Szemat.Pro | DEMO AI Agent Eval Harness

Select your language

Access the demo

I have a demo key Enter your key to start

Request access Request a demo key with your email

Enter your demo key

Request demo access

Email address Full name Organization (optional) What brings you to this demo? What is

When you request a demo key we store the email, name, and organization you submit, to process your request and (if you opt in) to contact you about a pilot. Controller: Waldemar Szemat (Chile). See the Privacy Notice.

I have read the Privacy Notice and consent to processing of the details I submit.

Terms of use and consent

See it answer, then verify

Verify your email

Enter the 6-digit code we sent to {email}.

Verified

Chat inputs are sent to US AI providers each turn and stored only as redacted, fingerprinted records. Names are kept for in-thread personalization and are not redacted. Contact details you submit at sign-up are stored as real personal data. Learn more Opt out

Agent Execution Graph

Loading the agent graph.

Multi-turn medication-adherence agent

Reference implementation. Not a medical device. 100% synthetic data, zero PHI.

The same pattern (eval contract, retrieval with citation, refusal templates, cost gate, tracing) transfers to other regulated industries. See the README for the wider map.

LangGraph 1.x RAG + citation gate 1596+ tests Apache 2.0

Evaluation dimensions 4 deterministic + 3 judge

Citation correctness grounded Citation coverage cited Refusal correctness safe Escalation correctness urgent Groundedness faithful Faithfulness accurate Hallucination honest

Type a message, or pick an example from the top right. ⌘ + Enter sends.

Select your language

Access the demo

Enter your demo key

Request demo access

Terms of use and consent

See it answer, then verify

Verify your email

Verified

Agent Execution Graph

Multi-turn medication-adherence agent

That's your demo session

Run a paid pilot on your own data

How likely are you to recommend this demo? (0–10)