Project: Tiny Embedding Search Lab
Build a dependency-free search lab that ranks tiny vectors and reports production warnings.
What you will build
- A cosine similarity function.
- A tiny in-memory vector index.
- A top-k retrieval function.
- A metadata safety warning for stale and restricted chunks.
- A retrieval report that explains why a chunk was selected.
Product-level outcome
Search by meaning, inspect with discipline
The project should show learners both sides: semantic search is useful, and production retrieval still needs metadata, warnings, and explainable reports.