agentica 0.0.6__tar.gz → 0.1.0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (87) hide show
  1. {agentica-0.0.6 → agentica-0.1.0}/PKG-INFO +37 -18
  2. {agentica-0.0.6 → agentica-0.1.0}/README.md +36 -17
  3. {agentica-0.0.6 → agentica-0.1.0}/agentica/__init__.py +1 -1
  4. {agentica-0.0.6 → agentica-0.1.0}/agentica/assistant.py +21 -23
  5. agentica-0.0.6/agentica/documents.py → agentica-0.1.0/agentica/knowledge_base.py +193 -135
  6. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/openai_llm.py +2 -4
  7. agentica-0.1.0/agentica/tools/search_arxiv.py +124 -0
  8. agentica-0.1.0/agentica/utils/file_parser.py +126 -0
  9. {agentica-0.0.6 → agentica-0.1.0}/agentica/utils/misc.py +0 -42
  10. {agentica-0.0.6 → agentica-0.1.0}/agentica/vectordb/lancedb.py +4 -2
  11. {agentica-0.0.6 → agentica-0.1.0}/agentica/vectordb/pgvector.py +3 -2
  12. agentica-0.1.0/agentica/version.py +1 -0
  13. {agentica-0.0.6 → agentica-0.1.0}/agentica/workflow.py +9 -10
  14. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/PKG-INFO +37 -18
  15. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/SOURCES.txt +3 -1
  16. agentica-0.1.0/tests/test_llm.py +31 -0
  17. agentica-0.0.6/agentica/version.py +0 -1
  18. agentica-0.0.6/tests/test_llm.py +0 -47
  19. {agentica-0.0.6 → agentica-0.1.0}/LICENSE +0 -0
  20. {agentica-0.0.6 → agentica-0.1.0}/agentica/config.py +0 -0
  21. {agentica-0.0.6 → agentica-0.1.0}/agentica/document.py +0 -0
  22. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/__init__.py +0 -0
  23. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/azure_emb.py +0 -0
  24. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/base.py +0 -0
  25. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/hash_emb.py +0 -0
  26. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/ollama_emb.py +0 -0
  27. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/openai_emb.py +0 -0
  28. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/text2vec_emb.py +0 -0
  29. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/together_emb.py +0 -0
  30. {agentica-0.0.6 → agentica-0.1.0}/agentica/emb/word2vec_emb.py +0 -0
  31. {agentica-0.0.6 → agentica-0.1.0}/agentica/file/__init__.py +0 -0
  32. {agentica-0.0.6 → agentica-0.1.0}/agentica/file/base.py +0 -0
  33. {agentica-0.0.6 → agentica-0.1.0}/agentica/file/csv.py +0 -0
  34. {agentica-0.0.6 → agentica-0.1.0}/agentica/file/txt.py +0 -0
  35. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/__init__.py +0 -0
  36. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/anthropic_llm.py +0 -0
  37. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/azure_llm.py +0 -0
  38. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/base.py +0 -0
  39. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/ollama_llm.py +0 -0
  40. {agentica-0.0.6 → agentica-0.1.0}/agentica/llm/together_llm.py +0 -0
  41. {agentica-0.0.6 → agentica-0.1.0}/agentica/memory.py +0 -0
  42. {agentica-0.0.6 → agentica-0.1.0}/agentica/message.py +0 -0
  43. {agentica-0.0.6 → agentica-0.1.0}/agentica/python_assistant.py +0 -0
  44. {agentica-0.0.6 → agentica-0.1.0}/agentica/references.py +0 -0
  45. {agentica-0.0.6 → agentica-0.1.0}/agentica/run_record.py +0 -0
  46. {agentica-0.0.6 → agentica-0.1.0}/agentica/sqlite_storage.py +0 -0
