Capability Manifest Generators

This page provides purely client-side low-code generators for conformance manifest scaffolding. They remove the friction of hand-writing JSON, but they do not decide which capabilities an implementation or adapter is allowed to claim publicly.

Protocol capability manifest

Use this generator for SDK repositories. Choose a protocol line, enter your implementation name, tick the protocol capabilities you have completed, and copy the generated JSON. After that, review the Capability Catalog for token semantics and combination rules.

The catalog is built from versioned protocol and case manifests. It emits only the SDK-owned capability manifest shape: implementation_name, protocol_version, and supports. It does not generate adapter plans, benchmark plans, or case definitions.

Low-Code Capability Manifest Generator

Choose a protocol line, enter your implementation identifier, and tick completed capabilities to generate a copy-ready capability manifest JSON.

The generator only assembles JSON. It does not decide which tokens you should claim publicly; the versioned capability catalog and the conformance baseline remain the source of truth.

Loading the current capability baseline...

Generated Output

{
  "implementation_name": "",
  "protocol_version": "",
  "supports": [],
  "$schema": "../../schemas/capability-manifest.schema.json"
}

OpenAI API profile capability manifest

Use this generator for adapters that expose an OpenAI-compatible NNRP API surface, such as the vLLM NNRP adapter. It generates the adapter-owned manifest used by the API profile runner: adapter, profile, schema_version, compatibility_levels, operations, and extensions.

The catalog is built from profiles/openai-compatible/1/manifest.json and its declarative recipe manifests. The generated file is used together with the profile suite manifest to select recipe-level tests.

OpenAI API Profile Capability Generator

Generate the adapter-owned declaration used to select declarative recipes for the OpenAI-compatible NNRP API profile.

Loading API profile baseline...

Generated Output

{
  "adapter": "",
  "profile": "",
  "schema_version": "",
  "compatibility_levels": [],
  "operations": [],
  "extensions": [],
  "$schema": "../../schemas/api-profile-capabilities.schema.json"
}

Wire-level conformance manifests

Use this generator when the conformance runner must test a real endpoint directly instead of calling an SDK-owned adapter. The Target Manifest output declares runner modes, frame transport endpoints, host-route providers, selected wire capabilities, and execution limits. The Host Route Scenarios output copies the frozen suite fixtures and expected evidence into a scenario manifest.

In a host-route fixture, application_endpoint remains an nnrp:// or nnrps:// endpoint while each provider keeps its own route-local locator. Security declarations record the mode and credential owner only; generated scenarios never contain certificate, private-key, or secret bytes.

This is intentionally separate from the SDK capability manifest and the OpenAI API profile manifest. Adapter manifests are useful for recipe selection, but wire-level conformance must be able to act as a client, server, or proxy and exchange protocol frames directly.

Wire-level Conformance Manifest Generator

Generate target declarations and host-route scenarios for frame-level E2E tests where the runner directly acts as client, server, or proxy.

Loading wire-level baseline...