Pinecone. The selected vector store provider for this configuration.

Required for normal operation. Create and copy the key from the Pinecone console (API keys). Used to authenticate all index and vector operations. If you enter a key here, treat the saved configuration as a secret.

Name of the Pinecone index for this workflow. If the index does not exist, the integration may create it automatically using your server or serverless settings below—verify auto-create support and naming rules in your environment.

Optional partition inside the index (default is often default). Use separate namespaces for logical isolation (for example multi-tenant data). Queries typically target one namespace unless your product merges results across namespaces.

Maximum number of vectors to return from a similarity search. Must be a positive integer when set.

Optional minimum similarity score; hits below this value are dropped. Often interpreted on a 0–1 scale depending on metric and client; leave empty to return all top-K matches without a score cutoff.

Serverless — Pinecone scales capacity automatically; billing is usage-based; a good default for many new workloads. Pod-based — Dedicated pod capacity with more predictable performance characteristics; may suit steady high-throughput or legacy setups. When you switch type, the UI may show different fields (for example pod size or environment)—complete all required fields for the selected type.

Vector length for the index. Must match your embedding model output (for example 1536 for OpenAI text-embedding-3-small at default dimension, or 384 for common local models). Cannot be changed after the index is created.

Cloud where Pinecone hosts the serverless index: for example aws, gcp, or azure. Choose a region available for that provider in your Pinecone project.

Region code for the index (for example us-east-1 on AWS). Pick a region close to your application and compliant with data residency requirements.

When enabled, reduces risk of accidental index deletion via API or console. Turn on for production indexes; you may disable in sandboxes when you expect to tear down indexes often.

None — skip embedding configuration only when valid for this feature. Pre-configured — use a profile from AI Services > Embedding models. Custom — configure embedding inline.

After you choose a source, select the embedding provider (OpenAI, Mistral, and so on). That drives vector generation; Pinecone settings above define where vectors are stored and how they are queried.