PHP-Python Prototype

A prototype demonstrating a Symfony application using a PHP library to communicate with LLM agents via WebSocket and JSON-RPC.

Architecture

┌─────────────────┐    Library     ┌──────────────────┐    WebSocket     ┌─────────────────┐
│  Symfony App    │ ←──────────→   │   PHP Library    │ ←─────────────→  │      Python     │
│   (Web Demo)    │   Integration  │ (Communication)  │   JSON-RPC       │   (LLM Logic)   │
└─────────────────┘                └──────────────────┘                  └─────────────────┘

• Symfony App: Web application with FrankenPHP and chatbot interface
• PHP-Python Library: Library for communicating with LLM services
• Python LLM Agent: Simulates an LLM agent that makes function calls
• PostgreSQL: Database for persistence

Prerequisites

Option 1: Docker (Recommended)

• Docker and Docker Compose v2

Option 2: Local Development

• PHP 8.4+ with Composer
• Python 3.11+ with aiohttp
• PostgreSQL 18 (optional)

Project Structure

php_python_proto/
├── php-python-lib/         # PHP library for LLM agent communication
├── symfony-app/            # Symfony 7.3 application with FrankenPHP
│   └── frankenphp/         # FrankenPHP configuration (Caddyfile, entrypoint)
├── agent-python/           # Python LLM agent server
└── compose.yaml            # Docker Compose configuration

Setup Instructions

Option 1: Docker Setup (Recommended)

1. Copy environment file

cp .env.example .env

2. Start all services

docker compose up -d

3. Access the application

• Symfony app: http://localhost (HTTP) or https://localhost (HTTPS)
• Python agent: ws://localhost:9000/ws
• PostgreSQL: localhost:5432
• Caddy metrics: http://localhost:2019/metrics

4. View logs

docker compose logs -f symfony-app

5. Stop services

docker compose down

Option 2: Local Development Setup

1. Install PHP dependencies

cd php-python-lib
composer install

cd symfony-app
composer install

2. Install Python dependencies

pip install aiohttp

Running the prototype

With Docker (Recommended)

# Start in detached mode
docker compose up -d

# Or with logs
docker compose up

Without Docker

Terminal 1: Start Python Server

cd agent-python
python agent_server.py

Terminal 2: Start Symfony Application

cd symfony-app
php -S localhost:8000 -t public

Testing the Application

1. Open your browser and go to: http://localhost (Docker) or http://localhost:8000 (local)
2. Enter your name
3. Send messages to the LLM agent (try words or phrases like "hello", "test message", etc.)
4. Watch the agent analyze your text using the available functions:
   • getStringLength: Gets the length of text
   • countWords: Counts words in text
   • reverseString: Reverses the text

Docker Stack

• FrankenPHP: Modern PHP application server with HTTP/2, HTTP/3, and automatic HTTPS
• PostgreSQL 18: Database server
• Python 3.11: LLM agent server with aiohttp
• Caddy: Built-in web server (via FrankenPHP)

Features

• Hot-reload for development
• Multi-stage Docker builds (dev/prod)
• Health checks for all services
• Persistent database volume
• Automatic HTTPS certificates

How It Works

1. PHP Library (php-python-lib)

The library provides an AgentClient class that:

• Connects to the Python WebSocket server
• Registers PHP functions that the agent can call
• Handles JSON-RPC communication
• Returns raw message exchange data for debugging

2. Symfony Application

• Provides a simple chatbot web interface
• Runs on FrankenPHP with worker mode for optimal performance
• Uses the PHP library to communicate with the agent
• Displays both the agent response and raw JSON-RPC messages
• Implements demo functions for text analysis

3. Python Agent

• Simulates an LLM that makes function calls
• Uses WebSocket for real-time communication
• Implements JSON-RPC 2.0 protocol