0.1.0](https://github.com/shibing624/agentica/releases/tag/0.1.0)
+
 
-## Features
+## 😊 Features
 `Agentica`是一个用于构建Agent的工具，具有以下功能：
 
-- **Agent编排**：通过简单代码快速编排Agent，支持 Reflection(反思）、Plan and Solve(计划并执行)、RAG、Agent、Multi-Agent、Multi-Role、Workflow等功能
+- **Agent编排**：通过简单代码快速编排Agent，支持 Reflection(反思）、Plan and Solve(计划并执行)、RAG、Agent、Multi-Agent、Team、Workflow等功能
 - **自定义prompt**：Agent支持自定义prompt和多种工具调用（tool_calls）
 - **LLM集成**：支持OpenAI、Azure、Deepseek、Moonshot、Claude、Ollama、Together等多方大模型厂商的API
 - **记忆功能**：包括短期记忆和长期记忆功能
-- **Multi-Agent协作**：支持多Agent和多角色的团队协作。
+- **Multi-Agent协作**：支持多Agent和任务委托（Team）的团队协作。
 - **Workflow工作流**：拆解复杂任务为多个Agent，基于工作流自动化串行逐步完成任务，如投资研究、新闻文章撰写和技术教程创建
 - **自我进化Agent**：具有反思和增强记忆能力的自我进化Agent
 - **Web UI**：兼容ChatPilot，可以基于Web页面交互，支持主流的open-webui、streamlit、gradio等前端交互框架
 
-## Installation
+## 💾 Install
 
 ```bash
 pip install -U agentica
@@ -65,7 +79,7 @@ cd agentica
 pip install .
 ```
 
-## Getting Started
+## 🚀 Getting Started
 
 #### Run the example
 ```shell
@@ -96,7 +110,7 @@ print(r)
 ```
 
 
-## Web UI
+## ▶️ Web UI
 
 [shibing624/ChatPilot](https://github.com/shibing624/ChatPilot) 兼容`agentica`，可以通过Web UI进行交互。
 
@@ -115,7 +129,7 @@ bash start.sh
 ```
 
 
-## Examples
+## 😀 Examples
 
 
 | 示例                                                                                                                                                    | 描述                                                                                                                                |
@@ -199,7 +213,7 @@ streamlit run 34_llm_os_demo.py
 
 <img alt="LLM OS" src="https://github.com/shibing624/agentica/blob/main/docs/llm_os_snap.png" width="800" />
 
-## Contact
+## ☎️ Contact
 
 - Issue(建议)
   ：[![GitHub issues](https://img.shields.io/github/issues/shibing624/agentica.svg)](https://github.com/shibing624/agentica/issues)
@@ -208,7 +222,7 @@ streamlit run 34_llm_os_demo.py
 
 <img src="https://github.com/shibing624/agentica/blob/main/docs/wechat.jpeg" width="200" />
 
-## Citation
+## 😇 Citation
 
 如果你在研究中使用了`agentica`，请按如下格式引用：
 
@@ -229,10 +243,10 @@ BibTeX:
 }
 ```
 
-## License
+## ⚠️ License
 
 授权协议为 [The Apache License 2.0](/LICENSE)，可免费用做商业用途。请在产品说明中附加`agentica`的链接和授权协议。
-## Contribute
+## 😍 Contribute
 
 项目代码还很粗糙，如果大家对代码有所改进，欢迎提交回本项目，在提交之前，注意以下两点：
 
@@ -241,11 +255,11 @@ BibTeX:
 
 之后即可提交PR。
 
-## Acknowledgements 
+## 💕 Acknowledgements
 
-- [https://github.com/langchain-ai/langchain](https://github.com/langchain-ai/langchain)
-- [https://github.com/simonmesmith/agentflow](https://github.com/simonmesmith/agentflow)
-- [https://github.com/phidatahq/phidata](https://github.com/phidatahq/phidata)
+- [langchain-ai/langchain](https://github.com/langchain-ai/langchain)
+- [simonmesmith/agentflow](https://github.com/simonmesmith/agentflow)
+- [phidatahq/phidata](https://github.com/phidatahq/phidata)
 
 
 Thanks for their great work!