  47. {agentica-0.0.6 → agentica-0.1.0}/agentica/task.py +0 -0
  48. {agentica-0.0.6 → agentica-0.1.0}/agentica/tool.py +0 -0
  49. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/__init__.py +0 -0
  50. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/airflow.py +0 -0
  51. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/analyze_image.py +0 -0
  52. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/apify.py +0 -0
  53. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/create_image.py +0 -0
  54. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/duckduckgo.py +0 -0
  55. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/file.py +0 -0
  56. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/jina.py +0 -0
  57. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/ocr.py +0 -0
  58. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/run_nb_code.py +0 -0
  59. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/run_python_code.py +0 -0
  60. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/search_exa.py +0 -0
  61. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/search_serper.py +0 -0
  62. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/shell.py +0 -0
  63. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/sql.py +0 -0
  64. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/url_crawler.py +0 -0
  65. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/wikipedia.py +0 -0
  66. {agentica-0.0.6 → agentica-0.1.0}/agentica/tools/yfinance.py +0 -0
  67. {agentica-0.0.6 → agentica-0.1.0}/agentica/utils/__init__.py +0 -0
  68. {agentica-0.0.6 → agentica-0.1.0}/agentica/utils/log.py +0 -0
  69. {agentica-0.0.6 → agentica-0.1.0}/agentica/utils/shell.py +0 -0
  70. {agentica-0.0.6 → agentica-0.1.0}/agentica/utils/timer.py +0 -0
  71. {agentica-0.0.6 → agentica-0.1.0}/agentica/vectordb/__init__.py +0 -0
  72. {agentica-0.0.6 → agentica-0.1.0}/agentica/vectordb/base.py +0 -0
  73. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/dependency_links.txt +0 -0
  74. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/entry_points.txt +0 -0
  75. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/not-zip-safe +0 -0
  76. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/requires.txt +0 -0
  77. {agentica-0.0.6 → agentica-0.1.0}/agentica.egg-info/top_level.txt +0 -0
  78. {agentica-0.0.6 → agentica-0.1.0}/setup.cfg +0 -0
  79. {agentica-0.0.6 → agentica-0.1.0}/setup.py +0 -0
  80. {agentica-0.0.6 → agentica-0.1.0}/tests/__init__.py +0 -0
  81. {agentica-0.0.6 → agentica-0.1.0}/tests/test_function_create_image.py +0 -0
  82. {agentica-0.0.6 → agentica-0.1.0}/tests/test_function_get_url.py +0 -0
  83. {agentica-0.0.6 → agentica-0.1.0}/tests/test_function_save_file.py +0 -0
  84. {agentica-0.0.6 → agentica-0.1.0}/tests/test_run_nb_code.py +0 -0
  85. {agentica-0.0.6 → agentica-0.1.0}/tests/test_sqlite_storage.py +0 -0
  86. {agentica-0.0.6 → agentica-0.1.0}/tests/test_write_code.py +0 -0
  87. {agentica-0.0.6 → agentica-0.1.0}/tests/test_write_plan.py +0 -0
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: agentica
3
- Version: 0.0.6
3
+ Version: 0.1.0
4
4
  Summary: LLM agents
5
5
  Home-page: https://github.com/shibing624/agentica
6
6
  Author: XuMing
@@ -63,9 +63,9 @@ License-File: LICENSE
63
63
  ## Features
64
64
  `agentica`是一个Agent构建工具,功能:
65
65
 
