use-lightcurve 0.5.0

Human-grounded end-to-end evaluation for Pipecat voice agents.

Lightcurve Python SDK

Install Lightcurve in a Pipecat voice-agent project:

pip install --upgrade use-lightcurve

Set the workspace API key:

LIGHTCURVE_API_KEY=lc_...

Load Lightcurve once, before constructing the Pipecat pipeline:

import lightcurve.auto

That is the complete agent integration. Lightcurve discovers the Pipecat worker, handles Playground WebRTC audio, captures pipeline events, and receives dataset replay jobs. No phone number, SIP service, room token, runner URL, callback, or provider credential is required.

Dataset automation is also available from Python:

from lightcurve import Lightcurve

lc = Lightcurve()
datasets = lc.list_datasets()
run = lc.run_dataset(dataset_id=datasets["datasets"][0]["id"])