{agentica-0.2.1 → agentica-0.2.2}/agentica/tools/jina_tool.py

@@ -28,7 +28,6 @@ class JinaTool(Toolkit):
             work_dir: str = None,
     ):
         super().__init__(name="jina_tool")
-
         self.api_key = api_key or getenv("JINA_API_KEY")
         self.max_content_length = max_content_length
         self.work_dir = work_dir or os.path.curdir

{agentica-0.2.1 → agentica-0.2.2}/agentica/tools/url_crawler_tool.py

@@ -6,6 +6,7 @@
 import hashlib
 import os
 import re
+import json
 from urllib.parse import urlparse
 
 import markdownify
@@ -17,9 +18,14 @@ from agentica.utils.log import logger
 
 
 class UrlCrawlerTool(Toolkit):
-    def __init__(self, base_dir: str = os.path.curdir):
+    def __init__(
+            self,
+            base_dir: str = os.path.curdir,
+            max_content_length: int = 8000,
+    ):
         super().__init__(name="url_crawler_tool")
         self.base_dir = base_dir
+        self.max_content_length = max_content_length
         self.register(self.url_crawl)
 
     @staticmethod
@@ -67,7 +73,14 @@ class UrlCrawlerTool(Toolkit):
         webpage_text = "# " + title + "\n\n" + webpage_text
         return webpage_text
 
-    def url_crawl(self, url: str) -> dict:
+    def _trim_content(self, content: str) -> str:
+        """Trim to the maximum allowed length."""
+        if len(content) > self.max_content_length:
+            truncated = content[: self.max_content_length]
+            return truncated + "... (content truncated)"
+        return content
+
+    def url_crawl(self, url: str) -> str:
         """Crawl a website and return the content of the website as a json string.
 
         Args:
@@ -81,7 +94,7 @@ class UrlCrawlerTool(Toolkit):
             print(url, '\n\n', r)
 
         Returns:
-            dict: The dict of url, content, save_path
+            str: The content of the website as a json string.
         """
         filename = self._generate_file_name_from_url(url)
         save_path = os.path.realpath(os.path.join(self.base_dir, filename))
@@ -92,7 +105,7 @@ class UrlCrawlerTool(Toolkit):
             headers = {
                 "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.110 Safari/537.36",
             }
-            response = requests.get(url, stream=True, headers=headers, timeout=30, verify=False)
+            response = requests.get(url, stream=True, headers=headers, timeout=60, verify=False)
             response.raise_for_status()
 
             content_type = response.headers.get("content-type", "")
@@ -113,10 +126,11 @@ class UrlCrawlerTool(Toolkit):
             logger.debug(f"Failed to crawl: {url}: {e}")
         crawler_result = {
             "url": url,
-            "content": content,
+            "content": self._trim_content(content),
             "save_path": save_path,
         }
-        return crawler_result
+        result = json.dumps(crawler_result, ensure_ascii=False)
+        return result
 
 
 if __name__ == '__main__':

{agentica-0.2.1 → agentica-0.2.2}/agentica/utils/file_parser.py

@@ -17,6 +17,7 @@ from agentica.utils.log import logger
 
 
 def read_json_file(path: Path) -> str:
+    """Read JSON file."""
     try:
         if path.suffix == ".jsonl":
             with path.open("r", encoding="utf-8") as f:
@@ -35,6 +36,7 @@ def read_json_file(path: Path) -> str:
 
 
 def read_csv_file(path: Path, row_limit: Optional[int] = None) -> str:
+    """Read CSV file."""
     try:
         with open(path, newline="") as csvfile:
             reader = csv.DictReader(csvfile)
@@ -49,6 +51,7 @@ def read_csv_file(path: Path, row_limit: Optional[int] = None) -> str:
 
 
 def read_txt_file(path: Path) -> str:
+    """Read text file."""
     try:
         return path.read_text("utf-8")
     except Exception as e:
@@ -56,85 +59,59 @@ def read_txt_file(path: Path) -> str:
         return ""
 
 
-def read_pdf_file(path: Path, enable_image_ocr: bool = False) -> str:
+def read_pdf_file(path: Path) -> str:
+    """Read PDF file."""
     try:
-        import pypdf
-    except ImportError:
-        raise ImportError("pypdf is required to read PDF files: `pip install pypdf`")
-
-    if enable_image_ocr:
-        try:
-            import imgocr
-        except ImportError:
-            raise ImportError("use `pip install imgocr`")
-
+        from docling.document_converter import DocumentConverter
+    except (ImportError, ModuleNotFoundError):
+        raise ImportError("`docling` not installed, please install using `pip install docling`")
     text = ""
     try:
-        # Read PDF text
-        reader = pypdf.PdfReader(str(path))
-        for page in reader.pages:
-            text += page.extract_text() + "\n"
-
-        if enable_image_ocr:
-            # Read images and perform OCR
-            ocr = imgocr.ImgOcr()
-            for page in reader.pages:
-                for image in page.images:
-                    if image is None or not image:  # Check if the image data is valid
-                        logger.warning(f"Invalid image data: {image}")
-                        continue
-                    img_array = np.array(image)
-                    if img_array.ndim == 3 and img_array.shape[-1] == 4:  # Check if the image has an alpha channel
-                        img_array = img_array[:, :, :3]  # Remove the alpha channel
-                    elif img_array.ndim != 3 or img_array.shape[-1] not in [3, 4]:
-                        logger.warning(f"Unexpected image shape: {img_array.shape}")
-                        continue
-                    ocr_result = ocr.ocr(img_array)
-                    for result in ocr_result:
-                        text += result["text"] + "\n"
+        logger.info(f"Reading: {path}")
+        converter = DocumentConverter()
+        res = converter.convert(path)
+        text = res.document.export_to_markdown()
     except Exception as e:
         logger.error(f"Error reading: {path}: {e}")
     return text
 
