Make your Alexa actually intelligent by connecting it to powerful LLMs.
A custom Alexa skill that routes your voice queries to an LLM (currently Gemini / Groq / Claude) instead of Alexa's default responses. Think of it as replacing Alexa's brain while keeping its ears and mouth.
Architecture:
Your Voice → Alexa (ears) → AWS Lambda (router) → LLM API (brain) → Alexa (mouth) → You hear the answer
- 🎙️ Voice-powered AI assistant through any Alexa device
- 🧠 Powered by your choice of LLM (Gemini, Groq, Claude)
- 💬 Multi-turn conversations — remembers context within a session
- 🆓 Runs entirely on free tiers (Lambda free tier + free LLM API)
- 🔧 You own the stack — customize system prompts, swap models, add memory
├── lambda/
│ ├── lambda_function.py # Main Lambda handler (Groq/Llama)
│ └── lambda_function_gemini.py # Alternative: Google Gemini version
├── skill-config/
│ └── interaction-model.json # Alexa skill interaction model
├── SETUP-GUIDE.md # Step-by-step deployment guide
└── README.md
- Get an API key (Groq or Google AI Studio)
- Create an AWS Lambda function (Python 3.13)
- Paste the Lambda code and set your API key as environment variable
- Create an Alexa skill and paste the interaction model
- Connect the skill to your Lambda
Full instructions in SETUP-GUIDE.md
| Service | Cost |
|---|---|
| AWS Lambda | Free (1M requests/month, forever) |
| Alexa Skill | Free |
| Groq API | Free tier (no credit card) |
| Gemini API | Free tier (no credit card) |
Total: $0/month for personal use.
- Phase 1: Basic Alexa → LLM pipeline
- Phase 2: Persistent memory across sessions (DynamoDB)
- Phase 3: Context-aware personal assistant (custom memory system)
- Phase 4: Multi-user support and productization
| Provider | Model | Status |
|---|---|---|
| Groq | Llama 3.3 70B | ✅ Working |
| Gemini 2.5 Flash | 🔄 Ready | |
| Anthropic | Claude Haiku | ✅ Working (needs credits) |
MIT