Refactor: magent-ui (#98)

* chore: update dependencies * refactor(langchain): core adaptor * feat(example): langchain example * chore: spell * feat(langchain): handle ai message * feat(chat): add rehype raw plugin * feat(langchain): input & output format * feat(example): interperter example * fix: chat bot width * fix: modify the presentation of the agent execution process * feat: launch agent in notebook and display qrcode * fix: add qrcode * docs: tongyi agent demo * fix: mobile view change * chore(langchain): update environment & version * chore(flow): tailwind auto update * docs: readme * chore: fix config * feat: magent-ui-core * refactor: magent-ui-core * fix: lint * docs: update readme * feat(core): magent-core --------- Co-authored-by: xujingli.xjl <[email protected]>
difizen · Dec 17, 2024 · 632a3ed · 632a3ed
1 parent 11a0abf
commit 632a3ed
Show file tree

Hide file tree

Showing 68 changed files with 3,035 additions and 986 deletions.
diff --git a/.vscode/settings.json b/.vscode/settings.json
@@ -131,7 +131,7 @@
       "source.fixAll.eslint": "explicit"
     }
   },
-  "cSpell.words": ["magent", "uvicorn"],
+  "cSpell.words": ["llms", "magent", "qwen", "tongyi", "uvicorn"],
   "python.analysis.typeCheckingMode": "basic",
   "python.analysis.autoImportCompletions": true
 }
diff --git a/README.md b/README.md
@@ -21,10 +21,9 @@ PEER Multi-Agent Dialogue
 
 #### Future Plans
 
-- November
-  - [ ] Workflow nodes support code nodes
-  - [ ] Workflow nodes support intent recognition
-  - [ ] Clearer debugging information
+- [ ] Workflow nodes support code nodes
+- [ ] Workflow nodes support intent recognition
+- [ ] Clearer debugging information
 
 #### Configuration
 

diff --git a/README.zh-CN.md b/README.zh-CN.md
@@ -4,7 +4,10 @@
 
 ## magent-ui
 
-为本地代码研发提供辅助产品化能力，方便用户调试
+为本地代码研发提供辅助产品化能力，方便用户调试。支持多种 Agent 研发框架。
+
+- [agentUniverse](https://github.com/alipay/agentUniverse)
+- [langchain](https://github.com/langchain-ai/langchain)
 
 ### agentUniverse
 
@@ -21,10 +24,9 @@ PEER 多智能体对话
 
 #### 后续计划
 
-- 11 月
-  - [ ] 工作流节点支持代码节点
-  - [ ] 工作流节点支持意图识别
-  - [ ] 更清晰的调试信息
+- [ ] 工作流节点支持代码节点
+- [ ] 工作流节点支持意图识别
+- [ ] 更清晰的调试信息
 
 #### 配置
 

diff --git a/api/pyproject.toml b/api/pyproject.toml
@@ -11,10 +11,7 @@ dependencies = [
   "pydantic-settings>=2.3.2",
   "psycopg2-binary>=2.9.9",
   "fastapi-pagination>=0.12.25",
-  "langchain>=0.1.20",
-  "langchain-community>=0.0.38",
   "sse-starlette>=2.1.2",
-  "agentuniverse>=0.0.11",
   "langchain_openai",
 ]
 readme = "README.md"

