Install dashboards and alerts for your infrastructure

An integration is a ready-made observability bundle for a piece of infrastructure you run: a standard dashboard, a set of health alerts, and the configuration needed to collect the data. The catalog covers common databases, caches, queues, streaming and compute systems, web infrastructure, tools that collect metrics, and GPU hardware.

Integrations are in Beta

The integrations catalog is in Beta. Its standard integrations are available in every project, but their dashboards, alerts, and setup instructions may change.

They are built on the metrics the OpenTelemetry Collector already scrapes from those services. The Collector is a separate program that gathers telemetry and forwards it to Logfire. Point it at Redis, open the catalog, and install: you get the dashboard and alerts without working out which attributes the receiver emits, writing the SQL behind each panel, or repeating that work for the next service.

Give every service instance a unique identity

Before using a setup snippet, replace its example identity values with values for your deployment. Follow the OpenTelemetry service semantic conventions when choosing them:

• service.namespace: the system or group containing related services, such as payments-platform or infrastructure.
• service.name: the logical service name shared by equivalent instances, such as redis.
• service.instance.id: one instance of that service, such as a UUID or an unambiguous scrape target like redis-01.example.com:6379.

The combination of all three values must be globally unique for each service instance the Collector scrapes. Keep that identity stable for the lifetime of the instance, and give every independently running instance a different combination. Reusing a complete identity across hosts merges their independent counters into one metric stream, which can make dashboard rates and alerts wrong.

What an integration installs

• A standard dashboard. A curated set of panels for the service. For example, the Redis dashboard covers memory against maxmemory, command throughput, keyspace hit ratio, evictions, connected clients, and replication. Standard dashboards render from one definition Logfire maintains, so installing enables it for your project rather than copying it. There is no per-project copy to drift. Installing the integration is how you enable its service dashboard.
• Alerts. Health alerts grounded in the service’s own operational guidance. For example, the Redis integration alerts on memory near maxmemory, high eviction rate, low keyspace hit ratio, rejected connections, and high memory fragmentation. Every integration also includes a not reporting metrics alert that fires when the service stops sending telemetry.
• Setup instructions. The OpenTelemetry Collector receiver configuration to scrape the service, ready to copy.
• Detection. A check that confirms the service’s metrics are already arriving in your project, so you only install what is relevant.

Which alerts start switched on

Alerts come in two tiers. Recommended alerts install active and begin evaluating right away. Diagnostic alerts, which are noisier or only meaningful for some workloads, install inactive so the default set stays high signal. Switch a diagnostic alert on when you want it.

Alerts are installed without a notification channel, so nothing pages anyone until you attach one.

Install an integration

1. Send telemetry. Point your OpenTelemetry Collector at the service. Each integration’s Set up tab has the configuration to copy. On self-hosted Logfire, substitute your own ingest endpoint.
2. Open the catalog. Open the Integrations catalog for your project.
3. Install in one of two ways:
   • Select Detect and Install at the top of the catalog. Logfire checks every integration against the telemetry in your project and installs the ones it finds.
   • Or find one integration and select Install on its row, or open it with View details and install from there.
4. Attach a notification channel. Review the installed set on the integration’s Alerts tab. Then open your project’s Alerts page and give each alert a channel and schedule so it can reach you.

Verify the installation

The integration’s status changes to Installed. Open View details, then check that its Dashboards tab marks the bundled dashboard as Added and its Alerts tab lists the bundled alerts. When the service’s metrics are arriving, select Open on the dashboard and confirm that its panels contain data for your selected time range.

Integrations stay current automatically

Logfire maintains installed integrations for you. We may revise dashboards and alert queries over time to improve them, and your project receives those changes automatically. Changes you make to an alert’s notification channels or enabled state are preserved.

Uninstall deletes the alerts the integration created and disables its bundled dashboards for the project. Alerts and dashboards you made yourself are left alone.

Read the catalog

The catalog groups the integrations visible to your project by service type. Each row shows the service, its tags, its contents as a count of dashboards and alerts, and its status for your project:

• Available: Logfire has not detected telemetry from this service.
• Detected: the service’s metrics are arriving and it is ready to install.
• Installed: its dashboards are enabled and its alerts exist.

Narrow the list with the search box, the status dropdown (All statuses, Available, Installed), and tag chips such as Database, Cache, Queue, Streaming, Web Server, Compute, Metrics, and GPU. In the dropdown, Available means not installed, so it shows rows with either an Available or Detected status. Selecting several tags matches any of them.

Detection runs on demand

Detection happens when you select Detect and Install. Nothing scans in the background. An integration counts as detected when its metrics (for example, the redis.* metrics) have arrived in your project recently.

Install from an AI assistant

The catalog is also reachable through the Logfire MCP server, so an AI coding assistant or an on-call agent can put monitoring in place for you. integration_list returns the catalog with your project’s install and detection state, and integration_install installs one. This is what an agent reaches for when it needs dashboards and alerts to exist before it can investigate a problem.

Available integrations

Integrations marked Beta are the standard set available in every project. Integrations marked Early access are newer and still being validated against the technology they monitor. Their dashboards and alerts work, but their metric names and thresholds may change. Contact Pydantic if you want to try them.

Integration	Tags	Availability
Redis	Database, Cache	Beta
Memcached	Cache	Beta
PostgreSQL	Database, SQL	Beta
MySQL	Database, SQL	Beta
MongoDB	Database, NoSQL	Beta
Elasticsearch	Database, Search	Beta
RabbitMQ	Queue	Beta
Apache Kafka	Queue, Streaming	Beta
NGINX	Proxy, Web Server	Beta
Apache HTTP Server	Web Server	Beta
ActiveMQ	Queue, Messaging, Java	Early access
Apache Cassandra	Database, NoSQL	Early access
Apache Flink	Compute, Streaming, Big Data	Early access
HAProxy	Proxy, Load Balancer	Early access
NVIDIA GPUs	GPU, CUDA, Kubernetes, Hardware	Early access
Prometheus	Metrics, Collector, Monitoring	Early access
Apache Spark	Compute, Big Data	Early access
Telegraf	Metrics, Collector, Agent	Early access
Apache Tomcat	Web Server, Java	Early access
ZooKeeper	Coordination, Distributed Systems	Early access

Integrations are added over time. Each is a data-only definition, so the catalog grows without new per-service code.