66
- - 简单代码快速编排Agent,支持RAG、WorkflowMultiAgentMultiRole等
66
+ - 简单代码快速编排Agent,支持 Reflection(反思)、Plan and Solve(计划并执行)、RAG、AgentMulti-AgentMulti-Role、Workflow等功能
67
67
  - Agent支持prompt自定义,支持多种工具调用(tool_calls)
68
- - 支持OpenAI API和Moonshot API(kimi)调用
68
+ - 支持OpenAI/Azure/Claude/Ollama/Together API调用
69
69
 
70
70
  ## Installation
71
71
 
@@ -104,27 +104,46 @@ m = Assistant(
104
104
  read_chat_history=True,
105
105
  debug_mode=True,
106
106
  )
107
- m.run("一句话介绍林黛玉")
108
- m.run("北京最近的新闻", stream=True, print_output=True)
109
- m.run("总结前面的问答")
107
+
108
+ r = m.run("一句话介绍林黛玉")
109
+ print(r, "".join(r))
110
+ r = m.run("北京最近的新闻", stream=True, print_output=True)
111
+ print(r, "".join(r))
112
+ r = m.run("总结前面的问答", stream=False, print_output=False)
113
+ print(r)
110
114
  ```
111
115
 
112
116
 
113
117
  ## Examples
114
118
 
115
- | 示例 | 描述 |
116
- |-------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
117
- | [examples/naive_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/naive_rag_demo.py) | 实现了基础版RAG,基于Txt文档回答问题 |
118
- | [examples/advanced_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/advanced_rag_demo.py) | 实现了高级版RAG,基于PDF文档回答问题,新增功能:pdf文件解析、query改写,字面+语义多路召回,召回排序(rerank) |
119
- | [examples/python_assistant_demo.py](https://github.com/shibing624/agentica/blob/main/examples/python_assistant_demo.py) | 实现了Code Interpreter功能,自动生成python代码,并执行 |
120
- | [examples/research_demo.py](https://github.com/shibing624/agentica/blob/main/examples/research_demo.py) | 实现了Research功能,自动调用搜索工具,汇总信息后撰写科技报告 |
121
- | [examples/run_flow_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/run_flow_news_article_demo.py) | 实现了写新闻稿的工作流,multi-agent的实现,定义了多个Assistant和Task,多次调用搜索工具,并生成高级排版的新闻文章 |
122
- | [examples/run_flow_investment_demo.py](https://github.com/shibing624/agentica/blob/main/examples/run_flow_investment_demo.py) | 实现了投资研究的工作流:股票信息收集 - 股票分析 - 撰写分析报告 - 复查报告等多个Task |
123
- | [examples/crawl_webpage.py](https://github.com/shibing624/agentica/blob/main/examples/crawl_webpage.py) | 实现了网页分析工作流:从Url爬取融资快讯 - 分析网页内容和格式 - 提取核心信息 - 汇总保存为md文件 |
124
- | [examples/find_paper_from_arxiv.py](https://github.com/shibing624/agentica/blob/main/examples/find_paper_from_arxiv.py) | 实现了论文推荐工作流:自动从arxiv搜索多组论文 - 相似论文去重 - 提取核心论文信息 - 保存为csv文件 |
125
- | [examples/remove_image_background.py](https://github.com/shibing624/agentica/blob/main/examples/remove_image_background.py) | 实现了自动去除图片背景功能,包括自动通过pip安装库,调用库实现去除图片背景 |
126
- | [examples/text_classification_demo.py](https://github.com/shibing624/agentica/blob/main/examples/text_classification_demo.py) | 实现了自动训练分类模型的工作流:读取训练集文件并理解格式 - 谷歌搜索pytextclassifier库 - 爬取github页面了解pytextclassifier的调用方法 - 写代码并执行fasttext模型训练 - check训练好的模型预测结果 |
119
+ | 示例 | 描述 |
120
+ |---------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
121
+ | [examples/naive_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/naive_rag_demo.py) | 实现了基础版RAG,基于Txt文档回答问题 |
122
+ | [examples/advanced_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/advanced_rag_demo.py) | 实现了高级版RAG,基于PDF文档回答问题,新增功能:pdf文件解析、query改写,字面+语义多路召回,召回排序(rerank) |
123
