whatsapp-ai-pix-agent

KO | EN

gitlite — search

#python #java #python3 #arduino #golang #machine-learning #rust #html #flask #javascript #seismology #nodejs

whatsapp-ai-pix-agent

★ 17 Open GitHub ↗

LangGraph state machine that routes WhatsApp messages through intent classification and triggers Pix charges via AbacatePay.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

<h1 align="center">WhatsApp LangChain</h1>

Follow me on Instagram:
 <a href="https://instagram.com/mjuniordev" target="_blank">
 @Mjuniordev
 </a>

WhatsApp · LangChain · LangGraph · OpenRouter · Evolution API · Abacate Pay

A personal WhatsApp automation built to respond to my wife with a little AI-powered negotiation. It classifies her messages, decides whether the request is a favor, an order, or just conversation, calculates a playful cost, creates a PIX QR code, and sends the response back through WhatsApp.

## What It Does

This project receives WhatsApp webhook events, classifies the message intent as `favor`, `ordem`, or `no request`, and uses a LangGraph workflow to decide what happens next.

When the message is actionable, the graph:

- identifies the intent with OpenRouter
- routes the request through the proper graph node
- calculates the request cost
- creates a PIX QR code with Abacate Pay
- sends text and QR code media through Evolution API
- blocks responses from numbers outside the allowlist

## Workflow

```mermaid
flowchart LR
  Start([Webhook]) --> Intent[Identify Intent]
  Intent -->|favor| Favor[Favor Node]
  Intent -->|ordem| Ordem[Ordem Node]
  Intent -->|no request| Response[Response Node]
  Favor --> Pay[Abacate Pay]
  Ordem --> Pay
  Pay --> Response
  Response --> End([WhatsApp Reply])
```

## Tech Stack

- Node.js 24
- TypeScript
- Fastify
- LangChain
- LangGraph
- OpenRouter
- Evolution API
- Abacate Pay

## Getting Started

Install dependencies:

```bash
npm install
```

Create your local environment file:

```bash
cp .env.example .env
```

Fill in the real values in `.env`, then start the server:

```bash
npm run dev
```

The API starts on port `3000`.

## Webhook Example

Send a WhatsApp webhook event to the local server:

```bash
curl -X POST http://localhost:3000/webhook \
  -H "Content-Type: application/json" \
  -d '{
    "data": {
      "key": {
        "remoteJid": "5511999999999@s.whatsapp.net"
      },
      "message": {
        "conversation": "Amor pega agua pra mim"
      }
    }
  }'
```

The server reads `data.key.remoteJid` to check the allowlist and `data.message.conversation` as the message sent to the LangGraph workflow.

## Environment

The project expects these values:

```bash
LANGSMITH_API_KEY=
LANGSMITH_TRACING=
LANGSMITH_PROJECT=
OPENROUTER_API_KEY=
ABACATE_PAY_API=
ABACATE_CUSTOMER_NAME=
ABACATE_CUSTOMER_CELLPHONE=
ABACATE_CUSTOMER_EMAIL=
ABACATE_CUSTOMER_TAX_ID=
EVOLUTION_API_HOST=
EVOLUTION_INSTANCE=
EVOLUTION_API_KEY=
ALLOWED_NUMBERS=
```

`ALLOWED_NUMBERS` should be a comma-separated list of phone numbers. Non-numeric characters are ignored by the app before matching.

## Scripts

```bash
npm run dev              # Start the Fastify server with .env loaded
npm test                 # Run tests
npm run langgraph:serve  # Open the LangGraph development server
```

## Project Structure

```text
src/
  config.ts
  index.ts
  server.ts
  graph/
    graph.ts
    nodes/
      identifyIntentNode.ts
      favorNode.ts
      abacateQrCodeNode.ts
      chatResponseNode.ts
  services/
    openrouter.service.ts
prints/
  slide_1.png
  slide_2.png
  slide_3.png
  lang_graph.png
  studio_example.png
```

whatsapp-ai-pix-agent

Explore Similar Repositories

Was this content helpful?

Recent Feedback

Download README