FAQ

Answers to the questions we hear most often.

Do you store our users’ conversations?

No. We hash content at capture and only use behavioral and structural signals (tool use, latency, outcomes). We cannot reconstruct what users said or what the model replied. See Privacy & data architecture and GDPR overview.

Where does data go?

With the open-source SDK and local SQLite, data stays on your machine. If you use our cloud API, only non-content signals are sent. With self-hosting, nothing leaves your VPC. See Self-hosting guide.

How much does it cost?

The core SDK is free and open source (Apache 2.0). Cloud dashboard, Slack alerts, and team features will be paid; pricing TBD. We’ll announce on the site and in the changelog.

Which frameworks do you support?

LangChain, LangGraph, AutoGen, CrewAI, and raw OpenAI (via HTTP proxy). See Framework integration guides.

How do I interpret the drift score?

0 = no change, 1 = max change. We often recommend alerting or gating around 0.2–0.3. See Drift score reference.

Can we use this for EU AI Act / DORA / GDPR?

We’re built to support documentation and transparency requirements without storing raw content. See EU AI Act, DORA guide, and GDPR overview. Have your legal or compliance team confirm for your use case.

More questions? Email hello@driftbase.com.