novafabric.ai/demo

Asset registry

interactive

Every model, agent, prompt, tool, dataset and evaluation suite gets a name@version identity. Promotion toward production is gated by eval results.

Runs in your browser on a fixture from the NovaFabric repository. It reads files; it does not talk to a running server. The real interface is the CLI.

prompt

code-review-prompt@0.2.0

development

Shorter, faster v0.2.0 of the code-review prompt. Failed eval suite — misses edge cases.

Versions:

Spec

{
  "template": "Review this code diff. Note bugs and style issues.\n\n{diff}",
  "variables": [
    "diff"
  ]
}

Eval results

  • edge-case-coverage
    0.62
  • tone-and-clarity
    0.84

Why is this still in development?

The edge-case-coverage suite failed (score 0.62, threshold 0.80). Promotion is blocked until eval passes. v0.1.0 stays in production.

On your machine

nova — bash
$ nova register code-review-prompt-0.2.0.yaml
$ nova eval agent code-review-prompt@0.2.0
$ nova promote direct code-review-prompt@0.2.0 --to production
$ nova list --type prompt