OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

Lintdb

LintDB is an Apache-2.0-licensed C++ vector database library that embeds in Python applications to store and retrieve multiple token-level vectors per document, with late-interaction support for ColBERT and PLAID.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Lintdb — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use LintDB when a Python application needs an embedded retrieval index for multi-vector, token-level search, particularly with ColBERTv2 and PLAID, and can install the library through conda.

Who it is for

  • Python developers building generative-AI or RAG retrieval components.
  • Teams whose retrieval design uses token-level embeddings or late-interaction scoring.
  • Developers who need schema-defined fields for indexing, storage or filtering.

Documented uses

  • Store multiple token-level vectors under a document ID and rank using maximum similarity.
  • Create an IVF index for ColBERT embeddings.
  • Search an index using vector queries and request ranked results.
  • Filter retrieval against fields defined in the index schema.
Before you choose

Requirements & limits

The README specifies conda as the supported installation route and provides a conda install command. The repository lists the Apache-2.0 license.

  • The library is embedded in Python rather than presented as a separately configured database service.
  • The documented installation route is conda; the project cites OpenBLAS as a dependency for accelerated matrix multiplication.
  • The README marks ColBERTv2 with PLAID as supported and XTR as unsupported.
  • The README reports bit-level compression that can store 128-dimensional embeddings in as little as 16 bytes.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Define an embedding index and query it

Set up the documented schema and index, then form a vector query and search the index.

Before you start

  • LintDB installed through conda
  • A Python application and a 128-dimensional ColBERT embedding field
  • An embedding value and an index path

Steps

  1. Define a schema containing a ColbertField with tensor data, 128 dimensions and the documented binarizer quantization settings.

  2. Create a Configuration and initialize an IndexIVF with the index path and schema.

  3. Create a Query containing a VectorQueryNode for the ColBERT tensor field.

  4. Call index.search(0, query, 10) to request results.

Expected outcome

The documented usage pattern returns search results from the index for a vector query.

Guide limitations

  • The example assumes an embedding has already been produced; it does not document setup for the embedding model.
  • The example does not provide a complete data-ingestion procedure.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
DeployQL/LintDB · Official homepage title · retained in its source language
How we review sourcesSuggest a correction

Related options

Vespa — The README names Vespa as another system supporting token-level embeddings and contrasts its search-engine setup with LintDB’s embedded approach.

Explore project →

Chroma — The README compares the projects as embedded vector databases and states that LintDB supports multi-tenancy while its documented interface is Python-only.

Explore project →