From book to answer
See what is prepared, and what has been tested.
Books, courses, conversations and evidence.
One private place to bring it all together.
Sign in to your central AI workspace.
A current snapshot of your books and connections. Evidence, not guesswork.
See what is prepared, and what has been tested.
Checking now…
Actual indexed books in this LMS workspace.
Project-wide saved tests. Not a live certification of this LMS.
Your LMS authorizes its students. Central AI handles books, answers and memory.
Connect the right knowledge to the right course.
PDFs & images · 100 MB local / 50 MB hosted OCR · 500 PDF pages · Hosted OCR for handwriting
Saved indexing jobs for this workspace.
Test PDF-to-text and handwriting-to-text without indexing books.
Private test uploads are scoped to this LMS and test course. No embeddings, exams or grades are created. Hosted OCR can incur charges. This is owner testing, not proof of teacher login or enrollment.
Uses the saved extraction connection.
Choose Visual OCR for handwriting PDFs, even when printed questions have selectable text. Automatic extraction does not capture all image-only writing or diagram labels. Originals remain available for review.
Load test history or upload a document.
Review draft—not certified OCR accuracy.
Ask, follow up, return to history—and see the book evidence.
Separate what is healthy now from what has actually passed a test.
Live checks use real, read-only API calls and negative authorization tests. They do not create accounts, start services or spend model quota. Use the chat playground to test real inference.
Not run in this browser session.
Project-wide acceptance results with timestamps.
Your LMS stays in charge of its users. Central AI stays in charge of knowledge.
No LMS database copy or direct browser access to AI secrets.
Choose a provider and model for each job. Connect a provider with its API key once.
Connect tests your selection before activating it. Test alone does not save. Keys are encrypted server-side and never returned to the browser. Tiny real tests may incur provider charges. No silent fallback.
Chat and extraction models can change without re-embedding. A different embedding model needs explicit approval to rebuild all books using saved text. Original files and the previous vector index are kept. Extraction changes apply to new uploads; use a book’s Re-extract action to extract it again.
Private originals and extracted text. No direct browser credentials.
Storage access stays behind the authenticated API. Existing local books are not migrated automatically. Teacher answer keys and homework workflows are not implemented yet.
Backend-only authentication.
Rotation immediately invalidates the previous token. The browser must never hold an LMS backend token.
Permanently remove this LMS's Central AI data and revoke its backend token. This does not delete the external LMS application or database.
Includes libraries, files, vectors, conversations, memories and standalone extraction uploads. Other workspaces and shared provider settings remain.
The LMS backend must verify login and course access on every request.
Complete frontend/backend examples are in docs/INTEGRATION.md. Interactive endpoint documentation is available at /docs.