Jinni
A conversational application exploring personal context, approved memories and the foundations of a more individual AI assistant.
Project illustrationThe challenge
A personal assistant needs useful context without treating every generated observation as a permanent fact. Conversation history, profile information and proposed memories need distinct roles.
The approach
Build an authenticated chat endpoint connected to OpenRouter.
Assemble context from conversation history, profile information and approved memories.
Keep proposed memory candidates separate from approved memory.
Record model usage and leave isolated per-user agent runtimes as a future integration.
What the work demonstrates
- Implemented authenticated model-backed chat and contextual memory handling.
- An application foundation for personal AI experiences.
- Dedicated per-user Hermes environments remain planned work.
