VibeMongo Admin — Dashboard IA para MongoDB
agrhub/vibe-mongo-admin
Admin MongoDB com SRE autônomo e Gemini integrado
Ver no GitHubDashboard moderno de administração MongoDB com IA integrada, construído sobre Vue 3, Express, Google ADK e MCP Server. Oferece monitoramento autônomo de performance, otimização de índices com IA, chat contextual para consultas em linguagem natural e observabilidade em tempo real via Arize Phoenix.
Instalação
npx convez-ai agrhub/vibe-mongo-admin1 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
Casos de uso
- Otimizar queries MongoDB com sugestões de IA baseadas em traces
- Monitorar performance de banco em tempo real e identificar slow queries
- Escrever agregações MongoDB em linguagem natural via chat
- Analisar e debugar operações de banco com IA judge evaluation
Veredito

Ferramental ambicioso que reimagina admin de MongoDB com IA e observabilidade. Excelente para times que precisam otimização autônoma de queries e monitoramento SRE, mas requer infraestrutura Google Cloud e conhecimento técnico substancial.
README
A tradução para o português deste README ainda está na fila.
VibeMongo Admin — Documentation

A modern, AI-first MongoDB administration dashboard powered by a triad of cutting-edge technologies: Google Cloud Agent (ADK & Gemini), the MongoDB MCP Server, and Arize Phoenix Cloud & MCP for autonomous monitoring and LLM evaluation.
📚 Table of Contents
| Document | Description |
|---|---|
| Premium AI SRE Upgrades | [NEW] Deep-dive into Schema Migrator, ERD, Webhooks, Index Sanitizer, and Mock Gen |
| Architecture Overview | System design, component diagram, and data flow |
| Client Guide | Vue 3 frontend — setup, structure, and configuration |
| Backend Guide | Express API + Google ADK Agent — setup, structure, and configuration |
| AI Agent Guide | Google ADK + MongoDB MCP Server integration deep-dive |
| Deploy — Cloud Run | Step-by-step deploy to Google Cloud Run |
| Deploy — Cloud Agent | Deploy the AI Agent as a Google Cloud Agent Engine service |
| Workflow Diagrams | Detailed sequence diagrams: VibeMongo ↔ ADK ↔ MCP ↔ MongoDB |
Inspiration
Managing MongoDB databases using traditional GUI clients often involves writing complex, syntax-heavy aggregation pipelines just to get simple answers. For developers who are not MongoDB experts, writing these pipelines can be a huge time sink. Furthermore, monitoring database performance and identifying slow queries usually requires jumping into completely separate APM (Application Performance Monitoring) tools.
We were inspired to build a modern, web-based admin interface that combines the robust features of traditional database clients with the power of generative AI and real-time observability. Our goal was to create a "Vibe" where managing databases feels effortless, intuitive, and conversational, while also introducing Autonomous Database SRE capabilities.
🚀 Why VibeMongo? (The Autonomous DB SRE)
While traditional tools like MongoDB Compass offer solid administration, VibeMongo rethinks the database management experience by integrating generative AI and real-time observability deeply into the workflow:
- DB-Guardian Copilot (AI-Driven SRE)
- VibeMongo actively monitors query performance in real-time.
- Using the Arize Phoenix MCP, it captures trace data and flags slow queries. The built-in AI agent (Google ADK) can then analyze the trace waterfall and suggest optimizations to turn slow
COLLSCANoperations into efficientIXSCANoperations.
- AI Judge Evaluate (LLM-as-a-judge)
- Click "AI Judge Evaluate" on any trace span to instantly open the Chatbot. The Gemini Agent analyzes the raw input and output, scores its safety and optimality, and provides a human-readable explanation of what the database command actually did.
- Context-Aware AI Chat Sidebar
- Forget writing complex aggregation pipelines manually. Just ask the agent: "Show me the top 5 products by revenue grouped by category in a pie chart."
- The agent securely translates natural language to MongoDB aggregations via the MongoDB MCP and instantly visualizes the result using ECharts—right in your browser.
- One-Click Index Optimization
- Every collection view features an "Optimize Index" magic button that instantly triggers the AI to evaluate your current schema, query usage, and indexes to suggest performance enhancements.
- "Vibe" Aesthetics (Minimalist & Modern)
- Built on Vue 3 and Vite, it drops the clunky Java-like interfaces of legacy tools in favor of a sleek, glassmorphic, responsive, dark-mode native dashboard.
- Full-Stack Feature Parity
- Supports Document CRUD, dynamic BSON parsing, complex schema analysis, mass deletion, multi-database connection management, and comprehensive zip-based Backup & Restore workflows.
🏗️ Architecture & Technology Stack
VibeMongo Admin runs as a unified containerized service hosting both backend API services and frontend assets.
graph TD
subgraph Client [Browser - Vue 3 SPA]
Dashboard[Dashboard & Explorer]
Chat[Agent Chat Sidebar]
Observability[Observability Charts]
end
subgraph Container [Docker Container - Node 20 Server]
Express[Express.js Server]
SQLite[(SQLite DB - Connections)]
NeDB[(NeDB - Telemetry & Stats)]
ADK[Google ADK Agent]
MongoService[MongoService Driver]
end
subgraph External [External APIs & Subprocesses]
Gemini[Gemini API / Vertex AI]
MCP[mongodb-mcp-server subprocess]
MongoDB[(MongoDB Database)]
Phoenix[Arize Phoenix Cloud]
end
Client -- HTTPS & SSE --> Express
Express -- Connection Profile --> SQLite
Express -- Query DB --> MongoService
MongoService --> MongoDB
Express -- Telemetry Data --> NeDB
ADK -- Model Calls --> Gemini
ADK -- Execute Tool --> MCP
MCP --> MongoDB
Express -- Traces/Spans --> Phoenix
ADK -- Telemetry Evaluator --> Phoenix
Express -- Serve Static Assets --> Client

