Async Evaluation

aevaluate()

::: ragas.evaluation.aevaluate

Async Usage

Ragas provides both synchronous and asynchronous evaluation APIs to accommodate different use cases:

Using aevaluate() (Recommended for Production)

For production async applications, use aevaluate() to avoid event loop conflicts:

import asyncio
from ragas import aevaluate

async def evaluate_app():
    result = await aevaluate(dataset, metrics)
    return result

# In your async application
result = await evaluate_app()

Using evaluate() with Async Control

For backward compatibility and Jupyter notebook usage, evaluate() provides optional control over nest_asyncio:

# Default behavior (Jupyter-compatible)
result = evaluate(dataset, metrics)  # allow_nest_asyncio=True

# Production-safe (avoids event loop patching)
result = evaluate(dataset, metrics, allow_nest_asyncio=False)

Migration from nest_asyncio Issues

If you're experiencing issues with nest_asyncio in production:

Before (problematic):

# This may cause event loop conflicts
result = evaluate(dataset, metrics)

After (fixed):

# Option 1: Use async API
result = await aevaluate(dataset, metrics)

# Option 2: Disable nest_asyncio
result = evaluate(dataset, metrics, allow_nest_asyncio=False)

1.3 KiB Raw Permalink Blame History