Loading…

Features Pricing

Log in Get started

Vector Database

Vector Database

Storage

A vector database is a database designed to store vectors (embeddings) and answer similarity or nearest-neighbor queries efficiently. DocLD uses Pinecone as its vector database for knowledge base chunks: vector search runs against Pinecone to power RAG retrieval.

Capabilities

Upsert — Store embeddings with metadata (e.g., document-id, page, chunk-id).
Query — Given a query embedding, return the most similar vectors (top-k) and optional metadata filtering.
Namespace — Organize vectors by namespace (e.g., per knowledge base).

DocLD’s vector index is backed by Pinecone; ingestion and reindex update the vector database.

Related Concepts

Vector database backs the vector index and vector search. Pinecone is DocLD’s vector database. Embedding and retrieval depend on it.

Related terms

Index
Pinecone
Embedding
Vector Search

See also

features › knowledge bases

Frequently Asked Questions

A vector database stores vector embeddings and answers similarity or nearest-neighbor queries efficiently. DocLD uses Pinecone as its vector database for knowledge base chunks and RAG retrieval.

Vector search over embeddings finds semantically relevant chunks quickly. The vector database stores chunk embeddings and metadata so queries can retrieve the top-k most similar chunks for the LLM context.

DocLD is built on Pinecone. Support for other vector databases would require a different deployment or integration; check the roadmap.

Pinecone and DocLD manage scaling. You add documents and query; the index scales with your plan and usage.

No. All queries go through the DocLD API (chat, search). This keeps security, scoping, and billing in one place.

Embedding vectors and metadata (document_id, page, chunk_index, etc.). Raw text is not stored in the vector DB; it is in DocLD storage.

Vectors can be recreated by reindexing from source documents. DocLD and Pinecone handle durability; check backup options for your plan.

Yes. Metadata filters (e.g., document type, date) narrow the set of vectors considered in a query. DocLD applies filters when you search.

Pinecone is optimized for low-latency similarity search. Query time is typically milliseconds; total RAG latency includes embedding and LLM.

See the DocLD documentation and glossary for details, and check the API reference for related endpoints and options.

On this page

Capabilities
Related Concepts
Frequently Asked Questions

Product

Features
Pricing
API Reference

Industries

Insurance
Construction
Professional Services
Finance
Energy & Utilities
View all

Company

About
Careers

Resources

Documentation
Blog
Help Center
Status

Legal

Privacy Policy
Terms of Service
Trust

Connect

X
GitHub
Discord

© 2026 DocLD, Inc.SOC audit in progress