精通LangGraph-Tools call-03

from typing import Annotatedfrom langchain_core.runnables import RunnableConfigfrom langchain_core.tools import InjectedToolArgfrom langgraph.store.base import BaseStorefrom langgraph.prebuilt import InjectedState, InjectedStore
# 可以是同步或异步；不需要 @tool 装饰器async def my_tool(    # 这些参数由 LLM 填充    some_arg: str,    another_arg: float,    # config: RunnableConfig 在 LangChain 调用中始终可用    # 这不会暴露给 LLM    config: RunnableConfig,    # 以下三个是特定于预构建的 ToolNode    # （以及 `create_react_agent` 扩展）。如果你在自己的节点中单独调用工具，    # 那么你需要自己提供这些。    store: Annotated[BaseStore, InjectedStore],    # 这会传递完整的状态。    state: Annotated[State, InjectedState],    # 你也可以从状态中注入单个字段    messages: Annotated[list, InjectedState("messages")]    # 以下与 create_react_agent 或 ToolNode 不兼容    # 你也可以排除其他不向模型显示的参数。    # 这些必须手动提供，如果你在自己的节点中调用工具/函数，它们会很有用    # some_other_arg=Annotated["MyPrivateClass", InjectedToolArg],):    """调用 my_tool 对现实世界产生影响。        参数:        some_arg: 一个非常重要的参数        another_arg: LLM 将提供的另一个参数    """      # 文档字符串成为工具的描述并传递给模型    print(some_arg, another_arg, config, store, state, messages)    # Config, some_other_arg, store 和 state 在传递给 bind_tools 或 with_structured_output 时    # 对 LangChain 模型是“隐藏”的    return "... some response"

from typing import List
# this is the state schema used by the prebuilt create_react_agent we'll be using belowfrom langgraph.prebuilt.chat_agent_executor import AgentStatefrom langchain_core.documents import Document

class State(AgentState):    docs: List[str]

from typing import List, Tuplefrom typing_extensions import Annotatedfrom langchain_core.messages import ToolMessagefrom langchain_core.tools import toolfrom langgraph.prebuilt import InjectedState
@tooldef get_context(question: str, state: Annotated[dict, InjectedState]):    """Get relevant context for answering the question."""    return "\n\n".join(doc for doc in state["docs"])

print(get_context.get_input_schema().model_json_schema())

{'description': 'Get relevant context for answering the question.', 'properties': {'question': {'title': 'Question', 'type': 'string'}, 'state': {'title': 'State', 'type': 'object'}}, 'required': ['question', 'state'], 'title': 'get_context', 'type': 'object'}

print(get_context.tool_call_schema.model_json_schema())

{'description': 'Get relevant context for answering the question.', 'properties': {'question': {'title': 'Question', 'type': 'string'}}, 'required': ['question'], 'title': 'get_context', 'type': 'object'}

from langchain_ollama import ChatOllamaimport base_conffrom langgraph.prebuilt import ToolNode, create_react_agentfrom langgraph.checkpoint.memory import MemorySavermodel = ChatOllama(base_url=base_conf.base_url,model=base_conf.model_name,temperature=0)tools = [get_context]# ToolNode 将自动负责将state注入tooltool_node = ToolNode(tools)checkpointer = MemorySaver()graph = create_react_agent(model, tools, state_schema=State, checkpointer=checkpointer)

docs = [    "FooBar company just raised 1 Billion dollars!",    "FooBar company was founded in 2019",]inputs = {    "messages": [{"type": "user", "content": "what's the latest news about FooBar"}],    "docs": docs,}config = {"configurable": {"thread_id": "1"}}for chunk in graph.stream(inputs, config, stream_mode="values"):    chunk["messages"][-1].pretty_print()

================================ Human Message =================================
what's the latest news about FooBar================================== Ai Message ==================================Tool Calls:  get_context (e1eeaa88-b5f4-45ae-abf3-ed7fd861ce66) Call ID: e1eeaa88-b5f4-45ae-abf3-ed7fd861ce66  Args:    question: latest news about FooBar================================= Tool Message =================================Name: get_context
FooBar company just raised 1 Billion dollars!
FooBar company was founded in 2019================================== Ai Message ==================================
The latest news about FooBar is that the company has just raised 1 Billion dollars! For reference, FooBar was founded in 2019.

from langgraph.store.memory import InMemoryStore
doc_store = InMemoryStore()
namespace = ("documents", "1")  # user IDdoc_store.put(    namespace, "doc_0", {"doc": "FooBar company just raised 1 Billion dollars!"})namespace = ("documents", "2")  # user IDdoc_store.put(namespace, "doc_1", {"doc": "FooBar company was founded in 2019"})

from langgraph.store.base import BaseStorefrom langchain_core.runnables import RunnableConfigfrom langgraph.prebuilt import InjectedStore

@tooldef get_context(    question: str,    config: RunnableConfig,    store: Annotated[BaseStore, InjectedStore()],) -> Tuple[str, List[Document]]:    """Get relevant context for answering the question."""    user_id = config.get("configurable", {}).get("user_id")    docs = [item.value["doc"] for item in store.search(("documents", user_id))]    return "\n\n".join(doc for doc in docs)

print(get_context.tool_call_schema.model_json_schema())

{'description': 'Get relevant context for answering the question.', 'properties': {'question': {'title': 'Question', 'type': 'string'}}, 'required': ['question'], 'title': 'get_context', 'type': 'object'}

tools = [get_context]# ToolNode 会自动把 Store 注入到 toolstool_node = ToolNode(tools)checkpointer = MemorySaver()graph = create_react_agent(model, tools, checkpointer=checkpointer, store=doc_store)messages = [{"type": "user", "content": "what's the latest news about FooBar"}]config = {"configurable": {"thread_id": "1", "user_id": "1"}}for chunk in graph.stream({"messages": messages}, config, stream_mode="values"):    chunk["messages"][-1].pretty_print()

================================ Human Message =================================
what's the latest news about FooBar================================== Ai Message ==================================Tool Calls:  get_context (89b2d78c-6b6d-4c46-a5a7-cee513bff5cb) Call ID: 89b2d78c-6b6d-4c46-a5a7-cee513bff5cb  Args:    question: latest news about FooBar================================= Tool Message =================================Name: get_context
FooBar company just raised 1 Billion dollars!================================== Ai Message ==================================
The latest news is that FooBar company has just raised 1 Billion dollars!

messages = [{"type": "user", "content": "what's the latest news about FooBar"}]config = {"configurable": {"thread_id": "2", "user_id": "2"}}for chunk in graph.stream({"messages": messages}, config, stream_mode="values"):    chunk["messages"][-1].pretty_print()

================================ Human Message =================================
what's the latest news about FooBar================================== Ai Message ==================================Tool Calls:  get_context (817da555-c13e-4fa1-8bbe-3854713fc643) Call ID: 817da555-c13e-4fa1-8bbe-3854713fc643  Args:    question: latest news about FooBar================================= Tool Message =================================Name: get_context
FooBar company was founded in 2019================================== Ai Message ==================================
The information I have currently states that FooBar company was founded in 2019. However, this doesn't provide the latest news. Could you please specify a date range or give me some more time to fetch the most recent updates?