langgraph-agent-toolkit 0.1.0__py3-none-any.whl

langgraph_agent_toolkit/agents/blueprints/interrupt_agent/agent.py

@@ -0,0 +1,140 @@
+from datetime import datetime
+from typing import Any, cast
+
+from langchain.prompts import SystemMessagePromptTemplate
+from langchain_core.language_models.chat_models import BaseChatModel
+from langchain_core.messages import AIMessage, BaseMessage, HumanMessage
+from langchain_core.runnables import RunnableConfig, RunnableLambda, RunnableSerializable
+from langgraph.checkpoint.memory import MemorySaver
+from langgraph.graph import END, MessagesState, StateGraph
+from langgraph.types import interrupt
+from pydantic import BaseModel, Field
+
+from langgraph_agent_toolkit.agents.agent import Agent
+from langgraph_agent_toolkit.core import settings
+from langgraph_agent_toolkit.core.models.factory import ModelFactory
+
+
+class AgentState(MessagesState, total=False):
+    """`total=False` is PEP589 specs.
+
+    documentation: https://typing.readthedocs.io/en/latest/spec/typeddict.html#totality
+    """
+
+    birthdate: datetime | None
+
+
+def wrap_model(model: BaseChatModel, system_prompt: BaseMessage) -> RunnableSerializable[Any, BaseMessage]:
+    preprocessor = RunnableLambda(
+        lambda state: [system_prompt] + state["messages"],
+        name="StateModifier",
+    )
+    return preprocessor | model
+
+
+background_prompt = SystemMessagePromptTemplate.from_template("""
+You are a helpful assistant that tells users there zodiac sign.
+Provide a one paragraph summary of the origin of zodiac signs.
+Don't tell the user what their sign is, you are just demonstrating your knowledge on the topic.
+""")
+
+
+async def background(state: AgentState, config: RunnableConfig) -> AgentState:
+    """Demonstrate doing work before the interrupt."""
+    m = ModelFactory.create(config["configurable"].get("model", settings.DEFAULT_MODEL))
+    model_runnable = wrap_model(m, background_prompt.format())
+    response = await model_runnable.ainvoke(state, config)
+
+    return {"messages": [AIMessage(content=response.content)]}
+
+
+birthdate_extraction_prompt = SystemMessagePromptTemplate.from_template("""
+You are an expert at extracting birthdates from conversational text.
+
+Rules for extraction:
+- Look for user messages that mention birthdates
+- Consider various date formats (MM/DD/YYYY, YYYY-MM-DD, Month Day, Year)
+- Validate that the date is reasonable (not in the future)
+- If no clear birthdate was provided by the user, return None
+""")
+
+
+class BirthdateExtraction(BaseModel):
+    birthdate: str | None = Field(
+        description="The extracted birthdate in YYYY-MM-DD format. If no birthdate is found, this should be None."
+    )
+    reasoning: str = Field(description="Explanation of how the birthdate was extracted or why no birthdate was found")
+
+
+async def determine_birthdate(state: AgentState, config: RunnableConfig) -> AgentState:
+    """Examine the conversation history to determine user's birthdate.
+
+    If no birthdate is found, it will perform an interrupt before proceeding.
+    """
+    m = ModelFactory.create(config["configurable"].get("model", settings.DEFAULT_MODEL))
+    model_runnable = wrap_model(
+        m.with_structured_output(BirthdateExtraction),
+        birthdate_extraction_prompt.format(),
+    ).with_config(tags=["skip_stream"])
+    response = await model_runnable.ainvoke(state, config)
+    response = cast(BirthdateExtraction, response)
+
+    # If no birthdate found, interrupt
+    if response.birthdate is None:
+        birthdate_input = interrupt(f"{response.reasoning}\nPlease tell me your birthdate?")
+        # Re-run extraction with the new input
+        state["messages"].append(HumanMessage(birthdate_input))
+        return await determine_birthdate(state, config)
+
+    # Birthdate found - convert string to datetime
+    try:
+        birthdate = datetime.fromisoformat(response.birthdate)
+    except ValueError:
+        # If parsing fails, ask for clarification
+        birthdate_input = interrupt(
+            "I couldn't understand the date format. Please provide your birthdate in YYYY-MM-DD format."
+        )
+        state["messages"].append(HumanMessage(birthdate_input))
+        return await determine_birthdate(state, config)
+
+    # Birthdate found
+    return {
+        "birthdate": birthdate,
+    }
+
+
+sign_prompt = SystemMessagePromptTemplate.from_template("""
+You are a helpful assistant that tells users there zodiac sign.
+What is the sign of somebody born on {birthdate}?
+""")
+
+
+async def determine_sign(state: AgentState, config: RunnableConfig) -> AgentState:
+    """Determine the zodiac sign of the user based on their birthdate."""
+    if not state.get("birthdate"):
+        raise ValueError("No birthdate found in state")
+
+    m = ModelFactory.create(config["configurable"].get("model", settings.DEFAULT_MODEL))
+    model_runnable = wrap_model(m, sign_prompt.format(birthdate=state["birthdate"].strftime("%Y-%m-%d")))
+    response = await model_runnable.ainvoke(state, config)
+
+    return {"messages": [AIMessage(content=response.content)]}
+
+
+# Define the graph
+agent = StateGraph(AgentState)
+agent.add_node("background", background)
+agent.add_node("determine_birthdate", determine_birthdate)
+agent.add_node("determine_sign", determine_sign)
+
+agent.set_entry_point("background")
+agent.add_edge("background", "determine_birthdate")
+agent.add_edge("determine_birthdate", "determine_sign")
+agent.add_edge("determine_sign", END)
+
+interrupt_agent = Agent(
+    name="interrupt-agent",
+    description="An agent the uses interrupts.",
+    graph=agent.compile(checkpointer=MemorySaver()),
+)
+interrupt_agent.graph.name = "interrupt-agent"

