RAG over Reviews

[yamirghofran]

Given a text query (re-written by an LLM to look like a review):

1. Generate the embedding for that query (call the endpoint serving the reviews embedding model - default EmbeddingGemma)
   Use Chroma SDK to find top_k similar reviews given to the query.
   Return a JSON of the reviews (including the metadata like review_id, book_id, text, etc.).