auditar-sdk

Zero-knowledge inference instrumentation for AI decision audit.

pip install auditar-sdk

• Captures inputs/outputs only — never touches model weights
• Works with scikit-learn, XGBoost, LightGBM, and custom APIs
• Decorator API or batch context manager
• Local SQLite queue for air-gapped deployments
• Sub-ms overhead

This SDK is part of the Auditar platform for regulated AI in Indian financial services.

Quick install

pip install auditar-sdk

Quickstart

from auditar_sdk import AuditarClient, trace_inference

client = AuditarClient(
    workspace="retail-lend-prod",
    endpoint="https://core.auditar.io",
    default_use_case="loan",
)

@trace_inference(client, model_name="Credit-Scout-V2", use_case="loan")
def predict(income, credit_score, age, existing_debt):
    return model.predict_proba([[income, credit_score, age, existing_debt]])[0][1]

prob = predict(income=75000, credit_score=680, age=35, existing_debt=12000)

From pip install to pilot

If you want audit-ready traces, RBI FREE-AI 7-Sutras mapping, and a governance dashboard without instrumenting every model yourself:

• Free 8-week pilot: we instrument your production models, you see results in 15 minutes.
• No credit card. No commitment.

Start here: auditar.in

After install, share your source so we can follow up with the right guide:

• GitHub / README: post-install attribution page
• Email outreach: ask your founder for the pilot timeline
• Hacker News / Reddit: community feedback track
• Twitter / LinkedIn: campaign-specific case study

If you came from an email, please reply to the founder who sent it so we can fast-track your pilot.

Email us directly: foundersoffice@auditar.in

License

MIT — see LICENSE.