Read-only reference material that informed NotebookAI's design. Preserved here so future contributors can trace the lineage of design decisions, but not part of the running product — nothing in this directory is imported by backend/, frontend/, or desktop/.
Moved into archive/ during Phase 14 (polish + audit) of the build, after the equivalent functionality was reimplemented natively in backend/notebookai/ and the karpathy-llm-wiki skill was packaged into skills/karpathy-llm-wiki/ (the canonical scaffold source).
The original NotebookLM-style RAG project that NotebookAI evolved out of. The pieces that survived the migration:
app/adapters/{pdf,url,youtube}.py→ reimplemented inbackend/notebookai/adapters/app/services/embeddings.py(sentence-transformers wrapper) → reimplemented inbackend/notebookai/index/embeddings.pyapp/services/chunking.py→ reimplemented inbackend/notebookai/index/builder.py- sqlite-vec wiring → reimplemented in
backend/notebookai/index/store.py
The pieces that were deliberately discarded (see VISION.md §"Migration from OpenNotebookLM"): JWT auth, multi-user routing, the Project → Document → Chunk → Conversation → Message SQLAlchemy hierarchy as primary state, the chat-centric four-pane layout, the cloud-deployable FastAPI surface.
Karpathy's original llm-wiki reference implementation. The wiki-as-substrate model (compile / cascade / lint / query / archive ops) is taken directly from this repo. The skill bundle at skills/karpathy-llm-wiki/ is the production version — it adapts the original for the agent-skills standard so any skill-aware CLI can use it.
These directories are kept verbatim for provenance. If you need to change behavior, change backend/notebookai/, skills/karpathy-llm-wiki/, or frontend/ — the live code paths.
If you're trying to delete archive/ to slim the repo: it's safe. Nothing in the running NotebookAI product depends on it.