Telemetry
To help us improve Weaviate and understand community usage trends, Weaviate collects telemetry data by default. This includes basic information like your server version, operating system, the number of objects in your instance, cluster topology, which client libraries talk to it, and cloud metadata. We may use this data to identify potential customers for our commercial services and may share it with third parties for this purpose. For full details, see our privacy policy.
We do not collect your data objects, collection names, property names, tenant names, or any content from your datasets. Only the information listed below is collected.
When data is sent
Each Weaviate node sends its own telemetry payload. A node sends a payload:
- once on startup (payload type
INIT), - every 24 hours while running (payload type
UPDATE), and - once on graceful shutdown (payload type
TERMINATE).
The startup payload does not include the object count, because shards may not be loaded yet. Client usage counters (see below) are reset after every payload, so each payload covers only the period since the previous one.
Payloads are sent as an HTTPS POST to https://telemetry.weaviate.io/weaviate-telemetry.
Data collected
The table lists every field in the payload and the Weaviate version that introduced it. Fields are only ever added; older servers send a subset.
| Field | Description | Since |
|---|---|---|
machineId | A random UUID generated when the process starts. It changes on every restart. | — |
type | INIT, UPDATE, or TERMINATE (see above). | — |
version | The Weaviate server version. | — |
os, arch | Host operating system and CPU architecture, as reported by the Go runtime (for example linux and amd64). | — |
usedModules | Names of the modules referenced in your collection definitions (vectorizers, generative, reranker, and so on). For Google modules, the name also indicates whether Vertex AI or AI Studio is used. Module configuration is not sent. | — |
objs | The number of objects on this node. | — |
collectionsCount | The number of collections in the schema. | — |
clientUsage | The number of requests received in the period, grouped by client library and client library version. The library is identified from the X-Weaviate-Client header that the official Weaviate clients send. Requests without that header are not counted. | v1.36 |
clientIntegrationUsage | The number of requests received in the period, grouped by integration name and version, taken from the X-Weaviate-Client-Integration header. Integrations such as framework connectors set this header to identify themselves. Requests without that header are not counted. | v1.38 |
cloudProvider, uniqueID | The cloud provider (AWS, GCP, or Azure) and the account identifier for that provider. See Cloud metadata for details. | — |
nodeId | The name of this node, which is the value of CLUSTER_HOSTNAME. If CLUSTER_HOSTNAME is not set, Weaviate uses the host's operating-system hostname. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
clusterId | A random UUID generated once per cluster and stored in the cluster's Raft state. Unlike machineId, it is the same on every node and persists across restarts. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
nodeCount | The number of nodes in the cluster. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
maxReplicationFactor, replicationEnabled | The highest replication factor of any collection, and whether it is greater than one. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
mtCollectionCount | The number of collections with multi-tenancy enabled. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
namedVectorCollectionCount | The number of collections that use named vectors. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
asyncIndexingEnabled | Whether ASYNC_INDEXING is on. | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
vectorIndexTypeCounts | The number of vector indexes of each type (for example hnsw, flat, dynamic). | v1.36.24, v1.37.14, v1.38.7, v1.39.0 |
Each payload is also logged as a summary at info level with action=telemetry_push.
Other requests to weaviate.io
When telemetry is enabled, Weaviate also makes these requests. They stop when telemetry is disabled.
- Startup banner. Since
v1.39.4, each node fetches the art for the startup banner fromhttps://weaviate.io/banner/v1.jsonon startup and again everyBANNER_INTERVAL. The request carries the Weaviate version in itsUser-Agentheader. You can keep telemetry on but turn off the banner withDISABLE_STARTUP_BANNER. - Cluster ID in documentation links. Since
v1.39.4, links todocs.weaviate.iothat Weaviate writes into log lines, error responses, and the startup banner carry theclusterIdas a?clusterid=query parameter. Following such a link sends the cluster ID to the docs site. Nothing is sent unless the link is opened.
Disabling telemetry data
To disable telemetry, add this line to your system configuration file:
DISABLE_TELEMETRY=true
This stops the telemetry payloads, the cloud metadata lookup, the banner fetch, and the cluster ID in documentation links. Set it on every node.
Two further variables adjust telemetry without disabling it: TELEMETRY_PUSH_INTERVAL changes how often UPDATE payloads are sent, and TELEMETRY_URL redirects payloads to your own endpoint, for example to inspect them.
Cloud metadata
When available, Weaviate collects basic metadata from metadata endpoints it has access to on selected cloud providers (Google Cloud, Amazon Web Services, Azure). On startup, Weaviate probes each provider's metadata endpoint with a one-second timeout and logs a notice when one responds.
More information about the metadata endpoints:
The cloud metadata contains:
- AWS: AWS Account ID
- GCP: Project ID
- Azure: Subscription ID
Questions and feedback
Have a question or feedback? Here's how to reach us.