Core Integration (The Tri-Partner Architecture)
- Frontend (UI/UX): Vue 3 (Composition API) with Vite, styled with modern Element Plus and ECharts. We built a dynamic parsing engine that reads specific JSON blocks outputted by the AI agent to render interactive components on the fly.
- Brain (Google Cloud Agent & Gemini): We utilized the Google Agent Development Kit (ADK) and Google's highly capable Gemini 3.1 Flash-Lite model on Vertex AI to act as the central reasoning engine.
- Execution (MongoDB MCP): To bridge the AI agent with MongoDB securely, we integrated the official MongoDB MCP Server (
mongodb-mcp-server). The Agent leverages this standardized Model Context Protocol to introspect schemas and execute aggregation pipelines dynamically based on the current active connection. - Observability & Evaluation (Arize Phoenix Cloud & MCP): We integrated the Arize Phoenix MCP and
@arizeai/phoenix-otelto capture OpenTelemetry traces of both MongoDB operations and LLM reasoning. When the user requests an "AI Judge Evaluation" or "Ask AI to Optimize", the UI fetches the trace data and sends it back to the Gemini Agent to provide human-readable, step-by-step diagnostic feedback directly in the chat interface.
Technology Stack Components
Frontend (Client)
| Component | Technology | Role / Description |
|---|---|---|
| Framework | Vue 3 + TypeScript | Core reactive Single Page Application (SPA) using Composition API |
| Build Tool | Vite | Ultra-fast Hot Module Replacement (HMR) and production build pipeline |
| UI Library | Element Plus | Premium component library for modern, dark-mode native dashboard layout |
| Charts & Dataviz | ECharts, Chart.js, vue-echarts | Interactive query results and performance metrics visualization |
| State Management | Pinia | Modular, type-safe global client state stores |
| Routing | Vue Router 4 | Client-side view routing & navigation guards |
| HTTP Client | Axios | RESTful API communication with the backend service |
| i18n | vue-i18n | Multi-language localization support |
| Styling | Sass (SCSS) | CSS preprocessor for custom glassmorphic styling and smooth variables control |
Backend (Server)
| Component | Technology | Role / Description |
|---|---|---|
| Runtime | Node.js 20+ | Fast, cross-platform JavaScript server runtime |
| Framework | Express 4 | Lightweight REST API endpoint router |
| Language | TypeScript | Strong typing for clean, maintainable backend business logic |
| Database Driver | MongoDB Node.js Driver v6 | Native driver for direct database management and CRUD operations |
| Observability DB | NeDB (@seald-io/nedb) | Lightweight, embedded datastore for telemetry, server stats & monitoring |
| Connection Store | SQLite (sqlite3) | Encrypted, server-side storage of MongoDB connection profiles |
| File Handler | Multer | Multipart form parser for uploading zip-based database backups |
AI Agent & Observability Layer
| Component | Technology | Role / Description |
|---|---|---|
| Agent SDK | @google/adk v1 | Google Agent Development Kit for AI agent orchestration and tool execution |
| LLM Model | Gemini (via Vertex AI / Google AI Studio) | High-performance reasoning engine for natural-language-to-BSON query translation |
| Tool Transport | MCP stdio subprocess | Secure local execution of the mongodb-mcp-server command via standard I/O |
| MCP SDK | @modelcontextprotocol/sdk | Node SDK implementing Model Context Protocol JSON-RPC standard |
| Database Tools | mongodb-mcp-server (npx) | Official MongoDB MCP tools for schema introspection and pipeline execution |
| Telemetry Tools | @arizeai/phoenix-mcp | Trace monitoring integration for real-time trace analysis & DB-Guardian |
| Observability APM | Arize Phoenix Cloud & OpenTelemetry | Telemetry spans capture for LLM evaluations, tracing bottlenecks & query debugging |
📸 Screenshots
Authentication & Connections

