style: drop the explanatory comments added with the HybridRetriever wrapper

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: kfaracik

__tests__/hybridRetrieval.test.ts

@@ -617,13 +617,6 @@ describe('HybridRetriever', () => {
     mockKeywordSearch.mockReset();
   });
 
-  // The cases above already exercise the hybrid logic; this proves the wrapper
-  // forwards 1:1 — query→prompt, store/embeddings from the constructor, and
-  // every option spread through. Inputs are chosen so the two options a naive
-  // spread could silently drop are load-bearing: sourceNamesById resolves doc
-  // 1's missing name, and attachmentSourceIds keeps doc 2's otherwise-gated
-  // low-similarity chunk and orders it first. A wrapper that dropped either
-  // would diverge from the raw call and fail the toEqual.
   it('forwards to hybridRetrieve 1:1, including attachmentSourceIds and sourceNamesById', async () => {
     const vectorResults = [
       {

utils/hybridRetrieval.ts

@@ -426,13 +426,6 @@ export const hybridRetrieve = async ({
   );
 };
 
-// Thin app↔library boundary. Binds the store + embeddings so retrieval is a
-// single retrieve(query, options) call, making the hybrid testable and portable
-// in isolation. It forwards to hybridRetrieve unchanged — no interface, because
-// there is one implementation and one caller; extract a Retriever interface only
-// if a second retriever ever appears. Deliberately NOT `implements VectorStore`:
-// the hybrid is read-only and its ContextChunk output drops id/embedding, so
-// coercing to QueryResult would change retrieval results.
 export type HybridRetrieveOptions = Omit<
   HybridRetrieveParams,
   'prompt' | 'vectorStore' | 'embeddings'