Skip to content
JsonEval
FRExplore the demo
THE PRODUCT JOURNAL

Less uncertainty.
One version at a time.

New features, improvements and fixes in JsonEval. A detailed look at what changes in your evaluation workbench.

The screenshots show the current demo with entirely fictional data. They are not archives of previous interfaces or real performance measurements.
Experience

An evaluation workbench built for developers.

A black-and-white identity, a need expressed in a few words and a demo to explore before creating an account.

  • A black-and-white identity combining JSON braces and evaluation in one mark.
  • A prompt composer with six use-case examples; your requirement is preserved through project creation.
  • An OCR Eval page for invoices, PDFs and scans, accessible from the Products menu. A demo using only fictional data.
  • CLI and AI tools plugin announced in the roadmap, marked as coming soon.
JsonEval demo: field and model JSON scores with fictional data
Explore a field-by-field comparison in the demo. Fictional data.Open the demo
Benchmark

Compare how you send documents, too.

Native PDFs or image pages, a selected provider: model variants enrich the comparison. Cost analytics become more precise.

  • Variantes propres au workspace : pages PDF en images ou fournisseur choisi.
  • Responses missing schema sections identified as retryable failures.
  • Failed-call costs included; historical estimates labelled.
  • Rescoring also includes models hidden from the comparison.
Demo of quality, cost per document and latency analytics for five models
The Cost view helps read the quality / cost trade-off. Fictional data.Open the demo
Data

Fewer clicks to improve your references.

Find values in documents and correct repeated ones in the Reviewer. References remain at the heart of evaluation.

  • Search text values in gold as well as document names.
  • Bulk replacement of a value for the same field, after confirmation.
  • Reviewer navigation by identity and parent-object preview.
  • Fields removed from the schema hidden from review queues without deleting their history.
Fictional invoice and reference values with validation states in JsonEval
Read documents and expected answers side by side. Fictional data.Open the demo
Review

Find the line behind an error.

The Reviewer can locate a line in a digital PDF or scan to help verify the expected value.

  • PDF reading fits the panel width, with identifier-based lookup.
  • On-demand browser text recognition for scans.
  • No highlighting when the line cannot be identified precisely enough.
Documents view of the current demo with a fictional invoice and its references
Illustration: the current demo. PDF lookup is available in the application. Fictional data.Open the demo
YOUR NEXT ITERATION STARTS HERE

Less “looks good”.
More “here are the results”.

Explore a fictional project. Then switch to your own documents.

Try the demo Create my project
No-account demo · Fictional data · No inference