GlassFlow exists to solve one problem: your data should arrive in ClickHouse clean, enriched, and query-ready — no matter where it comes from. Until today, that meant your data had to come from Kafka.

Starting today, it doesn't have to.

With GlassFlow v3.0.0, we're launching native OpenTelemetry (OTLP) ingestion as a first-class source connector in GlassFlow. Every transformation capability you rely on — deduplication, stateful processing, schema mapping — is now available for your OTel traces, logs, and metrics, with the same performance and reliability guarantees you'd expect from any GlassFlow pipeline.

Why OpenTelemetry, why now

OpenTelemetry has become the industry's default instrumentation standard. Most engineering teams building or expanding their observability stack today start from an OTel Collector—not from a Kafka cluster. And as ClickHouse has grown into the analytical backend of choice for observability data, the demand for a reliable transformation layer between the two has grown with it.

The OTel Collector handles collection and export well. ClickHouse handles storage and querying well. But the processing gap in between — deduplication across time windows, span enrichment, attribute normalisation, PII redaction — has been left to custom glue code or bolted-on tooling. That's exactly the gap GlassFlow was built to close, and now it closes it for OTel data too.

GlassFlow's value proposition doesn't change with this release.
We transform data at scale before it hits ClickHouse. The OTel connector simply extends our existing capability to a new and rapidly growing class of data sources.

What you can do with it

The OTel connector puts the full GlassFlow transformation engine to work on your telemetry data:

• Span and event deduplication — eliminate duplicate OTel spans across a configurable time window before they inflate your ClickHouse storage layer

• Filtering and sampling — drop health-check noise, redact PII attributes, and downsample high-volume metric streams before they reach the database

• Automatic schema mapping — OTel semantic conventions, nested resource attributes, and high-cardinality fields are mapped to your ClickHouse schema without manual configuration

What this means for you

For platform and infra engineers, the connector integrates with any standard OTLP setup in minutes — no infrastructure changes are required beyond adding a new exporter to your Collector config. SREs and observability leads get deduplication and downsampling before the data ever reaches ClickHouse, keeping ClickHouse clean and query-ready without sacrificing signal coverage or fidelity. And for teams building AI or LLM applications, GlassFlow can prepare and pre-process GenAI spans before they are enriched with token counts, model cost attribution, and latency buckets as they flow through the pipeline. As a result, everything lands in ClickHouse ready to query with full SQL.

If you're already running GlassFlow pipelines with Apache Kafka, nothing changes. Your existing pipelines are unaffected. If your team runs both Kafka and OTel instrumentation, you can now handle both from a single GlassFlow deployment.

Getting started

The OTel connector is available today across all deployments. Setup takes under five minutes.

1. In the GlassFlow UI, create a new pipeline and select OpenTelemetry as the source.

2. Configure OpenTelemetry Collectors to send data to the GlassFlow OTLP receiver:

HTTP Exporter: 

gRCP Exporter: 

3. Define your transformation logic in the GlassFlow editor — deduplication window, enrichment rules, sampling rate, or PII redaction.

4. Connect your ClickHouse sink and deploy. Data arrives clean, enriched, and schema-mapped from the first event.

Full documentation and a step-by-step getting started guide are available in the docs. Below is an example of how to connect your OTel collector to GlassFlow: 

We will also publish a demo walking through the full OTel → GlassFlow → ClickHouse flow in the next few days for our users to test the new functionality in a real-world scenario.

What's next

The OTel connector marks the beginning of a broader push to support more data sources in GlassFlow natively. We're actively expanding our integrations library this quarter — and we'd rather build what you actually need than guess.

Which connector should we build next?

View the integrations roadmap and submit your request →