Store and search high-dimensional vectors at scale.
about 1 month ago
Pinecone was the only game in town for a while. It's still the easiest to use. But with pgvector (Postgres) and Chroma becoming so good, it's hard to justify a separate specialized database sometimes. We are considering moving to pgvector just to keep our stack simpler.
4 months ago
We migrated from a self-hosted FAISS setup to Pinecone and haven't looked back. It's fully managed, so we don't worry about scaling or uptime. The latency is incredibly low for our RAG application. The 'serverless' pricing model is great because we only pay for what we use, rather than provisioning huge instances.
8 months ago
Pinecone is fast and reliable. Setting up an index takes seconds. The documentation is top-tier. However, if you have a high volume of writes/upserts, the costs can creep up. For read-heavy workloads it's efficient. The new pod-based architecture options give more control, but it's still a premium product.
