Model & Assumptions Experimental
- Model
- Semantic search over course-note passages using all-MiniLM-L6-v2 embeddings (in-browser). Optional AI synthesis via Claude Haiku using retrieved passages as context.
- Assumptions
- Retrieval quality depends on the embedding model and the corpus (course notes on computational photonics methods). AI answers are grounded in retrieved passages but may contain inaccuracies.
- Limitations
- Corpus is limited to available course notes — not exhaustive. AI mode requires a user-provided Anthropic API key (stored in session or localStorage). Answers should be verified against the source passages.
Methods Assistant
Nanophotonics solver recommendationsAsk about methods
Ask a physics or numerical-method question. I search a corpus of
computational-photonics course notes and return relevant passages.
With an API key, Claude synthesizes an answer from the retrieved context.
First query downloads the embedding model (~20 MB, cached).
How it works: your query is embedded in the browser with
all-MiniLM-L6-v2 (via Transformers.js)
and matched against pre-computed chunks of course notes.
With an API key, Claude Haiku
synthesizes an answer from the retrieved passages. Without a key, passages are shown verbatim.