Dashboard & Databases

Collections

Documents

Indexes

Backup & Restore

AI Agent (MongoDB MCP)

DB-Guardian (AI-Driven SRE)

AI Judge (Observability & Telemetry)

Observability Dashboard & Telemetry

Personalization (i18n & Theming)

GCP Cloud Run Production Deployment

🏗️ Quick Start (Local Development)
Prerequisites
- Node.js >= 20
- pnpm >= 9
- MongoDB (Atlas or local instance)
- Google Cloud Project with Vertex AI API enabled
1. Clone and install
git clone https://github.com/camtruong512/vibe-mongo-admin.git
cd vibe-mongo-admin
pnpm install
2. Configure environment
cp .env.example .env
# Edit .env with your credentials (see Backend Guide for details)
3. Run development servers
pnpm run dev
# Client → http://localhost:3000
# Backend → http://localhost:4000
🔑 Environment Variables Summary
| Variable | Required | Description |
|---|---|---|
PORT | Yes | Backend server port (default: 4000) |
HOST | Yes | Backend host (default: 0.0.0.0) |
PASSWORD | Yes | Admin dashboard password |
VITE_API_URL | Yes | Frontend → Backend URL |
GOOGLE_CLOUD_PROJECT | Yes | GCP Project ID for Vertex AI |
GOOGLE_CLOUD_LOCATION | Yes | Vertex AI region (e.g. global, us-central1) |
GOOGLE_GENAI_USE_VERTEXAI | Yes | Set to 1 to use Vertex AI |
GOOGLE_API_KEY | Conditional | Google API Key (if not using Vertex AI) |
AGENT_MODEL | Yes | Gemini model name (e.g. gemini-3.1-flash-lite) |
ENCRYPTION_KEY | Yes | 32-char key for encrypting MongoDB connection strings |
NODE_ENV | Yes | development or production |
PHOENIX_PROJECT_NAME | Yes | Arize Phoenix project name for tracing (Get from https://app.phoenix.arize.com) |
PHOENIX_COLLECTOR_ENDPOINT | Yes | Arize Phoenix collector endpoint (E.g. https://app.phoenix.arize.com/s/your_workspace) |
PHOENIX_API_KEY | Yes | Arize API Key for tracing (Get from https://app.phoenix.arize.com/settings?tab=api-keys) |
👤 Developer
- Name: Minh Truong
- GitHub: @camtruong512
👤 Supporter
- Name: Tan Do
- GitHub: @dmtan90
📝 License
This project is licensed under the MIT License.
See the LICENSE file for more information.