databuff

DataBuff — AI-Native APM Built On Opentelemetry

Mission: build the strongest OpenTelemetry APM backend

AI-native · OTLP-native · multi-agent out of the box · self-hosted traces, metrics, topology

Live Demo  ·  Documentation  ·  Open Source APM  ·  简体中文  ·  Community

AI multi-agent troubleshooting · Service health · Call graph topology

Keywords: AI-Native APM · OpenTelemetry APM · OTLP Backend · Multi-Agent · Distributed Tracing · AIOps · Open Source APM · MCP · Self-Hosted Observability

What is DataBuff

DataBuff is an AI-native APM backend built on Opentelemetry — ingest traces, metrics, and logs via OTLP, with full-stack monitoring, service topology, RED metrics, and out-of-the-box multi-agent AI troubleshooting.

The goal is not "yet another observability UI," but to build the strongest OpenTelemetry APM backend — letting LLMs query live telemetry directly, from natural-language query → multi-agent inspection → root-cause analysis → ops remediation.

Listed on OpenTelemetry official Vendors (Native OTLP) and CNCF Landscape.

⭐ Star the project · Live Demo · Documentation

Why DataBuff

If you are evaluating an OpenTelemetry APM backend, or want AI that actually closes the troubleshooting loop, DataBuff is built for you:

• AI-native, not bolted on — LLMs query traces, metrics, topology, and alerts directly; add an API key after install, no separate AI platform needed
• Multi-agent out of the box — AI Brain orchestrates query, inspection, ops, and Q&A experts in parallel
• Fully embrace OpenTelemetry — OTLP-native ingestion with existing instrumentation as-is; SkyWalking compatible for smooth migration
• Clear mission — build the strongest OpenTelemetry APM backend: self-hosted, production-grade, extensible
• Minimal deployment — Ingest + Doris + Web, one Docker command to run

Key Features

🤖 AI-Native

• Not a bolt-on chat box — LLM answers from real telemetry, not hallucination
• Multi-agent collaboration out of the box — AI Brain orchestrates query, inspection, ops, and Q&A experts
• AI application observability (Roadmap) — LLM call chains · token analytics · agent topology · skill/tool/model tracing
• MCP both ways — expose capabilities to Cursor / Claude; ingest external MCPs like Prometheus
• Bring your own model — Kimi, DeepSeek, GLM, Ollama, and other OpenAI-compatible APIs

📊 OpenTelemetry-Native

• OTLP-native ingestion — gRPC 4317 / HTTP 4318 for Traces + Metrics + Logs
• eBPF APM — kernel-level, non-intrusive collection; call chains and performance data without code changes
• Dual-protocol support — OTLP native + SkyWalking native gRPC (11800); switch by changing exporter address
• Alerting loop — threshold detection, scheduled evaluation, alert event history

🐳 Engineering Foundation

• 3-component stack — Ingest + Doris + Web, no middleware sprawl
• Skill extensibility — custom digital experts without touching core code

Feature Gallery

Out-of-the-box multi-agent AI on top of a full OpenTelemetry APM.

Natural Language Query

Ask "which service was slowest" in plain language — AI ranks results, no query language required.

Multi-Agent Collaboration

Complex tasks dispatched to multiple experts in parallel, synthesized into a forwardable incident report.

Root Cause Analysis

Pull topology, rank metrics, attribute bottlenecks by share — conclusions ready for your incident report.

APM UI

Global topology, service list, and trace drill-down — AI reads the data, the UI confirms it.

More capabilities in Documentation

Architecture & Ingestion

Quick Start in 5 Minutes

1. Install platform — one command to launch Ingest + Doris + Web

curl -fsSL https://databuff.ai/databuff/ai-apm-install.sh | bash

2. Install demo (optional) — auto-report traces and see topology quickly

curl -fsSL https://databuff.ai/databuff/ai-apm-demo-install.sh | bash

3. Connect a model → start troubleshooting — open http://YOUR_HOST:27403, login admin / Databuff@123, add API key to enable AI

Offline install / Kubernetes install

Offline install — download the bundle from the install page:

tar -zxvf databuff-ai-apm-offline--.tar.gz
cd databuff-ai-apm-offline-- && sudo ./install.sh

Kubernetes install

curl -fsSL https://databuff.ai/databuff/ai-apm-k8s-install.sh | bash
curl -fsSL https://databuff.ai/databuff/ai-apm-demo-k8s-install.sh | bash  # demo optional

Documentation

Contributing

If you are building AI-native APM / OpenTelemetry observability, help make the strongest OpenTelemetry APM backend a real, usable open-source foundation:

• ⭐ Star this repo and Watch for updates

• 🐛 Open an Issue to report bugs or request features

• 🤝 Read CONTRIBUTING.md and submit a PR

• 💬 Scan the QR code to join our WeChat community

  Scan to join the DataBuff community

License

This repository is licensed under AGPL-3.0.