Exercises: Compare Tiny Vectors
Use starter.py to implement cosine similarity and rank candidate chunks.
1 Implement cosine similarity
Compute dot product, vector length, and normalized similarity.
2 Rank candidate chunks
Sort chunks by similarity to the query vector.
3 Add a production warning
Flag when similarity is high but metadata says a chunk is stale.
4 Explain the result
Return a human-readable retrieval note.