langgraph_agent_toolkit/agents/blueprints/react/agent.py

@@ -0,0 +1,67 @@
+from langchain_community.tools import DuckDuckGoSearchResults
+from langgraph.checkpoint.memory import MemorySaver
+
+from langgraph_agent_toolkit.agents.agent import Agent
+from langgraph_agent_toolkit.agents.components.creators.create_react_agent import create_react_agent
+from langgraph_agent_toolkit.agents.components.tools import add, multiply
+from langgraph_agent_toolkit.agents.components.utils import AgentStateWithRemainingSteps, pre_model_hook_standard
+from langgraph_agent_toolkit.core.models.factory import ModelFactory
+from langgraph_agent_toolkit.core.observability.factory import ObservabilityFactory
+from langgraph_agent_toolkit.core.observability.types import ChatMessageDict, MessageRole, ObservabilityBackend
+from langgraph_agent_toolkit.core.prompts.chat_prompt_template import ObservabilityChatPromptTemplate
+from langgraph_agent_toolkit.core.settings import settings
+
+
+PROMPT_NAME = "react-assistant"
+
+
+# observability = ObservabilityFactory.create(settings.OBSERVABILITY_BACKEND)
+observability = ObservabilityFactory.create(ObservabilityBackend.LANGFUSE)
+
+# Push the prompt template to the observability platform first
+observability.push_prompt(
+    name=PROMPT_NAME,
+    prompt_template=[
+        ChatMessageDict(
+            role=MessageRole.SYSTEM,
+            content=(
+                "You are a team support agent that can perform calculations and search the web. "
+                "You can use the tools provided to help you with your tasks. "
+                "You can also ask clarifying questions to the user. "
+            ),
+        ),
+        ChatMessageDict(
+            role=MessageRole.PLACEHOLDER,
+            content="messages",
+        ),
+    ],
+    create_new_version=True,
+)
+
+print(observability.pull_prompt(name=PROMPT_NAME))
+
+# Create ObservabilityChatPromptTemplate that will load from observability platform at runtime
+prompt = ObservabilityChatPromptTemplate.from_observability_platform(
+    prompt_name=PROMPT_NAME,
+    observability_platform=observability,
+    load_at_runtime=True,
+    template_format="jinja2",
+    input_variables=["messages"],
+)
+
+react_agent = Agent(
+    name="react-agent",
+    description="A react agent.",
+    graph=create_react_agent(
+        model=ModelFactory.create(settings.DEFAULT_MODEL),
+        tools=[add, multiply, DuckDuckGoSearchResults()],
+        prompt=prompt,
+        pre_model_hook=pre_model_hook_standard,
+        state_schema=AgentStateWithRemainingSteps,
+        checkpointer=MemorySaver(),
+        immediate_step_threshold=5,
+    ),
+    observability=observability,
+)
+
+__all__ = ["react_agent"]

