Hello CryptPad team,

I’d like to propose the integration of a connector compatible with OpenAI-like APIs (via LiteLLM) to bring optional, privacy-conscious generative AI capabilities to CryptPad.

Motivation

CryptPad is a pioneer in privacy-first collaboration. Many organizations and users are now exploring generative AI to boost productivity. By adding a modular AI connector compatible with LiteLLM — which supports 100+ LLM providers including on-premise and sovereign models — CryptPad can offer smart features without compromising user privacy.

Proposal

Introduce a configurable AI backend (LiteLLM or compatible OpenAI APIs) that can be self-hosted and optionally enabled by administrators. This would allow:

Email app enhancements:

Draft assistance for email composition
Thread summarization for long conversations
etc.

Spreadsheet (Sheets) integration:

Chat assistant to help with formula generation and data manipulation
etc

Optional in-app chat assistant:

Context-aware help across CryptPad modules (notes, kanban, etc.)

All prompts and responses can be encrypted and processed locally or through trusted, self-hosted LLMs (e.g. Mistral, LLaMA, Claude, etc.).

Benefits

Enhances productivity with intelligent features
Fully optional
Compatible with the project’s privacy principles when paired with on-premise LLMs
Makes CryptPad more competitive against commercial AI-powered suites (while remaining FOSS)

Implementation Notes

Leverage LiteLLM’s abstraction layer to avoid tight coupling to a specific provider

integration should respect CryptPad's design language and minimalism
Clear warnings/disclaimers for users when AI is used, especially if remote endpoints are configured

Would the team be open to discussing a possible prototype or roadmap for this feature?

Thank you for your time and for your work on this project.

Best regards,

François from France