diff --git a/examples/langchain-example/.gitignore b/examples/langchain-example/.gitignore
@@ -0,0 +1,11 @@
+# python generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# venv
+.venv
+
diff --git a/examples/langchain-example/.python-version b/examples/langchain-example/.python-version
@@ -0,0 +1 @@
+3.10
diff --git a/examples/langchain-example/MANIFEST.in b/examples/langchain-example/MANIFEST.in
@@ -0,0 +1 @@
+recursive-include src/magent_ui/static *
diff --git a/examples/langchain-example/README.md b/examples/langchain-example/README.md
@@ -0,0 +1 @@
+# magent_ui
diff --git a/examples/langchain-example/openai_agent.ipynb b/examples/langchain-example/openai_agent.ipynb
@@ -0,0 +1,223 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "9bf3d1a7-bbbe-445c-b74d-cd31692ae30e",
+   "metadata": {
+    "execution": {
+     "shell.execute_reply.end": "2024-11-19T11:05:54.981070Z",
+     "shell.execute_reply.started": "2024-11-19T11:05:54.911264Z",
+     "to_execute": "2024-11-19T11:05:54.969Z"
+    },
+    "libroFormatter": "formatter-string"
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "AgentExecutor(verbose=True, tags=['zero-shot-react-description'], agent=ZeroShotAgent(llm_chain=LLMChain(verbose=False, prompt=PromptTemplate(input_variables=['agent_scratchpad', 'input'], input_types={}, partial_variables={}, template='Answer the following questions as best you can. You have access to the following tools:\\n\\nhash_string(word: str) - This function to get the hash value of a word.\\n\\n:param input_string: The string to be hashed\\n:return: The hash value of the input string\\n\\nUse the following format:\\n\\nQuestion: the input question you must answer\\nThought: you should always think about what to do\\nAction: the action to take, should be one of [hash_string]\\nAction Input: the input to the action\\nObservation: the result of the action\\n... (this Thought/Action/Action Input/Observation can repeat N times)\\nThought: I now know the final answer\\nFinal Answer: the final answer to the original input question\\n\\nBegin!\\n\\nQuestion: {input}\\nThought:{agent_scratchpad}'), llm=ChatOpenAI(client=<openai.resources.chat.completions.Completions object at 0x11ba8d460>, async_client=<openai.resources.chat.completions.AsyncCompletions object at 0x11ba12b20>, root_client=<openai.OpenAI object at 0x11ba59dc0>, root_async_client=<openai.AsyncOpenAI object at 0x11ba8d490>, model_name='gpt-4o', temperature=0.0, model_kwargs={}, openai_api_key=SecretStr('**********')), output_parser=StrOutputParser(), llm_kwargs={}), output_parser=MRKLOutputParser(), allowed_tools=['hash_string']), tools=[StructuredTool(name='hash_string', description='This function to get the hash value of a word.\\n\\n:param input_string: The string to be hashed\\n:return: The hash value of the input string', args_schema=<class 'langchain_core.utils.pydantic.hash_string'>, func=<function hash_string at 0x11ba78d30>)])"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from langchain_openai import ChatOpenAI\n",
+    "from langchain_core.tools import tool\n",
+    "from langchain.agents import initialize_agent\n",
+    "from langchain.agents import AgentType\n",
+    "\n",
+    "\n",
+    "@tool\n",
+    "def hash_string(word:str):\n",
+    "    \"\"\"\n",
+    "    This function to get the hash value of a word.\n",
+    "\n",
+    "    :param input_string: The string to be hashed\n",
+    "    :return: The hash value of the input string\n",
+    "    \"\"\"\n",
+    "    import hashlib\n",
+    "    hash_object = hashlib.sha256()\n",
+    "    hash_object.update(word.encode())\n",
+    "    hash_value = hash_object.hexdigest()\n",
+    "    return hash_value\n",
+    "\n",
+    "\n",
+    "llm = ChatOpenAI(\n",
+    "    model=\"gpt-4o\",\n",
+    "    temperature=0,\n",
+    "    max_tokens=None,\n",
+    "    timeout=None,\n",
+    "    max_retries=2,\n",
+    ")\n",
+    "\n",
+    "tools = [hash_string]\n",
+    "\n",
+    "agent = initialize_agent(\n",
+    "    tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)\n",
+    "\n",
+    "agent\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "5e16c965-cedc-45f8-ac41-d0111abf5b5d",
+   "metadata": {
+    "execution": {
+     "shell.execute_reply.end": "2024-11-19T11:05:56.472219Z",
+     "shell.execute_reply.started": "2024-11-19T11:05:56.466182Z",
+     "to_execute": "2024-11-19T11:05:56.523Z"
+    },
+    "libroFormatter": "formatter-string"
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "ChatOpenAI(client=<openai.resources.chat.completions.Completions object at 0x11ba8d460>, async_client=<openai.resources.chat.completions.AsyncCompletions object at 0x11ba12b20>, root_client=<openai.OpenAI object at 0x11ba59dc0>, root_async_client=<openai.AsyncOpenAI object at 0x11ba8d490>, model_name='gpt-4o', temperature=0.0, model_kwargs={}, openai_api_key=SecretStr('**********'))"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "agent.agent.llm_chain.llm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "b70afa86-39c9-4278-96de-da12dedb5ed5",
+   "metadata": {
+    "execution": {
+     "shell.execute_reply.end": "2024-11-19T11:05:59.907734Z",
+     "shell.execute_reply.started": "2024-11-19T11:05:57.155502Z",
+     "to_execute": "2024-11-19T11:05:57.193Z"
+    },
+    "libroFormatter": "formatter-string"
