IvanJS17 locks down bring-your-own AI keys
This fork turns personal AI credentials into a governed choice made once, rather than a setting that can drift mid-conversation.
IvanJS17 has added encrypted storage for personal provider keys and a controlled menu of approved AI routes. When someone starts a chat, they choose the provider, model and key up front; that combination stays fixed for the life of the conversation.
- OpenAI, Anthropic and Gemini options for major general-purpose AI models.
- OpenRouter, a service that provides access to multiple AI models through one account.
- DeepSeek, plus OpenCode Zen and Go, available as separate approved options.
If the chosen key is no longer available, the chat stops before it records a message or starts responding. That is a small but meaningful control: no silent switch to another model or credential after the fact.
So what Legal teams testing personal AI access should care because the fork makes each chat's AI route explicit, stable and easier to govern.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?