+ | [examples/python_assistant_demo.py](https://github.com/shibing624/agentica/blob/main/examples/python_assistant_demo.py) | 实现了Code Interpreter功能,自动生成python代码,并执行 |
124
+ | [examples/research_demo.py](https://github.com/shibing624/agentica/blob/main/examples/research_demo.py) | 实现了Research功能,自动调用搜索工具,汇总信息后撰写科技报告 |
125
+ | [examples/team_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/team_news_article_demo.py) | 实现了写新闻稿的team协作,multi-role实现,委托不用角色完成各自任务:研究员检索分析文章,撰写员根据排版写文章,汇总多角色成果输出结果 |
126
+ | [examples/workflow_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/workflow_news_article_demo.py) | 实现了写新闻稿的工作流,multi-agent的实现,定义了多个Assistant和Task,多次调用搜索工具,并生成高级排版的新闻文章 |
127
+ | [examples/workflow_investment_demo.py](https://github.com/shibing624/agentica/blob/main/examples/workflow_investment_demo.py) | 实现了投资研究的工作流:股票信息收集 - 股票分析 - 撰写分析报告 - 复查报告等多个Task |
128
+ | [examples/crawl_webpage_demo.py](https://github.com/shibing624/agentica/blob/main/examples/crawl_webpage_demo.py) | 实现了网页分析工作流:从Url爬取融资快讯 - 分析网页内容和格式 - 提取核心信息 - 汇总保存为md文件 |
129
+ | [examples/find_paper_from_arxiv_demo.py](https://github.com/shibing624/agentica/blob/main/examples/find_paper_from_arxiv_demo.py) | 实现了论文推荐工作流:自动从arxiv搜索多组论文 - 相似论文去重 - 提取核心论文信息 - 保存为csv文件 |
130
+ | [examples/remove_image_background_demo.py](https://github.com/shibing624/agentica/blob/main/examples/remove_image_background_demo.py) | 实现了自动去除图片背景功能,包括自动通过pip安装库,调用库实现去除图片背景 |
131
+ | [examples/text_classification_demo.py](https://github.com/shibing624/agentica/blob/main/examples/text_classification_demo.py) | 实现了自动训练分类模型的工作流:读取训练集文件并理解格式 - 谷歌搜索pytextclassifier库 - 爬取github页面了解pytextclassifier的调用方法 - 写代码并执行fasttext模型训练 - check训练好的模型预测结果 |
132
+ | [examples/llm_os_demo.py](https://github.com/shibing624/agentica/blob/main/examples/llm_os_demo.py) | 实现了LLM OS的初步设计,基于LLM设计操作系统,可以通过LLM调用RAG、代码执行器、Shell等工具,并协同代码解释器、研究助手、投资助手等来解决问题。 |
133
+
134
+
135
+ ### LLM OS
136
+ The LLM OS design:
137
+ <img alt="LLM OS" src="https://github.com/shibing624/agentica/blob/main/docs/llmos.png" width="600" />
138
+
139
+ #### Run the LLM OS App
140
+
141
+ ```shell
142
+ cd examples
143
+ streamlit run llm_os_demo.py
144
+ ```
127
145
 
146
+ ![llm_os](https://github.com/shibing624/agentica/blob/main/docs/llm_os_snap.png)
128
147
 
129
148
  ## Contact
130
149
 
@@ -41,9 +41,9 @@
41
41
  ## Features
42
42
  `agentica`是一个Agent构建工具,功能:
43
43
 
44
- - 简单代码快速编排Agent,支持RAG、WorkflowMultiAgentMultiRole等
44
+ - 简单代码快速编排Agent,支持 Reflection(反思)、Plan and Solve(计划并执行)、RAG、AgentMulti-AgentMulti-Role、Workflow等功能
45
45
  - Agent支持prompt自定义,支持多种工具调用(tool_calls)
46
- - 支持OpenAI API和Moonshot API(kimi)调用
46
+ - 支持OpenAI/Azure/Claude/Ollama/Together API调用
47
47
 
48
48
  ## Installation
49
49
 
@@ -82,27 +82,46 @@ m = Assistant(
82
82
  read_chat_history=True,
83
83
  debug_mode=True,
84
84
  )
85
- m.run("一句话介绍林黛玉")
86
- m.run("北京最近的新闻", stream=True, print_output=True)
87
- m.run("总结前面的问答")
85
+
86
+ r = m.run("一句话介绍林黛玉")
87
+ print(r, "".join(r))
88
+ r = m.run("北京最近的新闻", stream=True, print_output=True)
89
+ print(r, "".join(r))
90
+ r = m.run("总结前面的问答", stream=False, print_output=False)
91
+ print(r)
88
92
  ```
89
93
 
