LangChain Integration¶

Native LangChain tools that wrap the OpenTime REST API.

Installation¶

pip install opentime[langchain]

Usage¶

from opentime.integrations.langchain import get_opentime_tools

# Get all 8 tools (default: http://localhost:8080)
tools = get_opentime_tools()

# Or with a custom base URL
tools = get_opentime_tools(base_url="http://myserver:8080")

# Add to your agent
from langchain.agents import create_react_agent
agent = create_react_agent(llm, tools)

Requires REST API server

The LangChain tools call the OpenTime REST API over HTTP. Start the server first with opentime-rest or docker compose up -d.

Available Tools¶

Tool Name	Description
`opentime_clock_now`	Get current UTC time
`opentime_task_start`	Start timing a task (returns correlation_id)
`opentime_task_end`	End a task (pass correlation_id)
`opentime_active_tasks`	List in-progress tasks
`opentime_get_stats`	Duration stats for a task type
`opentime_recommend_timeout`	Data-driven timeout recommendation
`opentime_check_timeout`	Check if a task is at risk of timeout
`opentime_compare_approaches`	Compare approaches by actual speed

Each tool has a Pydantic args_schema for structured tool calling, so LangChain agents can discover parameter names and types automatically.

Example: Full Agent¶

from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_core.prompts import ChatPromptTemplate
from opentime.integrations.langchain import get_opentime_tools
from opentime.prompts import get_system_prompt

# Set up the LLM and tools
llm = ChatOpenAI(model="gpt-4o")
tools = get_opentime_tools()

# Include OpenTime instructions in the prompt
prompt = ChatPromptTemplate.from_messages([
    ("system", f"You are a helpful assistant.\n\n{get_system_prompt('function_calling')}"),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}"),
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools)
result = executor.invoke({"input": "Generate a Python function and time yourself"})