VanillaMeta

Lightweight, embeddable open-source dashboards — Apache-2.0, no "Powered by" badge required.

VanillaMeta is an open-source business intelligence web application for building chart widgets and dashboards on top of your databases and files — without the heavyweight footprint of traditional BI platforms.

Why VanillaMeta?

If you need a dashboard engine you can drop into your own product or internal tools without license friction, VanillaMeta is built for that job.

User guide

데이터 연결부터 AI 분석 생성·수정·저장·게시·공유까지는 온라인 사용자 가이드 또는 저장소 가이드를 참고하세요. 처음이라면 샘플로 시작하기를 따라 할 수 있습니다.

검색·기능별 상세 목차를 갖춘 VitePress 문서는 npm --prefix docs/user-manual ci 후 npm run docs:dev로 실행합니다. 전체 목차와 실행 방법을 참고하세요.

Quickstart

Local development and CI use Node.js 24 LTS (nvm use, or install Node 24.11+). The dashboard Assistant uses AI SDK 7 (@ai-sdk/react 4) and the existing Codex CLI connection to gpt-6-luna. Configuration and supported runtime limits are in the Assistant verification guide. Legacy Lambda deployment files are outside the persistent DuckDB worker deployment path.

docker compose up

Then open http://localhost:3000 and sign in:

ID: guest
PW: Admin!@12

That's it — no external database, no env files. The backend runs on embedded SQLite, auto-seeds chart components, and creates a sample dashboard on first signup.

Set your own secrets before exposing it beyond localhost: ACCESS_SECRET, REFRESH_SECRET, JWT_ACCESS_SECRET, URL_ACCESS_SECRET, CONFIG_SECRET_KEY (see backend-api/.env.example).

Local development

# Backend — http://localhost:4000 (SQLite, API prefix /v1)
cd backend-api && npm run start:local

# Frontend — http://localhost:3000
cd frontend-web && npm start

# Seed (idempotent — safe to re-run)
cd backend-api && npm run seed

See AGENTS.md for the full dev guide including tests and E2E.

Features

• No-code chart widgets — 50+ chart types (bar, line, pie, radar, treemap, numeric boards…)
• Dashboards — drag-and-drop layout, template recommendations, sharing & public embed (/share/:uuid, /embed/:uuid)
• SQL editor — query builder with live preview and read-only guardrails
• File analysis — upload CSV/JSON/Parquet, analyze in-browser via DuckDB-WASM or save as a server-side dataset
• AI assistant — natural-language → SQL and chart recommendation (bring your own OpenAI-compatible API key)
• Metrics — reusable semantic metrics (SUM/COUNT/AVG) with evaluation
• Operational touches — query progress over WebSocket, optional Redis result cache, rate limiting, Sentry hooks
• i18n — Korean & English

Data sources

Honest support matrix — 🔷 connectors share the standard knex/driver path but haven't all been exercised end-to-end. PRs for connector verification and fixes are very welcome.

Screenshots

Diverse chart types	SQL editor
No-code widget builder	Template recommendations

MCP for coding agents

The local stdio MCP server connects authenticated coding agents to source/metric discovery, structured AI previews and explicitly approved private artifact saves. Returns Studio links; no arbitrary SQL or automatic publishing. Setup and prompt walkthrough. Codex and Claude Code hosts verified with synthetic local data, separate preview/save approval and Studio reload.

Roadmap

• Dashboard templates & scheduled reports
• Excel / Google Sheets connectors
• Scoped embed tokens & tenant-aware workspaces

License

Apache-2.0 — embed it, brand it, ship it in your product. See LICENSE.

docker compose up