convez.ai
MCP ServerDados & AnalyticsIntermediário
asta-nguyen

KamehaDB

asta-nguyen/kamehadb

GUI de banco de dados com IA que fala SQL na sua linguagem

Ver no GitHub
BomPluginScore · de 0 a 100

MCP server que oferece uma interface gráfica de banco de dados multi-plataforma com suporte a 12+ engines (PostgreSQL, MySQL, MongoDB, Redis, ClickHouse, DuckDB, Qdrant e outros). Integra IA conversacional para gerar queries em linguagem natural, rastreamento de schema com histórico e geração automática de migrações.

Instalação

$npx convez-ai asta-nguyen/kamehadb

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
14
Forks
0
Acompanhando
0
Saúde
57%
Último commit
há 3 semanas
Versão
v2.1.0
Issues abertas
11
Contribuidores
3

Casos de uso

  • Explorar e consultar múltiplos bancos de dados simultaneamente
  • Gerar queries SQL/MongoDB usando chat com IA contextualizada no schema
  • Rastrear mudanças de schema e gerar DDL de migrações automaticamente
  • Monitorar saúde e performance de conexões em tempo real

Veredito

Ferramenta robusta que unifica gestão de múltiplos engines de banco de dados em um único workspace. A integração com IA para geração de queries torna-a especialmente valiosa, mas exige configuração de providers e conhecimento dos engines suportados.

README

A tradução para o português deste README ainda está na fila.

KamehaDB

KamehaDB is a local-first, cross-platform desktop database workspace that connects to PostgreSQL, MySQL, MariaDB, SQLite, SQL Server, Oracle, ClickHouse, DuckDB, MongoDB, Redis, Qdrant, and TigerBeetle — browse schemas, run queries, chat with AI, capture schema history, and generate migrations, all in one Tauri app with a local Node sidecar.

Features

  • 12+ engines, one app — SQL, document, cache, vector, and ledger systems side by side: PostgreSQL, MySQL, MariaDB, SQLite, SQL Server, Oracle, ClickHouse, DuckDB, MongoDB, Redis, Qdrant, TigerBeetle
  • Local-first — All metadata stored locally in a SQLite database; nothing leaves your machine. No hosted proxy, no telemetry
  • AI with real schema context — multi-provider chat (OpenAI, Ollama local + cloud, 9Router) reads your actual DDL, indexes, and constraints to generate SQL, MongoDB pipelines, Redis commands, and Qdrant searches. Streaming responses with stop/cancel
  • Schema timeline + migration assistant — capture on-demand snapshots, see what changed between them, and auto-generate DDL to migrate from one state to the next
  • Global search (Ctrl+K) — fuzzy search across connections, tables, columns, open tabs, and quick actions across every connected engine
  • Engine-native tooling — PostgreSQL stats, MongoDB explorer with chart view, Redis key browser, Qdrant 3D vector map, TigerBeetle accounts/transfers
  • Query history with favorites — normalized patterns, duration per group, per-connection
  • Charts — bar, line, area, pie views on any query result or Mongo collection (powered by recharts)
  • Health monitoring — live status badges with latency tracking; slow/reconnecting/offline states with tooltips
  • Schema browser & ER diagram — auto-generated ER graphs via ReactFlow + dagre
  • Pin & workspace memory — pinned connections and open tabs persisted to localStorage and restored on refresh
  • Schema-aware SQL editor — Monaco with FK-aware JOIN/ON suggestions and autocomplete across tables, columns, functions, keywords
  • Connection URL — paste postgresql://, mysql://, redis://, etc. to auto-fill connection fields

Quick Start

# Install dependencies
pnpm install

# Start dev services
docker compose up -d

# Run the app (sidecar + desktop)
pnpm dev

Then open the desktop app, create a new connection pointing at localhost, and start exploring.

