fix checkpointer missing in fastapi langgraph toolbasedgenui

Author: maxkorp

typescript-sdk/apps/dojo/src/files.json

@@ -570,7 +570,7 @@
     },
     {
       "name": "agent.py",
-      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\n\ngraph = workflow.compile()",
+      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nimport os\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\nworkflow.add_edge(\"chat_node\", END)\n\n\n# Conditionally use a checkpointer based on the environment\n# Check for multiple indicators that we're running in LangGraph dev/API mode\nis_fast_api = os.environ.get(\"LANGGRAPH_FAST_API\", \"false\").lower() == \"true\"\n\n# Compile the graph\nif is_fast_api:\n    # For CopilotKit and other contexts, use MemorySaver\n    from langgraph.checkpoint.memory import MemorySaver\n    memory = MemorySaver()\n    graph = workflow.compile(checkpointer=memory)\nelse:\n    # When running in LangGraph API/dev, don't use a custom checkpointer\n    graph = workflow.compile()\n",
       "language": "python",
       "type": "file"
     },
@@ -759,7 +759,7 @@
     },
     {
       "name": "agent.py",
-      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\n\ngraph = workflow.compile()",
+      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nimport os\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\nworkflow.add_edge(\"chat_node\", END)\n\n\n# Conditionally use a checkpointer based on the environment\n# Check for multiple indicators that we're running in LangGraph dev/API mode\nis_fast_api = os.environ.get(\"LANGGRAPH_FAST_API\", \"false\").lower() == \"true\"\n\n# Compile the graph\nif is_fast_api:\n    # For CopilotKit and other contexts, use MemorySaver\n    from langgraph.checkpoint.memory import MemorySaver\n    memory = MemorySaver()\n    graph = workflow.compile(checkpointer=memory)\nelse:\n    # When running in LangGraph API/dev, don't use a custom checkpointer\n    graph = workflow.compile()\n",
       "language": "python",
       "type": "file"
     }
@@ -971,7 +971,7 @@
     },
     {
       "name": "agent.py",
-      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\n\ngraph = workflow.compile()",
+      "content": "\"\"\"\nAn example demonstrating tool-based generative UI using LangGraph.\n\"\"\"\n\nimport os\nfrom typing import Any, List\nfrom typing_extensions import Literal\nfrom langchain_openai import ChatOpenAI\nfrom langchain_core.messages import SystemMessage\nfrom langchain_core.runnables import RunnableConfig\nfrom langgraph.graph import StateGraph, END\nfrom langgraph.types import Command\nfrom langgraph.graph import MessagesState\nfrom langgraph.prebuilt import ToolNode\n\n\nclass AgentState(MessagesState):\n    \"\"\"\n    State of the agent.\n    \"\"\"\n    tools: List[Any]\n\nasync def chat_node(state: AgentState, config: RunnableConfig) -> Command[Literal[\"tool_node\", \"__end__\"]]:\n    \"\"\"\n    Standard chat node based on the ReAct design pattern. It handles:\n    - The model to use (and binds in CopilotKit actions and the tools defined above)\n    - The system prompt\n    - Getting a response from the model\n    - Handling tool calls\n\n    For more about the ReAct design pattern, see:\n    https://www.perplexity.ai/search/react-agents-NcXLQhreS0WDzpVaS4m9Cg\n    \"\"\"\n\n    model = ChatOpenAI(model=\"gpt-4o\")\n\n    model_with_tools = model.bind_tools(\n        [\n            *state.get(\"tools\", []), # bind tools defined by ag-ui\n        ],\n        parallel_tool_calls=False,\n    )\n\n    system_message = SystemMessage(\n        content=f\"Help the user with writing Haikus. If the user asks for a haiku, use the generate_haiku tool to display the haiku to the user.\"\n    )\n\n    response = await model_with_tools.ainvoke([\n        system_message,\n        *state[\"messages\"],\n    ], config)\n\n    return Command(\n        goto=END,\n        update={\n            \"messages\": [response],\n        }\n    )\n\nworkflow = StateGraph(AgentState)\nworkflow.add_node(\"chat_node\", chat_node)\n# This is required even though we don't have any backend tools to pass in.\nworkflow.add_node(\"tool_node\", ToolNode(tools=[]))\nworkflow.set_entry_point(\"chat_node\")\nworkflow.add_edge(\"chat_node\", END)\n\n\n# Conditionally use a checkpointer based on the environment\n# Check for multiple indicators that we're running in LangGraph dev/API mode\nis_fast_api = os.environ.get(\"LANGGRAPH_FAST_API\", \"false\").lower() == \"true\"\n\n# Compile the graph\nif is_fast_api:\n    # For CopilotKit and other contexts, use MemorySaver\n    from langgraph.checkpoint.memory import MemorySaver\n    memory = MemorySaver()\n    graph = workflow.compile(checkpointer=memory)\nelse:\n    # When running in LangGraph API/dev, don't use a custom checkpointer\n    graph = workflow.compile()\n",
       "language": "python",
       "type": "file"
     },

typescript-sdk/integrations/langgraph/examples/python/agents/tool_based_generative_ui/agent.py

@@ -2,6 +2,7 @@
 An example demonstrating tool-based generative UI using LangGraph.
 """
 
+import os
 from typing import Any, List
 from typing_extensions import Literal
 from langchain_openai import ChatOpenAI
@@ -61,5 +62,19 @@ async def chat_node(state: AgentState, config: RunnableConfig) -> Command[Litera
 # This is required even though we don't have any backend tools to pass in.
 workflow.add_node("tool_node", ToolNode(tools=[]))
 workflow.set_entry_point("chat_node")
+workflow.add_edge("chat_node", END)
 
-graph = workflow.compile()
+
+# Conditionally use a checkpointer based on the environment
+# Check for multiple indicators that we're running in LangGraph dev/API mode
+is_fast_api = os.environ.get("LANGGRAPH_FAST_API", "false").lower() == "true"
+
+# Compile the graph
+if is_fast_api:
+    # For CopilotKit and other contexts, use MemorySaver
+    from langgraph.checkpoint.memory import MemorySaver
+    memory = MemorySaver()
+    graph = workflow.compile(checkpointer=memory)
+else:
+    # When running in LangGraph API/dev, don't use a custom checkpointer
+    graph = workflow.compile()