90
94
 
91
95
  ## Examples
92
96
 
93
- | 示例 | 描述 |
94
- |-------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
95
- | [examples/naive_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/naive_rag_demo.py) | 实现了基础版RAG,基于Txt文档回答问题 |
96
- | [examples/advanced_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/advanced_rag_demo.py) | 实现了高级版RAG,基于PDF文档回答问题,新增功能:pdf文件解析、query改写,字面+语义多路召回,召回排序(rerank) |
97
- | [examples/python_assistant_demo.py](https://github.com/shibing624/agentica/blob/main/examples/python_assistant_demo.py) | 实现了Code Interpreter功能,自动生成python代码,并执行 |
98
- | [examples/research_demo.py](https://github.com/shibing624/agentica/blob/main/examples/research_demo.py) | 实现了Research功能,自动调用搜索工具,汇总信息后撰写科技报告 |
99
- | [examples/run_flow_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/run_flow_news_article_demo.py) | 实现了写新闻稿的工作流,multi-agent的实现,定义了多个Assistant和Task,多次调用搜索工具,并生成高级排版的新闻文章 |
100
- | [examples/run_flow_investment_demo.py](https://github.com/shibing624/agentica/blob/main/examples/run_flow_investment_demo.py) | 实现了投资研究的工作流:股票信息收集 - 股票分析 - 撰写分析报告 - 复查报告等多个Task |
101
- | [examples/crawl_webpage.py](https://github.com/shibing624/agentica/blob/main/examples/crawl_webpage.py) | 实现了网页分析工作流:从Url爬取融资快讯 - 分析网页内容和格式 - 提取核心信息 - 汇总保存为md文件 |
102
- | [examples/find_paper_from_arxiv.py](https://github.com/shibing624/agentica/blob/main/examples/find_paper_from_arxiv.py) | 实现了论文推荐工作流:自动从arxiv搜索多组论文 - 相似论文去重 - 提取核心论文信息 - 保存为csv文件 |
103
- | [examples/remove_image_background.py](https://github.com/shibing624/agentica/blob/main/examples/remove_image_background.py) | 实现了自动去除图片背景功能,包括自动通过pip安装库,调用库实现去除图片背景 |
104
- | [examples/text_classification_demo.py](https://github.com/shibing624/agentica/blob/main/examples/text_classification_demo.py) | 实现了自动训练分类模型的工作流:读取训练集文件并理解格式 - 谷歌搜索pytextclassifier库 - 爬取github页面了解pytextclassifier的调用方法 - 写代码并执行fasttext模型训练 - check训练好的模型预测结果 |
97
+ | 示例 | 描述 |
98
+ |---------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------|
99
+ | [examples/naive_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/naive_rag_demo.py) | 实现了基础版RAG,基于Txt文档回答问题 |
100
+ | [examples/advanced_rag_demo.py](https://github.com/shibing624/agentica/blob/main/examples/advanced_rag_demo.py) | 实现了高级版RAG,基于PDF文档回答问题,新增功能:pdf文件解析、query改写,字面+语义多路召回,召回排序(rerank) |
101
+ | [examples/python_assistant_demo.py](https://github.com/shibing624/agentica/blob/main/examples/python_assistant_demo.py) | 实现了Code Interpreter功能,自动生成python代码,并执行 |
102
+ | [examples/research_demo.py](https://github.com/shibing624/agentica/blob/main/examples/research_demo.py) | 实现了Research功能,自动调用搜索工具,汇总信息后撰写科技报告 |
103
+ | [examples/team_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/team_news_article_demo.py) | 实现了写新闻稿的team协作,multi-role实现,委托不用角色完成各自任务:研究员检索分析文章,撰写员根据排版写文章,汇总多角色成果输出结果 |
104
+ | [examples/workflow_news_article_demo.py](https://github.com/shibing624/agentica/blob/main/examples/workflow_news_article_demo.py) | 实现了写新闻稿的工作流,multi-agent的实现,定义了多个Assistant和Task,多次调用搜索工具,并生成高级排版的新闻文章 |
105
+ | [examples/workflow_investment_demo.py](https://github.com/shibing624/agentica/blob/main/examples/workflow_investment_demo.py) | 实现了投资研究的工作流:股票信息收集 - 股票分析 - 撰写分析报告 - 复查报告等多个Task |
106
+ | [examples/crawl_webpage_demo.py](https://github.com/shibing624/agentica/blob/main/examples/crawl_webpage_demo.py) | 实现了网页分析工作流:从Url爬取融资快讯 - 分析网页内容和格式 - 提取核心信息 - 汇总保存为md文件 |
107
+ | [examples/find_paper_from_arxiv_demo.py](https://github.com/shibing624/agentica/blob/main/examples/find_paper_from_arxiv_demo.py) | 实现了论文推荐工作流:自动从arxiv搜索多组论文 - 相似论文去重 - 提取核心论文信息 - 保存为csv文件 |
108
+ | [examples/remove_image_background_demo.py](https://github.com/shibing624/agentica/blob/main/examples/remove_image_background_demo.py) | 实现了自动去除图片背景功能,包括自动通过pip安装库,调用库实现去除图片背景 |
109
+ | [examples/text_classification_demo.py](https://github.com/shibing624/agentica/blob/main/examples/text_classification_demo.py) | 实现了自动训练分类模型的工作流:读取训练集文件并理解格式 - 谷歌搜索pytextclassifier库 - 爬取github页面了解pytextclassifier的调用方法 - 写代码并执行fasttext模型训练 - check训练好的模型预测结果 |
110
+ | [examples/llm_os_demo.py](https://github.com/shibing624/agentica/blob/main/examples/llm_os_demo.py) | 实现了LLM OS的初步设计,基于LLM设计操作系统,可以通过LLM调用RAG、代码执行器、Shell等工具,并协同代码解释器、研究助手、投资助手等来解决问题。 |
111
+
112
+
113
+ ### LLM OS
114
+ The LLM OS design:
115
+ <img alt="LLM OS" src="https://github.com/shibing624/agentica/blob/main/docs/llmos.png" width="600" />
116
+
117
+ #### Run the LLM OS App
118
+
119
+ ```shell
120
+ cd examples
121
+ streamlit run llm_os_demo.py
122
+ ```
105
123
 
