AI & Machine Learning
Bring the power of large language models to your edge devices. Analyze sensor data, generate reports, detect anomalies, and create intelligent automation with natural language.
openai
Connect to OpenAI GPT models for text generation, analysis, and embeddings. The most widely used AI API with excellent performance.
Configuration
Output Example
{
"payload": "AI response text",
"usage": {
"prompt_tokens": 50,
"completion_tokens": 100
}
}Use GPT-4o for best quality/speed balance. Enable streaming for long responses. Cache responses in Redis to reduce API costs.
anthropic
Integrate Claude models for advanced reasoning and long-context understanding. Excellent for complex analysis tasks.
Configuration
Output Example
{
"payload": "Claude response",
"stop_reason": "end_turn",
"usage": {...}
}Claude excels at analysis, coding, and long documents. Use Haiku for speed, Opus for complex reasoning, Sonnet for balanced performance.
ollama
Run local LLMs with Ollama for privacy-focused, offline AI processing. No data ever leaves your device.
Configuration
Output Example
{
"payload": "Local LLM response",
"model": "llama3",
"eval_count": 150
}Install Ollama first: ollama.ai. Pull models with ollama pull llama3. For Raspberry Pi, use smaller models like phi or tinyllama.
AI-Powered IoT Use Cases
Transform your edge devices with intelligent automation
Analyze sensor patterns and detect unusual behavior that might indicate equipment failure.
Automatically summarize daily sensor data into human-readable reports.
Parse and understand error logs to identify root causes and suggest fixes.
Use vision models to classify camera images for quality control or security.
Control devices with voice or text: "Turn on the lights when it's dark".
Predict equipment failures before they happen using historical sensor data.
Cloud vs Local AI
Choose the right approach for your use case
Quick Tips
Use function nodes to pre-process data before sending to AI models.
Store AI responses in Redis to avoid redundant API calls for similar inputs.
For Raspberry Pi, try Ollama with phi or tinyllama - optimized for edge devices.
Store API keys in environment variables, never hardcode them in flows.