███╗   ██╗███████╗██╗  ██╗████████╗
 ████╗  ██║██╔════╝╚██╗██╔╝╚══██╔══╝
 ██╔██╗ ██║█████╗   ╚███╔╝    ██║
 ██║╚██╗██║██╔══╝   ██╔██╗    ██║
 ██║ ╚████║███████╗██╔╝ ██╗   ██║
 ╚═╝  ╚═══╝╚══════╝╚═╝  ╚═╝   ╚═╝

IA que atende por você

Intelligent WhatsApp secretary — an AI agent on your personal number,
powered by any OpenAI-compatible LLM, with a full web UI for management.

_{Self-hosted · No Business API · Any LLM · Zero build step}

Features · Quick Start · Built-in Tools · Architecture · Development · Tech Stack · License

---

✨ Features

📱 WhatsApp via QR code Uses your personal number, no Business API	🤖 Any OpenAI-compatible LLM OpenAI, Groq, Ollama, Together, LM Studio…
🔧 16 built-in tools Plus custom API tools and MCP protocol	📚 Knowledge base (RAG) FTS5 full-text search + embedding similarity
👥 Up to 10 agents Independent personality, model, provider + chaining	🛡️ Guardrails Whitelist/blacklist, anti-injection, PII filtering
⏳ Message debounce Groups rapid messages before sending to AI	🧠 Session management Auto-summaries prevent hallucination
🖥️ Full web UI Config, conversations, logs, knowledge, reports, chat	🔐 Multi-user auth Admin/user roles, bcrypt, persistent sessions
⏰ Scheduled messages Hourly, daily, weekly, monthly, cron expressions	🔌 MCP server Exposes all tools via SSE protocol
💬 WhatsApp groups Optional, configurable per group	🗄️ External databases Query MySQL and PostgreSQL from the AI

---

🚀 Quick Start

git clone https://github.com/renesul/Next.git
cd Next
make run

1. Open http://localhost:8080
2. Login with admin / admin123
3. Set your AI provider API key and base URL
4. Scan the WhatsApp QR code
5. Done — start chatting

🐳 Docker

docker compose up -d

Or build manually:

docker build -t next .
docker run -d -p 8080:8080 -v next-data:/data next

📋 Requirements

Requirement	Details
Go	1.25+
CGO	Enabled (gcc / build-essential must be installed)
OS	Linux, macOS (Windows via WSL)

⚙️ Configuration

Variable	Default	Description
PORT	8080	HTTP server port
DB_PATH	~/.next/	Data directory (databases, logs)

Both are optional. Set them in .env or as environment variables.

Everything else (AI provider, system prompt, tools, guardrails, agents, etc.) is configured through the web UI and stored in SQLite.

---

🛠️ Built-in Tools

Category	Tool	Description
Time	get_datetime	Current date, time, and day of week
Tasks	create_task	Create a task/reminder for the contact
	list_tasks	List tasks, optionally filtered by status
	complete_task	Mark a task as completed
Knowledge	search_knowledge	Search the knowledge base (RAG)
	save_note	Save persistent notes about a contact
	get_notes	Retrieve saved notes
Web	search_web	Web search via DuckDuckGo
	fetch_url	Fetch and extract text from a URL
	weather	Current weather for any location
	currency	Currency conversion with live rates
Scheduling	schedule_message	Schedule a message with optional recurrence
	list_scheduled	List pending scheduled messages
	cancel_scheduled	Cancel scheduled messages
Data	calculate	Evaluate math expressions
	query_database	Read-only SQL queries (local + external DBs)

You can also add custom API tools (any REST endpoint) and MCP tools (via SSE transport) through the web UI.

---

🏗️ Architecture

Message Flow

flowchart TD
    A["📱 WhatsApp Message"] --> B["🔍 Filter"]
    B --> C["⏳ Debounce"]
    C --> D["🛡️ Guardrails Pre-filter"]
    D --> E["👥 Agent Lookup"]
    E --> F["🧠 Session"]
    F --> G["📚 RAG Search"]
    G --> H["🤖 LLM + Tools"]
    H --> I["🛡️ Guardrails Post-filter"]
    I --> J["📱 WhatsApp Response"]

    style A fill:#25D366,color:#fff
    style J fill:#25D366,color:#fff
    style H fill:#2563eb,color:#fff

Loading

Project Structure

main.go                       Entry point, wiring, signal handling, migrations

app/                          Business logic
  types/types.go              Shared types (Message, Agent, Contact, etc.)
  ai/ai.go                   OpenAI-compatible client (reply, summarize, embed)
  memory/memory.go            Sessions, history, summaries, token budget
  rag/rag.go                  Knowledge base with FTS5 + embeddings
  guardrails/guardrails.go    Pre/post-AI message filtering
  tools/tools.go              Function calling registry, built-in + custom tools
  tools/mcp.go                MCP client + MCP server
  pipeline/pipeline.go        Message processing pipeline

internal/                     Infrastructure
  config/config.go            Config struct + SQLite key-value store
  logger/logger.go            JSON file (always) + SQLite (debug mode)
  auth/auth.go                Multi-user auth (bcrypt, sessions, middleware)
  whatsapp/whatsapp.go        WhatsApp connection (QR, send, receive)
  debounce/debounce.go        Message grouping per contact
  web/web.go                  HTTP routes and API handlers

templates/                    HTML pages (inline CSS/JS, no build step)

---

🧑‍💻 Development

make build               # Build binary (output: next)
make test                # Run all tests
make lint                # Run golangci-lint
make fmt                 # Format code
make check               # deps + fmt + vet + test (full CI check)
make install             # Build and install to ~/.next
make generate            # Run go generate
make run                 # Build and run
make clean               # Remove binary

Databases

Stored in ~/.next/ (or DB_PATH):

File	Purpose
next.db	App data (config, messages, summaries, knowledge, tasks, tools, agents, users, sessions, logs)
whatsapp.db	WhatsApp session store (managed by whatsmeow)

---

📦 Tech Stack

Component	Technology
Language	Go
Database	SQLite with FTS5
WhatsApp	whatsmeow (Web multidevice protocol)
AI Client	go-openai (any OpenAI-compatible API)
MCP	mcp-go (SSE transport)
Auth	bcrypt (golang.org/x/crypto)
External DBs	MySQL (go-sql-driver), PostgreSQL (lib/pq)

---

📄 License

This project is licensed under the MIT License.

---

_{No frameworks, no build steps, no nonsense.}

_{GitHub · MIT License}