AlphaAvatar WhatsApp Channel

A pluggable WhatsApp channel for AlphaAvatar, built with a driver abstraction architecture to support multiple backend APIs (Baileys / Meta / Twilio).

Built with a Driver Abstraction Architecture for future API replacement.

---

🏗 Architecture

This module is divided into two independent layers:

WhatsApp Backend API
        │
        ▼
Driver Layer (replaceable)
        │  WebSocket
        ▼
Bridge Core (Python)
        │  LiveKit DataChannel
        ▼
AlphaAvatar Agent

Layer Responsibilities

🔌 Driver Layer (Replaceable)

Located at:

drivers/
  baileys/

Responsible for:

• Connecting to WhatsApp backend
• Receiving inbound messages
• Sending outbound messages
• Converting raw events → unified schema
• Communicating with Bridge Core via WebSocket

This layer can be replaced without modifying the Agent or Bridge Core.

---

🧠 Bridge Core (Stable API Layer)

Located at:

src/alphaavatar/channels/whatsapp/

Responsible for:

• WebSocket hub
• Message deduplication
• Session mapping
• LiveKit publish/subscribe
• Routing whatsapp.in / whatsapp.out

This layer does not depend on a specific WhatsApp API provider.

---

📦 Current Driver: Baileys (WhatsApp Web)

Development & local testing only.

When to Use

• Development
• Internal testing
• Personal bot

Not Recommended For

• Production
• Public bots
• High-volume systems

---

🚀 Quick Start

1️⃣ Start Python Bridge Core

From repo root:

uv run alphaavatar-whatsapp-core

Expected output:

WhatsApp Core WS listening on ws://127.0.0.1:18789

Verify port:

ss -lntp | grep 18789

---

2️⃣ Start Baileys Driver

cd drivers/baileys
pnpm install
pnpm dev

Expected logs:

Connected to Core WS

Then a QR code will appear.

---

3️⃣ Link WhatsApp

On your phone:

1. WhatsApp → Settings
2. Linked Devices
3. Link a device
4. Scan terminal QR

Successful login:

WhatsApp connection opened

---

4️⃣ Test Message

Send a message to the linked account.

You should receive:

[echo] your message

If echo works, the channel pipeline is healthy.

---

🔄 Unified Event Schema

All drivers must emit events in the following format.

Inbound Event

{
  "v": 1,
  "channel": "whatsapp",
  "direction": "in",
  "from": "+9715xxxxxxx",
  "chat_id": "wa:xxx",
  "message_id": "xxx",
  "ts": 1700000000,
  "type": "text",
  "text": "hello",
  "meta": { "driver": "baileys" }
}

Outbound Event

{
  "v": 1,
  "channel": "whatsapp",
  "direction": "out",
  "to": "+9715xxxxxxx",
  "chat_id": "wa:xxx",
  "text": "hi!"
}

---

🔌 Driver Abstraction Model

Future drivers should implement:

• Inbound handler → send to Bridge Core
• Outbound handler → send via API
• Reconnect logic
• Auth lifecycle

Recommended structure:

drivers/
  baileys/
  meta/
  twilio/

Only the driver directory changes when switching backend API.

---

🔁 Switching to Official APIs

For production usage, replace Baileys with:

Meta Cloud API

• Webhook-based inbound
• HTTPS outbound
• Business account required

Twilio WhatsApp API

• Twilio webhook
• Twilio REST outbound

What Does NOT Change

• Python Bridge Core
• LiveKit topics
• Agent routing logic
• Memory / RAG / MCP plugins

Only driver implementation changes.

---

🧠 LiveKit Integration

Bridge Core publishes:

• whatsapp.in
• whatsapp.out

Agent subscribes to whatsapp.in Agent replies to whatsapp.out

This decouples channel transport from AI logic.

---

🛠 Production Hardening Checklist

When moving beyond development:

• Persistent deduplication (Redis / DB)
• Per-user message queue
• Rate limiting
• Allowlist / pairing
• Monitoring & health checks
• Containerization
• Replace Baileys with official API

---

📁 Project Structure

avatar-channels/
  avatar-channels-whatsapp/
    drivers/
      baileys/
        src/
    src/
      alphaavatar/channels/whatsapp/
    pyproject.toml
    README.md

---

📌 Design Philosophy

This module is designed with:

• Clean separation of concerns
• Replaceable transport layer
• Stable AI core
• Production upgrade path

It ensures that WhatsApp API provider changes do not cascade into AlphaAvatar core logic.