🚀 Platform Updates | June 27, 2025

[ashlynn-brown]

🚀 Platform Updates | June 27, 2025

This week brought major advances across access control, document ingestion, and infrastructure observability, laying critical groundwork for broader enterprise deployment and deeper AI mentor configurability.

🤖 AI Mentor Platform Enhancements

• Full RBAC Enforcement Across Mentor Interfaces: RBAC controls now extend across mentor settings, detail views, prompt management, and analytics history, ensuring fine-grained permissions at both UI and API levels. Students can now safely access settings endpoints when granted.
• Tenant-Specific Cost Tracking by User & Timestamp: A new cost tracking system supports per-user, per-session, and ISO datetime filters, empowering platforms to measure and optimize LLM usage down to the session level.
• Smarter Document Indexing with Crawler Controls: Document ingestion now accepts configurable crawl depth, limits, and pattern filters, both via API and Django admin. This upgrade is crucial for scalable multi-page site indexing and selective content inclusion.
• Credential Injection for Google Drive & Dropbox: Teams can now dynamically update credentials when reprocessing docs from connected cloud storage, streamlining secure document refresh cycles.
• Langfuse Filtering for Advanced Debugging: Traces and observations can now be queried by platform key, session ID, user ID, and mentor, giving teams precision observability into AI interactions across the stack.

💻 Frontend & Embeds

• NEXTJS Compatibility for Embeds: All message types are now namespaced with MENTOR: for compatibility with the new mentorAI frontend architecture.
• Dynamic Document Filters for Embedded Mentors: Mentors can now receive filter props to limit source documents at runtime, enabling tailored AI responses based on current context or user role.

🧰 Backend Improvements

• Object-Level RBAC Support + Action Scoping: Roles now define both endpoint-level and field-level actions (get, post, patch, etc.) along with data-level access (objects and fields). Token endpoints and API access patterns now fully respect these scopes.
• Declarative Swagger Schema Categorization: API endpoints are now semantically grouped within generated schema files, paving the way for better developer tooling, auto-generated clients, and IDE integration.
• Student Serializer Support in Mentor Settings: Student data is now formally supported in mentor settings, opening the door for more personalized logic in AI behavior or feedback.