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "\n",
+      "\n",
+      "\u001b[1m> Entering new None chain...\u001b[0m\n",
+      "{'actions': [AgentAction(tool='hash_string', tool_input='Hello, how are you today?', log='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"')], 'messages': [AIMessage(content='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"', additional_kwargs={}, response_metadata={})]}\n",
+      "<class 'langchain_core.runnables.utils.AddableDict'>\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n",
+      "\u001b[32;1m\u001b[1;3mThe word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\n",
+      "Action: hash_string\n",
+      "Action Input: \"Hello, how are you today?\"\u001b[0m\n",
+      "Observation: \u001b[36;1m\u001b[1;3m6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48\u001b[0m\n",
+      "Thought:{'steps': [AgentStep(action=AgentAction(tool='hash_string', tool_input='Hello, how are you today?', log='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"'), observation='6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48')], 'messages': [HumanMessage(content='6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48', additional_kwargs={}, response_metadata={})]}\n",
+      "<class 'langchain_core.runnables.utils.AddableDict'>\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n",
+      "\u001b[32;1m\u001b[1;3mI now know the final answer.\n",
+      "Final Answer: The hash value of the sentence \"Hello, how are you today?\" is 6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48.\u001b[0m\n",
+      "\n",
+      "\u001b[1m> Finished chain.\u001b[0m\n",
+      "{'output': 'The hash value of the sentence \"Hello, how are you today?\" is 6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48.', 'messages': [AIMessage(content='I now know the final answer.\\nFinal Answer: The hash value of the sentence \"Hello, how are you today?\" is 6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48.', additional_kwargs={}, response_metadata={})]}\n",
+      "<class 'langchain_core.runnables.utils.AddableDict'>\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n",
+      "--- False\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "{'actions': [AgentAction(tool='hash_string', tool_input='Hello, how are you today?', log='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"')],\n",
+       " 'messages': [AIMessage(content='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"', additional_kwargs={}, response_metadata={}),\n",
+       "  HumanMessage(content='6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48', additional_kwargs={}, response_metadata={}),\n",
+       "  AIMessage(content='I now know the final answer.\\nFinal Answer: The hash value of the sentence \"Hello, how are you today?\" is 6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48.', additional_kwargs={}, response_metadata={})],\n",
+       " 'steps': [AgentStep(action=AgentAction(tool='hash_string', tool_input='Hello, how are you today?', log='The word \"你好\" is Chinese and translates to \"Hello\" in English. I will now create a sentence starting with \"Hello\" and then get the hash value of the result.\\nAction: hash_string\\nAction Input: \"Hello, how are you today?\"'), observation='6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48')],\n",
+       " 'output': 'The hash value of the sentence \"Hello, how are you today?\" is 6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48.'}"
+      ]
+     },
+     "execution_count": 7,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "msg_iterator=agent.stream('Translate the follow word to English, and then write a sentence starting with the word, and then get the hash value of the result, \"你好\"')\n",
+    "\n",
+    "from langchain_core.agents import AgentAction, AgentStep, AgentFinish\n",
+    "from langchain_core.messages import  BaseMessageChunk\n",
+    "\n",
+    "first = True\n",
+    "gathered = None\n",
+    "for chunk in msg_iterator:\n",
+    "    print(chunk)\n",
+    "    print(type(chunk))\n",
+    "    print('---', isinstance(chunk, BaseMessageChunk))\n",
+    "    print('---', isinstance(chunk, AgentAction))\n",
+    "    print('---', isinstance(chunk, AgentStep))\n",
+    "    print('---', isinstance(chunk, AgentFinish))\n",
+    "    \n",
+    "    if first:\n",
+    "        gathered = chunk\n",
+    "        first = False\n",
+    "    else:\n",
+    "        gathered = gathered + chunk\n",
+    "gathered\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "id": "c40a71ab-e0d4-4dd1-9fc5-7458eddf1d46",
+   "metadata": {
+    "execution": {
+     "shell.execute_reply.end": "2024-11-19T16:50:27.226125Z",
+     "shell.execute_reply.started": "2024-11-19T16:50:27.221123Z",
+     "to_execute": "2024-11-19T16:50:27.275Z"
+    },
+    "libroFormatter": "formatter-string"
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'6e766a49e512e0ba0bc935e2aacd3e5a4a34add17f83afc4c9e669c70241cd48'"
+      ]
+     },
+     "execution_count": 22,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "gathered.get('steps')[0].observation"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "884e869b-8c1f-40f2-8bc6-23b3a29ab545",
+   "metadata": {
+    "libroFormatter": "formatter-string"
+   },
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {},
+ "nbformat": 4,
+ "nbformat_minor": 5
+}