124
+ ![llm_os](https://github.com/shibing624/agentica/blob/main/docs/llm_os_snap.png)
106
125
 
107
126
  ## Contact
108
127
 
@@ -8,7 +8,7 @@ from agentica.config import DOTENV_PATH, SMART_LLM, FAST_LLM # noqa, isort: ski
8
8
  from agentica.assistant import Assistant
9
9
  from agentica.python_assistant import PythonAssistant
10
10
  from agentica.document import Document
11
- from agentica.documents import Documents
11
+ from agentica.knowledge_base import KnowledgeBase
12
12
  from agentica.llm.openai_llm import OpenAILLM
13
13
  from agentica.llm.azure_llm import AzureOpenAILLM
14
14
  from agentica.task import Task
@@ -29,7 +29,7 @@ from uuid import uuid4
29
29
  from pydantic import BaseModel, ConfigDict, field_validator, ValidationError
30
30
 
31
31
  from agentica.document import Document
32
- from agentica.documents import Documents
32
+ from agentica.knowledge_base import KnowledgeBase
33
33
  from agentica.llm.base import LLM
34
34
  from agentica.memory import AssistantMemory, Memory
35
35
  from agentica.message import Message
@@ -81,7 +81,7 @@ class Assistant(BaseModel):
81
81
  update_memory_after_run: bool = True
82
82
 
83
83
  # -*- Assistant Knowledge Base
84
- knowledge_base: Optional[Documents] = None
84
+ knowledge_base: Optional[KnowledgeBase] = None
85
85
  # Enable RAG by adding references from the knowledge base to the prompt.
86
86
  add_references_to_prompt: bool = False
87
87
 
@@ -790,6 +790,7 @@ class Assistant(BaseModel):
790
790
  *,
791
791
  stream: bool = True,
792
792
  messages: Optional[List[Union[Dict, Message]]] = None,
793
+ print_output: bool = True,
793
794
  **kwargs: Any,
794
795
  ) -> Iterator[str]:
795
796
  logger.debug(f"*********** Assistant Run Start: {self.run_id} ***********")
@@ -865,6 +866,8 @@ class Assistant(BaseModel):
865
866
  self.llm = cast(LLM, self.llm)
866
867
  if stream and self.streamable:
867
868
  for response_chunk in self.llm.response_stream(messages=llm_messages):
869
+ if print_output:
870
+ print_llm_stream(response_chunk)
868
871
  llm_response += response_chunk
869
872
  yield response_chunk
870
873
  else:
@@ -922,6 +925,8 @@ class Assistant(BaseModel):
922
925
 
923
926
  # -*- Yield final response if not streaming
924
927
  if not stream:
928
+ if print_output:
929
+ print(llm_response)
925
930
  yield llm_response
926
931
 
927
932
  def run(
@@ -929,8 +934,8 @@ class Assistant(BaseModel):
929
934
  message: Optional[Union[List, Dict, str]] = None,
930
935
  *,
931
936
  stream: bool = True,
932
- print_output: bool = True,
933
937
  messages: Optional[List[Union[Dict, Message]]] = None,
938
+ print_output: bool = True,
934
939
  **kwargs: Any,
935
940
  ) -> Union[Iterator[str], str, BaseModel]:
936
941
  # Convert response to structured output if output_model is set
@@ -959,16 +964,11 @@ class Assistant(BaseModel):
959
964
  return self.output or json_resp
960
965
  else:
961
966
  if stream and self.streamable:
962
- resp = self._run(message=message, messages=messages, stream=True, **kwargs)
963
- if print_output:
964
- for chunk in resp:
965
- print_llm_stream(chunk)
967
+ resp = self._run(message=message, messages=messages, stream=True, print_output=print_output, **kwargs)
966
968
  return resp
967
969
  else:
968
- resp = self._run(message=message, messages=messages, stream=False, **kwargs)
970
+ resp = self._run(message=message, messages=messages, stream=False, print_output=print_output, **kwargs)
969
971
  resp = next(resp)
970
- if print_output:
971
- print(resp)
972
972
  return resp
973
973
 