langgraph_agent_toolkit/agents/blueprints/react_so/agent.py

@@ -0,0 +1,39 @@
+from langchain_community.tools import DuckDuckGoSearchResults
+from langgraph.checkpoint.memory import MemorySaver
+from pydantic import BaseModel, Field
+
+from langgraph_agent_toolkit.agents.agent import Agent
+from langgraph_agent_toolkit.agents.components.creators.create_react_agent import create_react_agent
+from langgraph_agent_toolkit.agents.components.tools import add, multiply
+from langgraph_agent_toolkit.agents.components.utils import AgentStateWithRemainingSteps, pre_model_hook_standard
+from langgraph_agent_toolkit.core import settings
+from langgraph_agent_toolkit.core.models.factory import ModelFactory
+
+
+class ResponseSchema(BaseModel):
+    response: list[str] = Field(
+        description="The response on user query.",
+    )
+    alternative_response: str = Field(
+        description="The alternative response on user query.",
+    )
+
+
+react_agent_so = Agent(
+    name="react-agent-so",
+    description="A react agent with structured output.",
+    graph=create_react_agent(
+        model=ModelFactory.create(settings.DEFAULT_MODEL),
+        tools=[add, multiply, DuckDuckGoSearchResults()],
+        prompt=(
+            "You are a team support agent that can perform calculations and search the web. "
+            "You can use the tools provided to help you with your tasks. "
+            "You can also ask clarifying questions to the user. "
+        ),
+        pre_model_hook=pre_model_hook_standard,
+        response_format=ResponseSchema,
+        state_schema=AgentStateWithRemainingSteps,
+        checkpointer=MemorySaver(),
+        immediate_step_threshold=5,
+    ),
+)

langgraph_agent_toolkit/agents/blueprints/supervisor_agent/agent.py

@@ -0,0 +1,44 @@
+from langchain_community.tools import DuckDuckGoSearchResults
+from langgraph.checkpoint.memory import MemorySaver
+from langgraph.prebuilt import create_react_agent
+from langgraph_supervisor import create_supervisor
+
+from langgraph_agent_toolkit.agents.agent import Agent
+from langgraph_agent_toolkit.agents.components.tools import add, multiply
+from langgraph_agent_toolkit.core import settings
+from langgraph_agent_toolkit.core.models.factory import ModelFactory
+
+
+model = ModelFactory.create(settings.DEFAULT_MODEL)
+
+math_agent = create_react_agent(
+    model=model,
+    tools=[add, multiply],
+    name="math_expert",
+    prompt="You are a math expert. Always use one tool at a time.",
+).with_config(tags=["skip_stream"])
+
+research_agent = create_react_agent(
+    model=model,
+    tools=[DuckDuckGoSearchResults()],
+    name="research_expert",
+    prompt="You are a world class researcher with access to web search. Do not do any math.",
+).with_config(tags=["skip_stream"])
+
+# Create supervisor workflow
+workflow = create_supervisor(
+    [research_agent, math_agent],
+    model=model,
+    prompt=(
+        "You are a team supervisor managing a research expert and a math expert. "
+        "For current events, use research_agent. "
+        "For math problems, use math_agent."
+    ),
+    add_handoff_back_messages=False,
+)
+
+supervisor_agent = Agent(
+    name="supervisor-agent",
+    description="A langgraph supervisor agent",
+    graph=workflow.compile(checkpointer=MemorySaver()),
+)

langgraph_agent_toolkit/agents/components/creators/__init__.py

@@ -0,0 +1,4 @@
+from langgraph_agent_toolkit.agents.components.creators.create_react_agent import create_react_agent
+
+
+__all__ = ["create_react_agent"]