Chunking Documents for RAG

Long documents must be split into chunks before embedding. Overlap between chunks preserves context that would otherwise be cut at a boundary.

Implement chunk_words(text, chunk_size, overlap) that:

• Splits text on whitespace into words.
• Returns a list of strings, each containing at most chunk_size words joined by single spaces.
• Consecutive chunks share exactly overlap words.
• Stops once the final word has been included (no trailing chunk that is fully contained in the previous one).
• Raise ValueError if overlap >= chunk_size.

Example

chunk_words("a b c d e f g", chunk_size=3, overlap=1)
# ['a b c', 'c d e', 'e f g']