 
-def read_pdf_url(url: str, enable_image_ocr: bool = False) -> str:
+def read_pdf_url(url: str) -> str:
+    """Read PDF file from URL."""
+    text = ""
     try:
-        from pypdf import PdfReader
-    except ImportError:
-        raise ImportError("`pypdf` not installed, use `pip install pypdf`")
-    if enable_image_ocr:
-        try:
-            import imgocr
-        except ImportError:
-            raise ImportError("use `pip install imgocr`")
+        from docling.document_converter import DocumentConverter
+    except (ImportError, ModuleNotFoundError):
+        raise ImportError("`docling` not installed, please install using `pip install docling`")
     try:
         logger.info(f"Reading: {url}")
-        # Create a default context for HTTPS requests (not recommended for production)
-        ssl._create_default_https_context = ssl._create_unverified_context
-        response = httpx.get(url)
-        doc_reader = PdfReader(BytesIO(response.content))
-        ocr_model = imgocr.ImgOcr() if enable_image_ocr else None
-        text_content = ""
-        for num, page in enumerate(doc_reader.pages):
-            page_text = page.extract_text() or ""
-            image_text_list = []
-            if enable_image_ocr:
-                for img_obj in page.images:
-                    img_data = img_obj.data
-                    # Perform OCR
-                    ocr_result = ocr_model.ocr(img_data)
-                    if ocr_result:
-                        image_text_list += [i.get('text') for i in ocr_result]
-            images_text = "\n".join(image_text_list)
-            content = page_text + "\n" + images_text
-            text_content += content
-        return text_content
+        converter = DocumentConverter()
+        res = converter.convert(url)
+        text = res.document.export_to_markdown()
     except Exception as e:
         logger.error(f"Error reading: {url}: {e}")
-        return ""
+    return text
+
+
+def read_docx_file_by_docling(path: Path) -> str:
+    """Read DOCX file using docling."""
+    text = ""
+    try:
+        from docling.document_converter import DocumentConverter
+    except (ImportError, ModuleNotFoundError):
+        raise ImportError("`docling` not installed, please install using `pip install docling`")
+    try:
+        logger.info(f"Reading: {path}")
+        converter = DocumentConverter()
+        res = converter.convert(path)
+        text = res.document.export_to_markdown()
+    except Exception as e:
+        logger.error(f"Error reading: {path}: {e}")
+    return text
 
 
 def read_docx_file(path: Path) -> str:
+    """Read DOCX file."""
     try:
         from docx import Document as DocxDocument
         from docx.oxml.ns import qn
@@ -237,36 +214,23 @@ def read_docx_file(path: Path) -> str:
 
 
 def read_excel_file(path: Path) -> str:
+    """Read excel file."""
+    text = ""
     try:
-        from openpyxl import load_workbook
-    except ImportError:
-        raise ImportError("`openpyxl` not installed, please install using `pip install openpyxl`")
-
+        from docling.document_converter import DocumentConverter
+    except (ImportError, ModuleNotFoundError):
+        raise ImportError("`docling` not installed, please install using `pip install docling`")
     try:
-        # 使用openpyxl读取Excel文件
-        wb = load_workbook(str(path))
-        sheet = wb.active
-        logger.debug(f"Reading: {path}, load active Sheet: {sheet.title}")
-
-        data = []
-        for row in sheet.iter_rows(values_only=False):
-            row_data = {}
-            for idx, cell in enumerate(row):
-                cell_name = sheet.cell(row=1, column=idx + 1).value
-                if hasattr(cell, 'hyperlink') and cell.hyperlink:
-                    row_val = f"[{cell.value}]({cell.hyperlink.target})"
-                    row_data[cell_name] = row_val
-                else:
-                    row_data[cell_name] = cell.value
-            data.append(row_data)
-
-        return json.dumps(data, ensure_ascii=False)
+        logger.info(f"Reading: {path}")
+        converter = DocumentConverter()
+        res = converter.convert(path)
+        text = res.document.export_to_markdown()
     except Exception as e:
         logger.error(f"Error reading: {path}: {e}")
-        return ""
+    return text
 
 
 if __name__ == '__main__':
     url = "https://arxiv.org/pdf/2412.15166"
     result = read_pdf_url(url)
-    print(result[:300])
+    print(result)

agentica-0.2.2/agentica/version.py

@@ -0,0 +1 @@
+__version__ = "0.2.2"