Asynchronous Agent Execution
Learn to implement asynchronous patterns for agents to perform parallel tasks and improve responsiveness.
Asynchronous Agent Execution is a free AI Agents with LangChain & Autonomous Workflows lesson on CoddyKit — lesson 2 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the AI Agents with LangChain & Autonomous Workflows learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.
Why Asynchronous Agents?
Imagine your AI agent needs to do several things at once: fetch data from two APIs, analyze text with an LLM, and then store a result. If it does these synchronously (one after another), it waits for each step to complete before starting the next.
This waiting can make your agent slow and unresponsive, especially when dealing with network calls or complex computations.
Synchronous vs. Asynchronous
Synchronous execution is like a single-lane road: only one car can pass at a time. If a car breaks down, all traffic stops.
- Synchronous: Tasks run one by one.
- Asynchronous: Tasks can start, pause while waiting for something (like an API response), and let other tasks run in the meantime. It's like a multi-lane highway or juggling multiple balls.
Asynchronous programming helps agents utilize idle time more effectively.
Python's Async/Await Keywords
Python uses the async and await keywords to enable asynchronous programming. Think of them as signals:
async def: Defines a function (called a coroutine) that can run asynchronously.await: Pauses the current coroutine until the awaited task is complete, allowing other tasks to run.
Let's see a basic example:
import asyncio
async def say_hello():
print("Hello ")
await asyncio.sleep(1) # Simulate a delay
print("World!")
async def main():
await say_hello()
if __name__ == "__main__":
asyncio.run(main())The asyncio Event Loop
Behind the scenes, Python's asyncio library manages how asynchronous tasks run. It uses an event loop.
- The event loop constantly monitors tasks.
- When an
awaitstatement pauses a task, the event loop switches to another ready task. - Once the awaited task is done, the event loop resumes the paused task.
This allows non-blocking operations, improving overall efficiency.
Async LLM Calls in LangChain
Many LangChain components, especially LLMs, offer asynchronous versions of their methods. For example, instead of .invoke(), you can often use .ainvoke() for an asynchronous call.
This is crucial for making your agent responsive when querying large language models, which can take time.
import asyncio
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage
async def main():
# Make sure you have your OpenAI API key set up
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
# Using .ainvoke() for an asynchronous call
response = await llm.ainvoke([HumanMessage(content="What is the capital of Canada?")])
print(response.content)
if __name__ == "__main__":
asyncio.run(main())Running Multiple LLM Calls Concurrently
The real power of async shines when you need to make multiple LLM calls. You don't have to wait for each one to finish before starting the next.
Use asyncio.gather() to run several asynchronous tasks in parallel and collect their results.
import asyncio
from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage
async def main():
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
# Create multiple asynchronous LLM invocation tasks
task1 = llm.ainvoke([HumanMessage(content="Tell me a fact about the sun.")])
task2 = llm.ainvoke([HumanMessage(content="Tell me a fact about the moon.")])
task3 = llm.ainvoke([HumanMessage(content="Tell me a fact about Earth.")])
# Run all tasks concurrently and wait for them to complete
results = await asyncio.gather(task1, task2, task3)
for i, res in enumerate(results):
print(f"Result {i+1}: {res.content}\n")
if __name__ == "__main__":
asyncio.run(main())Asynchronous Tool Execution
Just like LLMs, your custom tools can also be asynchronous! If your tool performs I/O-bound operations (like fetching data from a database or an external API), making it asynchronous will significantly improve agent performance.
To create an async tool, implement the _arun method in your BaseTool subclass.
import asyncio
from langchain.tools import BaseTool
class AsyncWebSearchTool(BaseTool):
name: str = "AsyncWebSearch"
description: str = "Searches the web asynchronously for a query."
async def _arun(self, query: str) -> str:
# Simulate an asynchronous web search API call
await asyncio.sleep(1.5)
return f"Results for '{query}': Found 5 articles."
def _run(self, query: str) -> str:
# Fallback for synchronous calls (optional but good practice)
return f"Sync results for '{query}': Found 4 articles."
async def main():
tool = AsyncWebSearchTool()
result = await tool.arun("latest AI news")
print(result)
if __name__ == "__main__":
asyncio.run(main())Building Async Chains & Agents
When you combine asynchronous LLMs and tools into chains or agents, LangChain automatically leverages their async capabilities. Most LangChain runnables and chains also provide an .ainvoke() method.
This means you can build entire asynchronous workflows, allowing your agent to process complex tasks involving multiple steps and external calls much faster.
import asyncio
from langchain_openai import ChatOpenAI
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnableSequence
async def main():
llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
prompt = ChatPromptTemplate.from_template("What is a unique fact about {animal}?")
# Define a simple chain
chain = prompt | llm
# Invoke the chain asynchronously
response = await chain.ainvoke({"animal": "platypus"})
print(response.content)
if __name__ == "__main__":
asyncio.run(main())Quick Check: Async Benefits
Let's check your understanding of asynchronous execution.
Recap: Mastering Async Agents
Great job! You've learned the fundamentals of asynchronous execution in Python and how to apply it to your LangChain agents.
- Asynchronous programming (
async/await) allows agents to perform tasks concurrently instead of waiting for each one. - This significantly improves responsiveness and throughput for I/O-bound operations.
- LangChain's LLMs, tools, and chains often provide asynchronous methods (e.g.,
.ainvoke(),.arun()) to leverage this power.
By integrating async patterns, you can build more efficient and high-performing autonomous workflows!
Frequently asked questions
Is the “Asynchronous Agent Execution” lesson free?
Yes — the full text of “Asynchronous Agent Execution” is free to read here on the web, and the AI Agents with LangChain & Autonomous Workflows course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the AI Agents with LangChain & Autonomous Workflows course, upgrade to CoddyKit PRO.
What will I learn in “Asynchronous Agent Execution”?
Learn to implement asynchronous patterns for agents to perform parallel tasks and improve responsiveness. You practise AI Agents with LangChain & Autonomous Workflows with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.
Do I need any experience to start AI Agents with LangChain & Autonomous Workflows?
No prior experience is required. AI Agents with LangChain & Autonomous Workflows on CoddyKit is structured for beginners through advanced learners; this is — lesson 2 of 4, so you can start here or from the beginning and move at your own pace.
How long does the “Asynchronous Agent Execution” lesson take?
Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.
Can I write and run code in this AI Agents with LangChain & Autonomous Workflows lesson?
Yes. Every AI Agents with LangChain & Autonomous Workflows lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.
All lessons in this course
- Designing Complex Workflows
- Asynchronous Agent Execution
- Error Handling & Resilience
- Human-in-the-Loop Approvals