Search foundation

Text2Cypher

For questions that live in relationships. Text2Cypher writes the Cypher query a graph database runs, so answers come back as exact records rather than passages that merely look similar.

In development

What is Text2Cypher?

Cypher is the query language of property-graph databases such as Neo4j. Asking a graph a question normally means writing Cypher by hand, which takes knowing both the language and the graph’s schema.

A Text2Cypher model takes the question and the schema — node labels, relationship types and properties — and writes the query. The database runs it and returns exact results, shown directly or handed to an LLM as grounded context: the retrieval step of GraphRAG.

How it works

Text2Cypher in 3 steps.

  1. 01

    Schema

    The model is given the graph schema: labels, relationship types and properties.

  2. 02

    Generate

    It translates the question into a Cypher query over that schema.

  3. 03

    Execute

    The database runs the query and returns exact records for your application or LLM.

Models

Text2Cypher models

No Text2Cypher model is released yet. Models appear here as soon as they are published on Hugging Face.

Compare

Text2Cypher and vector search

Text2CypherVector search
ReturnsRecords that satisfy the queryPassages similar to the query
Strongest atRelationships, filters, counts, multi-hop questionsOpen questions over unstructured text
NeedsA graph database and its schemaA vector index over documents
Answer typeExact: a record matches the query or it does notApproximate: ranked by similarity

FAQ

Text2Cypher, explained.

What is GraphRAG?
Retrieval-augmented generation grounded in a knowledge graph. Instead of, or alongside, similar passages, the LLM receives the entities and relationships a graph query returns.
Which databases use Cypher?
Neo4j, and through the openCypher project other graph databases such as Memgraph and Amazon Neptune.
How do I get access?
Text2Cypher is in development. To hear when it is released, get in touch.

Start a conversation

Put Text2Cypher to work
on your search.

Tell us what you retrieve and how you measure quality.

Get in touch →