YouTipLangChain Tool Access — InjectedState and InjectedStore
Sometimes tools need to access more context information, such as the current conversation state, user's persistent data, etc.
LangChain uses a dependency injection mechanism to allow tool functions to automatically obtain this information.
InjectedState — Accessing Agent State in Tools
By default, tools can only receive data from the model through parameters. However, sometimes tools need to know the context of the current conversation — such as previous conversation history, user-confirmed information, etc.
InjectedState allows tools to directly read the complete state of the Agent.
Example
from typing import Annotated, Any
from dotenv import load_dotenv
load_dotenv()
from langchain.tools import tool, InjectedState
from langchain.agents import create_agent
from langchain.chat_models import init_chat_model
from langchain.messages import HumanMessage
@tool
def remember_preference(
preference: str,
state: Annotated[dict[str, Any], InjectedState],
) ->str:
"""Remember the user's preference settings.
Args:
preference: The user's preference content
state: The automatically injected current Agent state
"""
# Get previous message history from state
messages = state.get("messages",[])
message_count =len(messages)
# Can read any field in the state
previous_prefs = state.get("user_preferences","None")
return(
f"Remembered preference: {preference}."
f"(Current conversation has {message_count} messages,"
f"previous preferences: {previous_prefs})"
)
# InjectedState is automatically injected, Agent doesn't need to pass this parameter
result = remember_preference.invoke({
"preference": "Dark theme",
# state doesn't need to be passed, automatically injected by the framework
})
print(result)
Running result:
Remembered preference: Dark theme. (Current conversation has 0 messages, previous preferences: None)Using InjectedState in Agents
Example
from typing import Annotated, Any
from langchain.tools import tool, InjectedState
from langchain.agents import create_agent
from langchain.chat_models import init_chat_model
from langchain.messages import HumanMessage
@tool
def conversation_stats(
state: Annotated[dict[str, Any], InjectedState],
) ->str:
"""Get current conversation statistics, such as message count, conversation length, etc.
No parameters needed, statistics are automatically read from the current state.
"""
messages = state.get("messages",[])
human_msgs =[m for m in messages if m.type=="human"]
ai_msgs =[m for m in messages if m.type=="ai"]
tool_msgs =[m for m in messages if m.type=="tool"]
return(
f"Conversation statistics: {len(messages)} total messages | "
f"User messages: {len(human_msgs)} | "
f"AI responses: {len(ai_msgs)} | "
f"Tool calls: {len(tool_msgs)} times"
)
InjectedState allows you to access all fields in AgentState. If you extend the state_schema (adding custom fields), these fields can also be read by InjectedState.
InjectedStore — Accessing Persistent Storage in Tools
Agent state is conversation-level and disappears when the conversation ends. Store is a cross-session persistent storage that can be used to save user preferences, learning progress, and other long-term information.
InjectedStore allows tools to directly read and write to Store.
Example
from typing import Annotated
from langgraph.store.base import BaseStore
from langgraph.store.mem