Distributed multimodal database

Four data models.
One database.

Bring structured queries, semantic search and connected data into one database. Give your application more ways to find the right information, with fewer separate systems for your team to manage.

Start building Explore the data models

One engine, four ways to query

SQL Structured records Vector Meaning and similarity Graph Relationships Full-text Words and phrases

OriginChainDBConnected by record IDs

Query capabilities

Choose the query
for the task.

Start with one model. Add another as your application needs more ways to find and understand its data.

01 / SQL

Filter and aggregate.

Look up records, apply conditions and calculate totals across your tables.

02 / Vector

Search by meaning.

Find semantically similar content using embeddings, even when the wording differs.

03 / Graph

Follow connections.

Traverse relationships between records to explore their context and dependencies.

04 / Full-text

Find the right words.

Search indexed text for matching terms and retrieve the passages they appear in.

In practiceSee how the four models can work together

For a charger lookup, graph finds compatible devices, SQL checks stock, vector finds the relevant setup guide and full-text locates a specific instruction. Your app connects the results using record IDs.

Explore the catalog code example

What you can build

Put your data
to work.

Bring search, similarity and relationships into the features people use every day.

Explore all solutions

Knowledge search

Find relevant passages by their wording or meaning, then retrieve the source records for your application or AI workflow.

Recommendations

Discover similar items, follow their relationships and check record attributes before making a suggestion.

Fraud investigation

Explore connections between accounts, devices and transactions to help investigators spot shared patterns.

Behind the queries

Keep every result
linked to its source.

Record IDs let your application connect a search match to the original data and its relationships.

Explore the architecture

Give your records a shared identity
Reference the same record ID in your tables, relationships and search indexes so results can be traced back to their source.
Prepare the indexes you need
Your ingestion code writes searchable text to the full-text index and stores embeddings in the vector index.
Bring results together in your app
Use returned IDs to fetch source records and combine the context your application needs.