Tram¶

A lightweight, extensible pipeline framework for logs, metrics and traces.

Tram is a cloud-native, in-memory data pipeline framework for building real-time data processing flows. It receives OpenTelemetry data, filters, validates and batches it, and routes it through tree-structured pipelines to any number of destinations.

flowchart LR
    otlp([OTLP source]) --> filter[cel_filter] --> batch[batch] --> backend([OTLP sink])
    otlp --> validator[cel_validator] --> nats([NATS sink])
    otlp --> debug([debug sink])

Features¶

Tree-Based Pipelines

Define data flows with a single source, optional stages and any number of sinks as leaves.
Cloud-Native

Configure streams through the HTTP/gRPC API, JSON files or Kubernetes custom resources. Everything runs in memory.
Extensible

Easily add custom sources, stages and sinks.
Real-Time

An event-driven architecture for low-latency data flows.
OTLP-Native

Logs, metrics and traces flow through the pipelines as OpenTelemetry Protocol data.
Observable

Records, durations and request sizes of every node are served as Prometheus metrics.

Quick Start¶

git clone https://codeberg.org/eljoth/tram.git
cd tram
go build -o tram ./cmd/tram
./tram server

The HTTP API then listens on localhost:8080. Create a first stream that logs a generated event every 5 seconds:

curl -X PUT http://localhost:8080/v1/streams/my-pipeline \
  -H "Content-Type: application/json" \
  -d '{
    "stream": {
      "source": { "name": "event-generator", "type": "generator", "spec": { "interval": "5s" } },
      "sinks": [ { "name": "console", "type": "debug", "from": "event-generator" } ]
    }
  }'

Learn More¶

Getting Started

Install Tram, run it and create your first streams
Pipelines

How sources, stages and sinks form a stream
Components

Every source, stage and sink with its configuration
Kubernetes Operator

Run streams described by Stream custom resources
Benchmarks

What pipelines cost, and how to size deployments
Contributing

Development setup, adding components and guidelines