Hi everyone, I'm a backend developer currently working on a project that requires RAG integration. The goal is to store thousands of users' preferences and behavioral history so that an LLM can deliver more personalized and relevant responses based on each user's profile. Our current testing volume is approximately 4,000 active users, each with 30–50 preference entries. These range from accommodation style (e.g., "prefers small but upscale lodging"), activity patterns (e.g., "tends to book evening/nightlife activities"), climate preferences (e.g., "favors warm destinations"), to booking history (e.g., "last three reservations were to X, Y, Z for 3, 1, and 10 days respectively"). I'm currently using Qdrant as my vector database, but Pinecone has been recommended to me with a lot of enthusiasm. I'd love to hear from the community about your real-world experience — particularly around latency, cost at this scale, and any tradeoffs you've encountered. Thanks for reading, and I appreciate any insights you're willing to share.

Hi @elegant-camera-66606 - happy to chat through your specific requirements here. I suspect Pinecone will be much less expensive than a production Qdrant setup for this workload.