Upgrade Guide

• The website viewer expects nest-graph-inspector to stay on a current release, so upgrading is required for continued use.
• Most updates are intended to be non-breaking; when a breaking change is introduced, the major version is incremented.
• The project is currently in a pivot phase, so updates may be shipped more frequently than usual.

npm update nest-graph-inspector

yarn upgrade nest-graph-inspector

pnpm update nest-graph-inspector

bun update nest-graph-inspector

Unreleased

Graph layout persistence moved to ui.layout.saveAs

layoutFilePath is gone — both the per-output field on a viewer or http output, and the module-wide NestGraphInspectorModuleOptions.layoutFilePath. Layout persistence is now one module-wide setting, ui.layout.saveAs: 'file' (the default) or 'runtime'.

// before
NestGraphInspectorModule.forRoot({
  outputs: [{ type: 'viewer', layoutFilePath: './nest-graph-layout.json' }],
});

// after
NestGraphInspectorModule.forRoot({
  outputs: [{ type: 'viewer' }],
  ui: { layout: { saveAs: 'file' } }, // 'file' is also the default — this line is optional
});

1. Delete layoutFilePath from every output and from your module options. It is no longer a recognized field, so leaving it in place is a type error.

2. In 'file' mode (the default), the saved layout now lives at ./.muse in your project root, rather than ./nest-graph-layout.json — and the path is fixed; it is no longer configurable. If you have an existing nest-graph-layout.json you want to keep, rename it to .muse in your project root after upgrading, so the viewer picks up your saved arrangement instead of starting from an empty layout.

3. If you'd rather not persist to disk at all, set ui: { layout: { saveAs: 'runtime' } } — the layout is then kept purely in memory and resets on every restart. See UI configuration for details.

AI chat moved into your browser

The AI chat no longer talks to Ollama. The model now runs inside the browser on WebGPU, so the library no longer opens a proxy to a model server — that proxy, and the option that configured it, are gone.

1. Delete ollama from your viewer output. The option was removed rather than deprecated, so leaving it in place is a type error.

// before
NestGraphInspectorModule.forRoot({
  outputs: [
    {
      type: 'viewer',
      ollama: { origin: 'http://localhost:11434', path: '/ollama' },
    },
  ],
});

// after
NestGraphInspectorModule.forRoot({
  outputs: [{ type: 'viewer' }],
});

2. Stop running Ollama for the inspector. Nothing connects to it any more, so it can come out of the Docker service, the start script, or the onboarding note that only existed to make the chat work.

3. Open the viewer in a WebGPU-capable desktop browser if you want the chat. Current Chrome and Edge qualify, and so does Safari 18 and later; Firefox needs WebGPU turned on manually. The browser window has to be at least 768px wide — the chat stays off on phones and tablets, and in any narrower window — and your GPU needs enough memory for the model you pick. The chat checks for WebGPU as soon as you open it, before you ask anything, and says what is missing. This applies to the chat only — the graph viewer, Direct Run, and every other output work in any browser they worked in before.

Nothing else about the viewer output changes. The graph endpoint, the access token, and Direct Run are untouched.