Choose ANN Vector search if:

• You want maximum speed and scalability
• Your queries are mostly semantic similarity (meaning-based search)
• You have clean embeddings and good vector quality
• You don’t need keyword precision or exact matches

Choose Hybrid RRF (Reciprocal Rank Fusion) if:

• You need better relevance quality out of the box
• You want to combine keyword + semantic search
• Your data includes IDs, technical terms, names, or rare tokens
• You want more robust ranking across different query types

Practical recommendation

In most real-world RAG / search setups, Hybrid RRF is the safer default because it:

• reduces “semantic drift”
  improves recall for exact terms
• handles mixed query styles better

👉 So:

• Default: Hybrid RRF
• Optimize for scale/latency: ANN Vector