AI Setup

Configure AI providers in the app settings (API Settings page):

ProviderDefault ModelNotes
Ollama (local)llama3.2Uses http://localhost:11434 by default
Ollama (cloud)—Requires base URL and API key
OpenAIgpt-4oRequires API key
9RouterAny modelRequires base URL and API key

Connection defaults (Docker)

EnginePortUserPasswordDatabase
PostgreSQL5432kamehakamehakamehadb
MySQL3306kamehakamehakamehadb
MariaDB3307kamehakamehakamehadb
Redis6379———
SQL Server1433saKameha1!kamehadb
Oracle1521SYSoracleORCLPDB1
ClickHouse8123defaultdefaultkamehadb
DuckDB5432———
TigerBeetle3000———
Qdrant6333———

Project Structure

kamehadb/
├── apps/
│   ├── desktop/          # Tauri v2 + React 19 frontend (Vite, Tailwind v4)
│   └── sidecar/          # Hono HTTP server + database adapters + metadata SQLite
├── packages/
│   ├── shared/           # Shared Zod schemas, app state types, adapter contracts
│   └── ui/               # Shared UI primitives
├── landing/              # Next.js marketing & docs site (managed with npm, separate lockfile)
├── docker-compose.yml    # Dev databases
└── docker-init/          # Seed SQL scripts for dev databases

Running Tests

pnpm test

Contributing

See CONTRIBUTING.md for development setup, project conventions, verification requirements, and pull request guidance.

macOS Gatekeeper

Unsigned macOS builds can still be used for local testing, but Gatekeeper may show messages such as "KamehaDB" is damaged and can't be opened or "KamehaDB" cannot be opened because Apple cannot check it for malicious software.

If you only need to test the app on your own Mac, try one of these:

  1. Right-click KamehaDB.app, choose Open, then confirm.
  2. Remove the quarantine attribute after copying the app to Applications:
xattr -dr com.apple.quarantine /Applications/KamehaDB.app

If the app is still being run directly from the mounted DMG:

xattr -dr com.apple.quarantine "/Volumes/KamehaDB/KamehaDB.app"

You can also disable Gatekeeper globally with sudo spctl --master-disable, but that is broader than necessary and should not be left enabled long-term.

Build for Local Machine

If you see "Target does not exist" error, add the required Rust targets first:

rustup target add aarch64-apple-darwin x86_64-apple-darwin

Then build for your local machine:

# Apple Silicon (M1/M2/M3)
cd apps/desktop && pnpm tauri build --target aarch64-apple-darwin

# Intel Mac
cd apps/desktop && pnpm tauri build --target x86_64-apple-darwin

Linux

# Build for Linux
cd apps/desktop && pnpm tauri build --target x86_64-unknown-linux-gnu

Windows

# Build for Windows
cd apps/desktop && pnpm tauri build --target x86_64-pc-windows-msvc

Tech Stack

  • Desktop: Tauri v2, React 19, Vite, Tailwind CSS v4, Base UI (@base-ui/react), TanStack (Query, Store, Table)
  • Editor: Monaco (via @monaco-editor/react), custom SQL autocomplete with FK-aware JOIN/ON hints, markdown rendering with react-markdown
  • Sidecar: Hono, PostgreSQL (pg), MySQL (mysql2), SQLite (better-sqlite3), MongoDB (mongodb), Redis (ioredis), SQL Server (mssql), Oracle (oracledb), ClickHouse, DuckDB (duckdb), TigerBeetle (tigerbeetle-node)
  • Graph: ReactFlow (@xyflow/react), dagre auto-layout
  • AI: Multi-provider abstraction (OpenAI, Ollama local/cloud, 9Router)

Roadmap

Currently Supported

  • PostgreSQL
  • MySQL
  • SQLite
  • MongoDB
  • Redis
  • Data export (CSV, JSON, SQL)
  • Qdrant (vectordb)
  • SQL Server
  • Oracle
  • ClickHouse
  • DuckDB
  • Tigerbeetle
  • MariaDB

In Progress

  • Query history with favorites
  • Data visualization / charts
  • Connection health monitoring
  • Global search (Ctrl+K)
  • Pin / favorites connections
  • Workspace tabs memory
  • Production safety mode — prevent accidental writes to production
  • Schema change timeline — track column additions, removals, index changes for debugging
  • Migration assistant

Future Ideas

  • Collaboration features

License

Apache-2.0 — see LICENSE.

Ver o original no GitHub

Bom para

Relacionados