SDK Examples

Interactive examples demonstrating PromptKit SDK features.

Overview

All examples use the Pipeline architecture - audio, TTS, VAD, and other features flow through Pipeline middleware rather than SDK-specific methods.

Available Examples

🎙️ streaming-tts

Demonstrates streaming text-to-speech through the Pipeline.

• Streaming LLM responses
• TTS middleware converting text to audio
• Real-time audio generation during streaming

🗣️ voice-chat

Demonstrates voice conversation through the Pipeline.

• VAD (Voice Activity Detection)
• Turn detection
• TTS responses
• Full voice conversation flow

🔄 streaming

Basic streaming example showing real-time LLM responses.

🎨 multimodal

Working with images, audio, and other media types.

🖼️ image-preprocessing

Automatic image resizing and optimization for vision models.

• WithAutoResize() for simple size limits
• WithImagePreprocessing() for full control
• Quality optimization for API costs and latency

🎥 realtime-video

Realtime video/image streaming for duplex sessions.

• SendFrame() for webcam-like scenarios
• WithStreamingVideo() for frame rate limiting
• Simulated frame capture example

🛠️ tools

Function calling and tool integration.

📊 variables

Dynamic variable substitution in prompts.

🔍 vad-demo

Voice Activity Detection examples.

👤 hitl

Human-in-the-loop workflows.

👋 hello

Simple getting started example.

Running Examples

All examples follow the pattern:

cd <example-name>
export OPENAI_API_KEY=your-key  # or appropriate provider key
go run .

Examples with the _interactive.go suffix are designed to be run interactively and demonstrate features visually.

Architecture Note

Pipeline-First Design:

These examples demonstrate the correct architecture where TTS, VAD, and audio processing happen through Pipeline middleware, not via SDK convenience methods.

✅ Correct: Configure Pipeline → Everything flows through middleware
❌ Old way: SDK methods bypass Pipeline

Benefits:

• Consistent interface for all processing
• Middleware composability
• Testable with mocks
• Observable via hooks
• Lower latency (streaming audio generation)

For integration tests and lower-level Pipeline usage, see tests/integration/.