convez.ai
MCP ServerBusca & PesquisaIntermediário
nomad-actual

Aetherium Browser Search

nomad-actual/aetherium-browser-search

Buscador privado com resumos IA, integrável ao navegador via SearXNG

Ver no GitHub
InicialPluginScore · de 0 a 100

MCP server que integra buscas web descentralizadas via SearXNG com resumos gerados por IA. Funciona como provedor de busca para navegadores, proxiando queries para SearXNG e enriquecendo resultados com análises automáticas de LLMs compatíveis com OpenAI.

Instalação

$npx convez-ai nomad-actual/aetherium-browser-search

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
0
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Integrar buscas descentralizadas como motor padrão do navegador
  • Gerar resumos automáticos de resultados de busca com LLMs locais
  • Hospedar instância privada de busca web com IA na própria infraestrutura

Veredito

Solução bem arquitetada para quem quer controle total sobre buscas e resumos IA. Requer infraestrutura própria (SearXNG + LLM) mas oferece privacidade e flexibilidade genuínas, ideal para selfhosting.

README

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

Aetherium Browser Search

A privacy-focused browser search bar service that proxies queries to SearXNG for decentralized web search results and enhances them with AI-generated overviews from any OpenAI-compatible LLM (llama.cpp, Ollama, vLLM, etc.).

Features

  • Browser Search Bar Integration -- Configure any browser to use your service as a search engine via the OpenSearch / keyword URL format
  • SearXNG Proxy -- Forward queries to a self-hosted or remote SearXNG instance, supporting categories (general, images, news, video, science, etc.)
  • AI Overview -- Automatically generate a concise AI summary from top search results using any OpenAI-compatible LLM endpoint, streamed live via SSE
  • SPA Architecture -- Client-side single-page application bundled with esbuild for fast, reactive rendering
  • Switchable Themes -- Built-in gruvbox, tokyonight, and dark-aero themes with cookie-persisted user preference
  • Collapsible Thinking Block -- View the LLM's reasoning process in a collapsible sidebar panel
  • Docker Ready -- Multi-stage build with health checks, ready to deploy via docker-compose

Architecture

Browser (search provider URL)
    │
    │ GET /search?q=<query>
    ▼
┌─────────────────────────────────┐
│  Fastify Service (TypeScript)   │
│  - SearXNG proxy                │
│  - LLM AI overview              │
│  - SPA HTML shell               │
└─────────┬───────────────┬───────┘
          │               │
          ▼               ▼
    ┌──────────┐   ┌──────────────┐
    │ SearXNG  │   │ OpenAI-Like  │
    │ Instance │   │ LLM API      │
    └──────────┘   └──────────────┘
          │               │
          └───────┬───────┘
                  ▼
         Client SPA (esbuild bundle)
         - Search results rendering
         - Theme switcher
         - SSE streaming client

Quick Start

Prerequisites

  • Node.js 22+ (for local development)
  • Docker & Docker Compose (for production)
  • A running SearXNG instance
  • An OpenAI-compatible LLM API (optional, e.g., llama.cpp, Ollama)

Docker Compose (Recommended)

  1. Copy and customize the environment file:
cp .env.example .env
  1. Edit .env with your SearXNG URL and optional LLM settings:
SEARXNG_URL=http://localhost:8080
LLM_API_URL=http://localhost:8081
LLM_MODEL=llama3.1-8b-instruct
  1. Start the service:
docker compose up -d
  1. Visit http://localhost:3000/health to verify.

Local Development

npm install
cp .env.example .env
# Edit .env with your configuration
npm run dev

Browser Configuration

Chrome / Edge / Brave

  1. Open Settings → Search engine → Manage search engines
  2. Under Site search, click Add
  3. Fill in:
    • Search engine: Aetherium Search
    • Keyword: @aether
    • URL with %s in place of query: https://your-domain.com/search?q=%s
  4. Now typing @aether something in the address bar will search via your service.

Firefox

  1. Open Settings → Search
  2. Scroll to Search Shortcuts, click Add
  3. Fill in:
    • Name: Aetherium Search
    • Shortcut: @aether
    • URL: https://your-domain.com/search?q=%s

Safari

Safari uses your default search engine. Set your default to point to:

https://your-domain.com/search?q=

Configuration

All settings are environment variables (see .env.example):

VariableRequiredDefaultDescription
SEARXNG_URLYes—Base URL of your SearXNG instance
SEARXNG_API_KEYNo—API key if SearXNG requires authentication
LLM_API_URLNo—Base URL of the OpenAI-compatible LLM API
LLM_API_KEYNo—API key for the LLM endpoint
LLM_MODELNollama3.1-8b-instructModel name to use
LLM_MAX_TOKENSNo1024Max tokens for LLM response
LLM_TEMPERATURENo0.7LLM sampling temperature
PORTNo3000Server port
HOSTNo0.0.0.0Server bind address
AI_OVERVIEW_PROMPTNoBuilt-inCustom prompt template for AI overview

AI Overview Prompt

Use {{query}} and {{results}} as placeholders:

AI_OVERVIEW_PROMPT=Synthesize the following search results into a concise answer for: {{query}}

Results:
{{results}}

API Reference

GET /search

Returns an HTML page with search results and AI overview.

Query Parameters:

ParamTypeDescription
qstringSearch query (required)
categorystringSearXNG category: general, images, news, video, science, it, music, files, maps
enginesstringComma-separated SearXNG engines
languagestringLanguage code (e.g., en, en-US)
pagenonumberPage number

Example:

GET https://search.yourdomain.com/search?q=what+is+aetherium&category=general

GET /config

Returns current configuration (API keys masked).

Response:

{
  "searxngUrl": "http://localhost:8080",
  "llmApiUrl": "http://localhost:8081",
  "llmModel": "llama3.1-8b-instruct",
  "llmApiKey": null,
  ...
}

POST /config

Validate configuration changes (requires restart to apply).

GET /health

Health check endpoint.

Response:

{
  "status": "ok",
  "timestamp": "2025-01-15T12:00:00.000Z"
}

Project Structure

├── src/
│   ├── server/
│   │   ├── app.ts            # Fastify server bootstrap
│   │   ├── config.ts         # Environment schema, config loader, AppConfig
│   │   ├── routes.ts         # Fastify route handlers
│   │   ├── searxng.ts        # SearXNG URL builder, result formatting
│   │   └── llm.ts            # OpenAI-compatible LLM call
│   ├── client/
│   │   ├── app.ts            # Client SPA entry point
│   │   ├── search.ts         # Search page rendering, SSE client
│   │   ├── ui.ts             # Theme switching, escapeHtml
│   │   ├── markdown.ts       # markdownToHtml (marked + DOMPurify)
│   │   ├── autocomplete.ts   # Autocomplete dropdown
│   │   └── types.ts          # Client-side SearXNGResult type
│   └── shared/
│       └── types.ts          # Shared TypeScript interfaces
├── public/
│   ├── index.html            # HTML shell
│   └── css/
│       ├── base.css          # Base styles, layout, result styling
│       └── themes.css        # Theme color palettes
├── docker-compose.yml
├── Dockerfile
├── tsconfig.server.json
├── tsconfig.client.json
└── .env.example

License

ISC

Ver o original no GitHub

Bom para

Relacionados