Quickstart

1 Prepare an API key

Dashboard access is coming soon. These examples require an existing API key. For access questions, contact api@nums.world.

Keep your key in your environment; never put a real key in a shared script or notebook.

export CAUSILO_KEY='csl_...'

2 Install the Python client

Use Python 3.11 or later.

pip install causilo-client

3 Send context and query tables

Save the complete example as first_call.py. It sends three context rows and two query rows, with two feature columns. The renewed target belongs only to the context.

import os
import pandas as pd
from causilo_client import Causilo

cx = Causilo("https://api.nums.world", token=os.environ["CAUSILO_KEY"])
context = pd.DataFrame({
    "region": ["EU", "US", "EU"],
    "spend": [120.0, 340.0, 0.0],
    "renewed": ["yes", "yes", "no"],
})
query = pd.DataFrame({
    "region": ["EU", "US"],
    "spend": [200.0, 50.0],
})
prediction, metadata = cx.predict_with_metadata(
    context, query, target="renewed",
    model="causilo-clf", output_type="preds",
)
print(prediction)
print(metadata["quota"])

python first_call.py

4 Read the result

predict_with_metadata returns (prediction, metadata). This example asks for output_type="preds", so classification returns one class label per query row. It prints your live answer; no output values are invented here.

For the established six-row input, see the original first-call example in the Cookbook.