Repolode
Submit
deepflowio

deepflowio/deepflow

Zero-code observability for cloud-native and AI applications, using eBPF and Wasm to collect metrics, traces, request logs and profiles.

Open-source alternative to:DatadogDynatrace

View on GitHubWebsite
Stars
4,316
Forks
492
Open issues
276
Last commit
19 hours ago

What is it?

DeepFlow is an observability platform for complex cloud-native and AI applications. It uses eBPF to collect metrics, distributed tracing, request logs and function profiles without requiring developers to modify application code.

Collected data is correlated across the full stack, from application services and AI workloads down to infrastructure such as gateways, databases, message queues, DNS and network interfaces. Tags such as Kubernetes containers, labels and cloud resources are injected into all signals to reduce data silos. Wasm plugins can be used to parse private protocols beyond the common ones.

The project provides SQL, PromQL and OTLP-compatible query interfaces, and can act as a storage backend for Prometheus, OpenTelemetry, SkyWalking and Pyroscope. It consists of an Agent running on each node and a Server that handles management, tag injection, ingest and queries.

Key features

  • 01Zero-code metrics and distributed tracing via eBPF for services in any language.
  • 02Continuous profiling with flame graphs for CPU, off-CPU, GPU, memory and network.
  • 03Wasm plugins for parsing private protocols.
  • 04Automatic tag injection from Kubernetes and cloud resources into all telemetry.
  • 05SQL, PromQL and OTLP query interfaces for popular observability stacks.

Use cases

  • Diagnosing latency across microservices without adding instrumentation to code.
  • Monitoring GPU and CUDA workloads for AI inference or training services.
  • Serving as a long-term storage backend for Prometheus or OpenTelemetry data.

Quick start

  1. 1

    Read the DeepFlow Community deployment documentation and install the all-in-one setup

  2. 2

    Compile deepflow-agent from source if needed, following the build guide

Things to know

  • Self-hosting requires deploying a Server in a Kubernetes cluster and an Agent on each node or host.
  • The Enterprise edition's documentation and demo are currently available only in Chinese.
  • The Cloud SaaS edition is still in beta.

In collections

Run and train AI on GPUsOpenTelemetry & observability platforms

Topics

#observability#ebpf#distributed-tracing#profiling#kubernetes#opentelemetry#apm#gpu#llm#wasm

FAQ

What is deepflow?
Zero-code observability for cloud-native and AI applications, using eBPF and Wasm to collect metrics, traces, request logs and profiles. DeepFlow is an observability platform for complex cloud-native and AI applications. It uses eBPF to collect metrics, distributed tracing, request logs and function profiles without requiring developers to modify application code.
Is deepflow free to use? What license does it have?
deepflow is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is deepflow still maintained?
Actively maintained. The last commit to deepflow was on Nov 6, 2026.
What can I use deepflow for?
Diagnosing latency across microservices without adding instrumentation to code. Monitoring GPU and CUDA workloads for AI inference or training services. Serving as a long-term storage backend for Prometheus or OpenTelemetry data.
How do I install deepflow?
Read the DeepFlow Community deployment documentation and install the all-in-one setup Compile deepflow-agent from source if needed, following the build guide
Where can I get deepflow?
The source code and installation instructions are on GitHub: https://github.com/deepflowio/deepflow

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects