examples/php-slim
| Name | Last commit message | Last commit date |
|---|---|---|
| .. | ||
README.md
TanStack AI PHP Slim Framework Example#
This is a PHP Slim Framework server example that demonstrates how to stream Anthropic and OpenAI API events in Server-Sent Events (SSE) format compatible with the TanStack AI client.
Features#
- PHP Slim Framework server with SSE streaming support
- Converts Anthropic and OpenAI API events to TanStack AI
StreamChunkformat - Compatible with
@tanstack/ai-client'sfetchServerSentEventsadapter - Supports tool calls and function calling
- Type-safe request/response handling
Setup#
Prerequisites#
- PHP 8.1 or higher
- Composer (PHP dependency manager)
Step-by-Step Setup#
- Navigate to the project directory:
cd examples/php-slim
- Install dependencies:
Install PHP dependencies:
composer install
This will install all required packages (Slim Framework, Anthropic SDK, OpenAI SDK, etc.).
Install Node.js dependencies (for running the UI):
pnpm install
Or if you're at the repository root, you can install all dependencies at once:
pnpm install
- Set up environment variables:
Copy env.example to .env and add your API keys:
cp env.example .env
# Edit .env and add your ANTHROPIC_API_KEY and/or OPENAI_API_KEY
- Run the server and UI:
For Anthropic:
pnpm start
Or using npm:
npm start
This will start both the PHP server and the vanilla-chat UI concurrently.
For OpenAI:
pnpm start-openai
Or using npm:
npm start-openai
The servers will start on:
- PHP Server (Anthropic):
http://localhost:8000 - PHP Server (OpenAI):
http://localhost:8001 - UI:
http://localhost:3001
Note: The UI (vanilla-chat) is required to interact with the PHP server. The start script runs both servers together.
API Endpoints#
POST /chat#
Streams chat responses in SSE format.
Request Body:
{
"messages": [
{
"role": "user",
"content": "Hello!"
}
]
}
Response: Server-Sent Events stream with StreamChunk format:
data: {"type":"content","id":"...","model":"claude-3-haiku-20240307","timestamp":1234567890,"delta":"Hello","content":"Hello","role":"assistant"}
data: {"type":"content","id":"...","model":"claude-3-haiku-20240307","timestamp":1234567890,"delta":" world","content":"Hello world","role":"assistant"}
data: {"type":"done","id":"...","model":"claude-3-haiku-20240307","timestamp":1234567890,"finishReason":"stop","usage":{"promptTokens":10,"completionTokens":2,"totalTokens":12}}
data: [DONE]
GET /health#
Health check endpoint.
Usage with TanStack AI Client#
This server is compatible with the TypeScript TanStack AI client:
import { ChatClient, fetchServerSentEvents } from '@tanstack/ai-client'
const client = new ChatClient({
connection: fetchServerSentEvents('http://localhost:8000/chat'),
})
await client.sendMessage('Hello!')
StreamChunk Format#
The tanstack/ai package converts provider events to the following StreamChunk types:
content: Text content updates with delta and accumulated contenttool_call: Tool/function call events with incremental argumentsdone: Stream completion with finish reason and usage statserror: Error events
See packages/typescript/ai/src/types.ts for the full TypeScript type definitions.
Supported Providers#
The converter currently supports:
- ✅ Anthropic (Claude models) - fully implemented
- ✅ OpenAI (GPT models) - fully implemented
Project Structure#
php-slim/
├── public/
│ ├── anthropic-server.php # Anthropic server
│ └── openai-server.php # OpenAI server
├── composer.json # PHP dependencies
├── env.example # Environment variables template
└── README.md # This file
Architecture#
The server uses the tanstack/ai package located at packages/php/tanstack-ai/:
anthropic-server.php/openai-server.php: Handles Slim setup, provider client initialization, and HTTP endpointstanstack/aipackage: ProvidesStreamChunkConverter, message formatters, and SSE utilities for converting provider events to TanStack AI format
The converter package is installed as a local dependency, making it easy to develop and test changes.
Notes#
- The server uses CORS middleware allowing all origins (configure for production)
- Default Anthropic model is
claude-3-haiku-20240307(can be made configurable) - Default OpenAI model is
gpt-4o(can be made configurable) - Supports system messages, tool calls, and tool results
- Error handling converts exceptions to error StreamChunks
- Uses PHP 8.1+ features including named arguments and match expressions
Development#
To use the local tanstack/ai package during development:
- Add to
composer.json:
{
"repositories": [
{
"type": "path",
"url": "../../packages/php/tanstack-ai"
}
],
"require": {
"tanstack/ai": "@dev"
}
}
- Run:
composer update tanstack/ai