playground / live decisions
Jev Playground
Describe a state, ask a typed question, and inspect the decision returned by your configured provider.
Quick Guide: What is the Jev Playground?
The Jev Playground is an interactive, zero-setup web workbench for evaluating typed decisions. Without configuring local developer dependencies or attaching billing keys, you can supply context state, edit decision criteria, and immediately observe raw categorical choice outcomes, probability distribution bars, and latency metrics in real time.
Route this customer support ticket to the best team.
Output
shipping
Sample output for orientation. Run the playground for a provider response.
Three-Step Live Evaluation Workflow
Verify your business logic assumptions through this interactive loop:
Provide Context State
Enter an inbound customer inquiry, support ticket, or forum snippet. Jev treats this text as immutable context state.
Define Criteria Guidelines
Select your output format, adjust categorical keys (e.g. shipping, refund), and customize clear criteria definitions.
Execute & Inspect Probabilities
Trigger the decision to review the chosen winner, full probability distributions (e.g. 91% vs 6%), model identifiers, and round-trip latency.
Frequently Asked Questions (FAQ)
Do Playground results mirror production API responses?
Yes. The web playground dispatches requests directly to active Jev inference endpoints. The underlying weights, probability calibrations, and token costs are identical to production workloads.
What do the probability bars signify in software practice?
The bars render normalized model probabilities across all candidate options. A score exceeding 0.90 denotes decisive certainty, while close splits (e.g., 45% vs 45%) indicate edge-case ambiguity that should trigger fallback routes in production.
Is it safe to test proprietary data in this public tool?
While communications are encrypted via HTTPS, do not paste unredacted passwords, payment cards, or sensitive personal identifiable information (PII). Reserve confidential data for private server-side pipelines.
Are there throughput caps or rate limits on the playground?
Yes. To prevent abuse and protect infrastructure for all builders, the playground applies rate-limiting and Cloudflare Turnstile verification. Standard manual testing operates smoothly without friction.
How do I export calibrated criteria into production code?
Switch to the Schema Builder view or copy the compiled JSON payload. The criteria structure pastes directly into your TypeScript route handlers or Python workers.