HYPHAby S33D
Documentation › Observe

Metrics and dashboards

A short query, formulas, functions, and a grid of widgets the whole team shares.

The query language

<aggregation>:<metric>{<filters>} by {<keys>}

Examples:

avg:system.cpu.usage{env:prod} by {host}
max:system.disk.in_use{role:db} by {host,mount}
sum:nginx.requests{*}.rollup(sum, 300)

HYPHA reads raw points up to two days, 5-minute aggregates up to 60 days, then hourly aggregates.

Formulas and functions

Each query has an identifier (a, b…); a formula combines them: a / b * 100. Functions, on a query or a formula:

FamilyFunctions
Arithmeticabs, log2, log10, cumsum, per_second, rate, derivative
Smoothingmoving_avg(x, n), ewma(x, n), median(x, n)
Comparisontimeshift(x, -86400) (same curve one day earlier)
Selectiontop(x, 5, 'mean', 'desc'), clamp_min, clamp_max, fill(x, 'zero')
Analysisforecast(x, 'linear', horizon), anomalies(x, 'agile', 2), outliers(x, 'mad', 3)

The explorer

The Metrics page finds a metric, suggests its tags, plots the query and the formula, and saves the view. Each metric can carry a unit and a description (catalogue provided for system.*, snmp.*, postgresql.*, haproxy.*, nginx.*).

Dashboards

Two dashboards come built in: Vue système (system) and Vue réseau (network), created with French titles you can rename. Yours are laid out on a 12-column grid.

Widgets: time series, query value, top list, log stream, monitor summary, note, host map, check status, heatmap, distribution, table, change, scatter plot, pie chart, event timeline, monitor graph with its thresholds, monitor value, service map, network map, problems, SLO list, image, embedded frame (same origin), group.

Dashboard features:

Ready-made dashboard catalogue

Dashboards › Catalog: about thirty ready-made dashboards, sorted by family, installed in one click into the workspace; an installed dashboard becomes a regular one (editable, clonable, exportable), and installing twice gives two dashboards. Their queries target the metrics the agent publishes: a dashboard stays empty until the integration is active. Dashboard titles are in French for now; any dashboard can be renamed.

FamilyDashboards
Systemsystem view, network view, system in detail, hardware (SMART, sensors), Windows (IIS, SQL Server)
Databases, search, queuesPostgreSQL, MySQL and MariaDB, Redis and Valkey, Memcached, MongoDB, OpenSearch and Elasticsearch, RabbitMQ
Web, mail, network, securitynginx and Apache, HAProxy, mail (Postfix, Dovecot, rspamd, Exchange), BIND DNS, SNMP devices, certificates and access
Containers, virtualization, cloudDocker, Kubernetes, KVM and VirtualBox, Proxmox VE, Hyper-V, VMware vSphere, clouds (AWS, Azure, Google Cloud, Scaleway, OVHcloud)
Generic collectionOpenMetrics targets, node_exporter, Prometheus server, Nagios and Centreon plugins, logs received by syslog (see Generic collection)

API: GET /api/v1/dashboards/catalogue, POST /api/v1/dashboards/catalogue/{id}.

Public read-only sharing

An administrator can publish a dashboard through a secret link (/share/?t=…). The link only runs the queries of the dashboard's widgets, read-only. It can be withdrawn at any time.

Send your metrics

Updated: 20 November 2026