974
974
  async def _arun(
@@ -977,6 +977,7 @@ class Assistant(BaseModel):
977
977
  *,
978
978
  stream: bool = True,
979
979
  messages: Optional[List[Union[Dict, Message]]] = None,
980
+ print_output: bool = True,
980
981
  **kwargs: Any,
981
982
  ) -> AsyncIterator[str]:
982
983
  logger.debug(f"*********** Run Start: {self.run_id} ***********")
@@ -1055,6 +1056,8 @@ class Assistant(BaseModel):
1055
1056
  response_stream = self.llm.aresponse_stream(messages=llm_messages)
1056
1057
  async for response_chunk in response_stream: # type: ignore
1057
1058
  llm_response += response_chunk
1059
+ if print_output:
1060
+ print_llm_stream(response_chunk)
1058
1061
  yield response_chunk
1059
1062
  else:
1060
1063
  llm_response = await self.llm.aresponse(messages=llm_messages)
@@ -1089,6 +1092,8 @@ class Assistant(BaseModel):
1089
1092
 
1090
1093
  # -*- Yield final response if not streaming
1091
1094
  if not stream:
1095
+ if print_output:
1096
+ print(llm_response)
1092
1097
  yield llm_response
1093
1098
 
1094
1099
  async def arun(
@@ -1096,8 +1101,8 @@ class Assistant(BaseModel):
1096
1101
  message: Optional[Union[List, Dict, str]] = None,
1097
1102
  *,
1098
1103
  stream: bool = True,
1099
- print_output: bool = True,
1100
1104
  messages: Optional[List[Union[Dict, Message]]] = None,
1105
+ print_output: bool = True,
1101
1106
  **kwargs: Any,
1102
1107
  ) -> Union[AsyncIterator[str], str, BaseModel]:
1103
1108
  # Convert response to structured output if output_model is set
@@ -1127,17 +1132,11 @@ class Assistant(BaseModel):
1127
1132
  return self.output or json_resp
1128
1133
  else:
1129
1134
  if stream and self.streamable:
1130
- resp = self._arun(message=message, messages=messages, stream=True, **kwargs)
1131
- if print_output:
1132
- async for chunk in resp:
1133
- print_llm_stream(chunk)
1135
+ resp = self._arun(message=message, messages=messages, stream=True, print_output=print_output, **kwargs)
1134
1136
  return resp
1135
1137
  else:
1136
- resp = self._arun(message=message, messages=messages, stream=False, **kwargs)
1137
- r = await resp.__anext__()
1138
- if print_output:
1139
- print(r)
1140
- return r
1138
+ resp = self._arun(message=message, messages=messages, stream=False, print_output=print_output, **kwargs)
1139
+ return await resp.__anext__()
1141
1140
 
1142
1141
  def chat(
1143
1142
  self, message: Union[List, Dict, str], stream: bool = True, **kwargs: Any
@@ -1334,7 +1333,6 @@ class Assistant(BaseModel):
1334
1333
  user: str = "User",
1335
1334
  emoji: str = ":sunglasses:",
1336
1335
  stream: bool = True,
1337
- markdown: bool = False,
1338
1336
  exit_on: Optional[List[str]] = None,
1339
1337
  **kwargs: Any,
1340
1338
  ) -> None:
@@ -1342,11 +1340,11 @@ class Assistant(BaseModel):
1342
1340
  _exit_on = exit_on or ["exit", "quit", "bye"]
1343
1341
  logger.debug(f"Enable cli, exit with {_exit_on[0]}")
1344
1342
  if message:
1345
- self.run(message=message, stream=stream, markdown=markdown, **kwargs)
1343
+ self.run(message=message, stream=stream, **kwargs)
1346
1344
 
1347
1345
  while True:
1348
1346
  message = Prompt.ask(f"[bold] {emoji} {user} [/bold]")
1349
1347
  if message in _exit_on:
1350
1348
  break
1351
1349
 
1352
- self.run(message=message, stream=stream, markdown=markdown, **kwargs)
1350
+ self.run(message=message, stream=stream, **kwargs)