DataBuff v0.1.4 is out: configurable seven AI ops entry points, multi-agent collaboration and deep inspection reports, plus SkyWalking/Trace data fidelity and Doris/upgrade hardening. Relative to v0.1.3: 45 commits / 274 files.

Highlights

• Seven AI ops entry points: See, Squad, Inspect, Diagnose, Repair, Predict, Answer — names, suggested prompts, and expert bindings are configurable.
• Multi-agent collaboration: parallel across experts, serial within one expert; HTML / Markdown / SVG reports previewable in the workspace.
• Deep service inspection: synthesizes metrics, logs, alerts, dependencies, error traces, instances, and JVM signals into an evidence-backed HTML report.
• Product Q&A expert: answers deploy/ingest/config questions from in-image docs and code.
• SkyWalking MQ virtual services: MQ producers normalized; RPC/entry stats and Trace time alignment fixed.
• Doris runtime recovery: AI troubleshooting entry stays available during outages; auto-hydrate after recovery without restarting Web.
• Upgrade compatibility: legacy Docker Compose auto-selected; backs up data/ before upgrade with rollback on failure.

Seven AI ops entry points

Go to AI Platform → AI Chat. Pick a capability from the home surface; configure names, prompts, and expert bindings in the admin page.

Fig. 1 · AI chat home with seven configurable entry points

Multi-agent collaboration

Hand complex tasks to the AI brain — it dispatches specialists in parallel (e.g. data queries + health inspection) and merges an actionable report.

Fig. 2 · Parallel dispatchExpertTask calls to data and inspection experts

Fig. 3 · P0/P1 summary with evidence chain, previewable in chat

Deep service inspection

Go to AI Platform → Smart Inspection. One sentence triggers a full HTML report — entry metrics, error logs, downstream health, and graded conclusions.

Fig. 4 · Trigger inspection and request an HTML report

Fig. 5 · HTML report: health scores, entry metrics, errors, downstream deps

Product Q&A expert

Deployment, ingest, port, and alert questions are answered from in-image DataBuff docs and code — not model memory alone.

Fig. 6 · Product Q&A expert grounded in in-image docs

Data fidelity: MQ virtual services & topology

v0.1.4 normalizes SkyWalking/OTel MQ producers into virtual services and fixes RPC CLIENT/outbound peer mapping, entry stats, and Trace end-time alignment.

Fig. 7 · MQ page: producers and consumers tracked separately

Fig. 8 · Global topology: checkout dependencies at a glance

Fig. 9 · Service flow: per-downstream contribution for RCA

Install & upgrade

New install: Docker installation. Upgrade an existing deployment: Upgrade and uninstall.

Try it

GitHub Release v0.1.4 · Live demo · Star on GitHub