Databuff AI Agent

<div align="center"> <img src="ai-apm-frontend/public/img/logo_login.png" alt="DataBuff" height="56" /> &nbsp;&nbsp; <img src="ai-apm-frontend/public/img/logo_wordmark.svg" alt="Databuff" height="32" /> <h3>DataBuff — AI-Native APM Built On Opentelemetry</h3> <p><strong>Mission: build the strongest OpenTelemetry APM backend</strong></p> <p>AI-native · OTLP-native · multi-agent out of the box · self-hosted traces, metrics, topology</p> <p align="center"> <a href="https://demo.databuff.ai">Live Demo</a> &nbsp;·&nbsp; <a href="docs/README_en.md">Documentation</a> &nbsp;·&nbsp; <a href="https://databuff.ai/en/opensource-apm/">Open Source APM</a> &nbsp;·&nbsp; <a href="README.cn.md">简体中文</a> &nbsp;·&nbsp; <a href="#community">Community</a> </p> <p align="center"> <img src="https://img.shields.io/github/stars/databufflabs/databuff?style=social" /> <img src="https://img.shields.io/badge/License-Apache--2.0-blue.svg" /> <img src="https://img.shields.io/badge/AI--Native-APM-7C3AED?style=flat-square" /> <img src="https://img.shields.io/badge/OpenTelemetry-Native_OTLP-000000?style=flat-square&logo=opentelemetry&logoColor=white" /> <img src="https://img.shields.io/badge/Multi--Agent-Out_of_the_box-2563EB?style=flat-square" /> <img src="https://img.shields.io/badge/Docker-One_command-2496ED?style=flat-square&logo=docker&logoColor=white" /> </p> </div> <p align="center"> <img src="docs/images/databuff-demo-en.gif" alt="DataBuff demo: AI chat, services, topology" width="880" /> </p> <p align="center">_{AI multi-agent troubleshooting · Service health · Call graph topology}</p>

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.

<img src="docs/images/screenshots/aiops-arc-en.svg" alt="AIOps roadmap: Visible → Legion → Inspect → Diagnose → Repair → Predict → Answer" width="900" />

Natural Language Query

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

<img src="docs/images/screenshots/nl-slowest.png" alt="Natural language query" width="720" />

Multi-Agent Collaboration

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

<img src="docs/images/screenshots/multi-agent-process.png" alt="Multi-agent collaboration" width="720" />

Root Cause Analysis

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

<img src="docs/images/screenshots/rca.png" alt="Root cause analysis" width="720" />

APM UI

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

<img src="docs/images/screenshots/global-topology.jpg" alt="Global topology" width="720" />

More capabilities in Documentation

---

Architecture & Ingestion

<img src="docs/images/screenshots/simple-architecture.jpg" alt="Minimal architecture: Ingest + Doris + Web" width="900" />

Protocol	Port / Endpoint	Signals
OTLP (OpenTelemetry native)	gRPC 4317 · HTTP 4318	Traces + Metrics + Logs
SkyWalking native gRPC	gRPC 11800	Trace + JVM metrics + Logs

---

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

<details> <summary><b>Offline install / Kubernetes install</b></summary> <br/>

Offline install — download the bundle from the install page:

tar -zxvf databuff-ai-apm-offline-<version>-<arch>.tar.gz
cd databuff-ai-apm-offline-<version>-<arch> && 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

</details>

---

Documentation

Doc	Description
Documentation	Product overview, user guides, ops reference
OTLP ingestion guide	OpenTelemetry SDK / Collector setup
eBPF ingestion	OBI DaemonSet, zero-instrumentation HTTP / gRPC
Nginx ingestion	ngx_otel_module entry traces
Competitive comparison	vs Jaeger, SigNoz, SkyWalking, and more
Migration guide	Migrate from other APM tools

---

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

<p align="center" id="community"> <img src="docs/images/community.png" alt="Scan to join the DataBuff community on WeChat" width="128" /> <br/> _{Scan to join the DataBuff community} </p>

---

License

This repository is licensed under AGPL-3.0.