@wizeworks/lens-hub

The local hub of WizeLens: one place for the traces and metrics of every service on your machine or team network, with the full dashboard, OTLP intake (traces, logs, and metrics), and a read-only MCP endpoint.

npx @wizeworks/lens-hub
# or
docker run -p 5960:5960 -v wizelens:/data ghcr.io/brandonkorous/wizelens-hub

Open http://localhost:5960/_wizelens/. Then point your services at it and restart them:

WIZELENS_HUB=http://localhost:5960                  # WizeLens add-ons (.NET and Node)
WIZELENS_PROJECT=my-shop                            # optional: group services by project
OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:5960   # any OpenTelemetry app

Options

| Flag | Environment variable | Default | |---|---|---| | --port <n> | WIZELENS_HUB_PORT | 5960 | | --host <addr> | WIZELENS_HUB_HOST | 127.0.0.1 (this machine only) | | --data <file> | WIZELENS_HUB_DATA | ~/.wizelens/hub/hub.db (memory for none) | | --token <text> | WIZELENS_HUB_TOKEN | none | | --retention <time> | WIZELENS_HUB_RETENTION | 7d | | --max-traces <n> | WIZELENS_HUB_MAX_TRACES | 50000 | | --max-metric-points <n> | WIZELENS_HUB_MAX_METRIC_POINTS | 2000000 raw points (they are also kept for --retention; rollups are not capped) | | --mcp | WIZELENS_HUB_MCP=1 | off | | --import <file> | | Copy an add-on's .wizelens/traces.db in at start |

To share the hub with a team, run it with --host 0.0.0.0 --token <text>. Services send WIZELENS_HUB_TOKEN; OTLP apps send OTEL_EXPORTER_OTLP_HEADERS=Authorization=Bearer <text>; in a browser, open /_wizelens/?token=<text> once.

Stats

The dashboard asks the hub for whole pages (GET /_wizelens/api/models/{page}?range=&project=&env=), computed from stats tables in the hub file: per-trace facts and rollups in 10 s, 5 min, and 1 h buckets. Ranges reach the hub's --retention. Traces are settled a few seconds after their last span (every 2 s, at most 150 ms at a time); pages say "Preparing stats" while traces wait. Files from older hubs upgrade when opened and backfill, newest traces first. See docs/CONTRACT.md, "Stats tables".

Metrics

OpenTelemetry apps send metrics to POST /v1/metrics (JSON or protobuf): gauges, sums, histograms, exponential histograms (kept as explicit buckets), and summaries (kept as count and sum). WizeLens add-ons send theirs in their batches. The Metrics page reads GET /_wizelens/api/models/metrics?range=1h&project=&env=&service=&metric=&group=&stat=&q=, computed in the hub from metric rollups (per series, in 10 s, 5 min, and 1 h buckets, written as points arrive), so every range reads all of its points. With --mcp, agents get list_metrics and get_metric.

In code

import { createHub } from "@wizeworks/lens-hub";

const hub = createHub({ port: 5960, storagePath: "./hub.db" });
console.log(await hub.listen());

The wire format is in CONTRACT.md. MIT licensed.