@nodellmcache/dashboard

An opt-in, real-time metrics dashboard for NodeLLMCache. A tiny Express server serves a self-contained, dark, futuristic UI that streams live cache metrics over Server-Sent Events from @nodellmcache/observability.

No build step, no external assets, no CDN: the whole UI is inline HTML/CSS/canvas.

Install

npm install @nodellmcache/dashboard @nodellmcache/observability @nodellmcache/core

Quick start

import { startDashboard } from '@nodellmcache/dashboard'
import { observability } from '@nodellmcache/observability'
import { PromptCache } from '@nodellmcache/prompt-cache'
import { MemoryAdapter } from '@nodellmcache/memory'

// Point your caches at the shared collector...
const cache = new PromptCache({ adapter: new MemoryAdapter(), metrics: observability })

// ...then open the dashboard.
const dash = await startDashboard() // http://127.0.0.1:4242
// later: await dash.close()

By default it reads the shared observability singleton, so any cache wired with metrics: observability shows up automatically.

What you get

• Hit rate ring + live line chart over time
• Tokens saved and estimated USD saved
• Latency (avg / p99) and compression ratio
• Embeddings reused
• Per-cache-type breakdown

Endpoints

| Route | Description | |-------|-------------| | GET / | The dashboard UI | | GET /api/snapshot | Current metrics snapshot (JSON) | | GET /api/compression | Aggregate compression stats (JSON) | | GET /api/stream | Server-Sent Events stream of snapshots |

Options

| Option | Default | Description | |--------|---------|-------------| | port | 4242 | Listen port (0 for an ephemeral port) | | host | 127.0.0.1 | Bind interface | | collector | observability | A MetricsCollector to read from | | pollIntervalMs | 1000 | How often the stream pushes a fresh snapshot |

startDashboard() resolves to a handle: { server, port, url, close() }.

License

MIT