A modular server that connects Alexa to language model providers behind one adapter interface, so the provider is a setting rather than a rewrite. Prompt components live in separate files, conversation history carries across Alexa sessions, and the prompting is tuned for answers that are spoken rather than read.
What stands out
- One adapter interface for OpenAI, Ollama and other providers
- Conversation history across Alexa sessions
- Prompting tuned for spoken answers