Parallel execution with RunnableParallel in LangChain

Introduction

RunnableParallel helps run multiple tasks at the same time. This makes programs faster by doing many things together instead of one after another.

When you want to ask several questions to different chatbots at once.

When you need to process multiple pieces of data at the same time.

When you want to speed up tasks that don't depend on each other.

When you want to combine results from different tools quickly.

Syntax

LangChain

from langchain.schema.runnable import RunnableParallel

parallel_runner = RunnableParallel({"r1": runnable1, "r2": runnable2, ...})
result = parallel_runner.invoke(input_data)

You create RunnableParallel by giving it a dict of keys to tasks (runnables) to run together.

Calling invoke runs all tasks at once and returns their results as a dict.

Examples

This runs two simple tasks in parallel and gets their outputs as a dict.

LangChain

from langchain.schema.runnable import RunnableParallel, RunnableLambda

r1 = RunnableLambda(lambda x: 'Hello')
r2 = RunnableLambda(lambda x: 'World')
parallel = RunnableParallel({"r1": r1, "r2": r2})
output = parallel.invoke(None)

Runs three tasks at the same time using the same input data.

LangChain

parallel = RunnableParallel({"a": runnableA, "b": runnableB, "c": runnableC})
results = parallel.invoke(input_data)

Sample Program

This example shows how to run two simple tasks at the same time and print their results. Each task just returns a message.

LangChain

from langchain.schema.runnable import RunnableParallel, RunnableLambda

# Define two simple runnables that return fixed strings
r1 = RunnableLambda(lambda x: 'Task 1 done')
r2 = RunnableLambda(lambda x: 'Task 2 done')

# Create a RunnableParallel to run both at the same time
parallel_runner = RunnableParallel({"r1": r1, "r2": r2})

# Run the tasks in parallel
results = parallel_runner.invoke(None)

# Print the results
for i, (key, res) in enumerate(results.items(), 1):
    print(f"Result from task {i}: {res}")

Important Notes

RunnableParallel runs tasks at the same time, so tasks should not depend on each other.

Results come back in the same order as the tasks were given (Python 3.7+ dict insertion order).

Use this to speed up independent tasks and improve performance.

Summary

RunnableParallel runs multiple tasks together to save time.

It returns a dict of results matching the keys of tasks.

Great for tasks that can happen at the same time without waiting.