Query Your Tables
as a Graph.
No ETL and no graph database. Run multi-hop queries, shortest paths, PageRank and centrality on the tables you already have, with one copy of your data you can query in SQL or as a graph. Then turn the results into decisions with rules you can trace.
Proven in production
Graph Questions,
Table Answers
Most relationship data already lives in warehouses and databases. A graph database asks you to copy it somewhere else first. A graph query layer skips the copy, but the business logic still lives somewhere else.
Core Algorithms.
One Line Each.
Graph functions are part of the Vadalog language. Point one at any predicate with edges, from a Parquet file to a Snowflake table, and it returns results as a new relation you can join, filter and reason over.
Paths and Reachability
Reachability, all paths, and shortest paths from one source or between all pairs.
Centrality
PageRank with tunable damping, degree centrality and betweenness centrality, with sampling for large graphs.
Components and Structure
Connected components find clusters and rings; reachability and all paths show how any two entities connect.
Reasoning on the Output
Feed algorithm results straight into rules: aggregate shortest paths, flag high-rank nodes, or combine them with business policy.
Answer graph questions this week, not after a migration. Retire the nightly sync into a separate graph store. Stop paying to store and secure the same data twice.
Ask in Plain English.
Keep Your Tables.
Ask the graph question in natural language and the AI assistant drafts the rules that answer it, over the tables you already have, for you to review. Teams who already know Cypher can keep using it.
Ask in Plain English
Describe the question, such as which accounts sit at the centre of the money flow, and the assistant drafts the rule. See natural language to query.
SQL Inside Graph Functions
Pass a SELECT straight into a graph function to filter or join the edge set before the algorithm runs.
Cypher If You Want It
Reuse Cypher patterns, shortest paths and graph procedures over files and databases, with no Neo4j server needed.
Logic Where Queries Stop
Recursion and explanations run as rules downstream of the graph step, in the same program.
Ask graph questions in plain English. Join graph results with warehouse data in one place. Reuse Cypher where you already have it.
From Tables to
Graph Insight
Connect Your Sources
Point Prometheux at the tables or files that hold your entities and relationships. No ETL, nothing copied.
Define the Edges
Say what counts as a connection in plain English, or in SQL or Cypher, across one source or many.
Run the Algorithm
Run multi-hop queries, or call PageRank, shortest paths or components in one line. Execution is parallel and distributed.
Reason on the Result
Join scores and paths with business rules, then serve them to apps and agents via API or MCP.
Graph Questions
Teams Ask First
Each one runs on tables you already have, with no ETL into a graph store, and every answer shows the path and rules behind it.
Financial Services
Follow money, accounts and exposure across every hop.
Supply Chain
See bottlenecks and knock-on effects before they reach customers.
Cybersecurity
Trace assets, identities and access as one connected picture.
Healthcare & Life Sciences
Connect drugs, targets, patients and providers across sources.
Graph Questions
You Already Have
Networks at Scale
A financial institution sped up critical reasoning over knowledge graphs of hundreds of millions of nodes and billions of edges by over 250%.
Finance & banking →Biomedical Knowledge Graphs
AstraZeneca reasons over 4.7M compound, gene and disease data points, accelerated 9× on NVIDIA GPUs.
NVIDIA blog →Rings and Hubs
Find connected groups with connected components, and rank the accounts at the centre of the flow with degree, betweenness and PageRank centrality.
Graph functions →Network Criticality
Use betweenness centrality to find the sites and links that most traffic depends on, straight from inventory tables.
Telecom →“With Prometheux, we achieved a speed-up of over 250% on critical tasks involving complex reasoning over knowledge graphs with hundreds of millions of nodes and billions of edges.”
Financial institution
“Prometheux technology powers highly scalable and explainable reasoning over some of the world's largest knowledge graphs.”
Accelerating Neurosymbolic AI with RAPIDS and Vadalog Parallel
“The ability to query heterogeneous databases through a flexible ontology makes it possible to infer deep knowledge and automate complex tasks, delivering trustworthy insights that far exceed the capabilities of traditional systems.”
Anna FavaFormer Senior Manager, Data & AI · Dompé
Questions,
Answered
Can I query my tables as a graph without a graph database?
Yes. Prometheux runs graph algorithms such as PageRank, shortest paths, centrality and connected components directly on relational tables and files. Any table with source and target columns can be queried as a graph, with no ETL into a separate store.
Which graph algorithms are supported?
Reachability, all paths, shortest paths, connected components, and degree, betweenness and PageRank centrality.
Can I use Cypher?
You don't need to: you can ask in plain English and the AI assistant drafts the rules. If your team already uses Cypher, MATCH patterns, shortestPath and graph procedures run over files and databases, with no Neo4j server needed.
How is this different from a graph query engine over my tables?
A graph query engine lets you query tables as a graph. Prometheux also reasons over the result: algorithm outputs feed recursive rules, aggregates and business policy in the same program, every derived fact keeps its provenance, and results can be written back to any connected source.
Do my analysts need to learn a new language?
No. They can ask in plain English and the AI assistant drafts the rules for review, or define edges in SQL or Cypher. Algorithms are one-line calls.
What if we already run Neo4j?
Neo4j is one of the supported sources, so you can query it alongside your warehouse and files in the same program instead of syncing data between them.
Does it scale to large graphs?
Yes. Graph functions run parallel and distributed. A financial institution reported a speed-up of over 250% on reasoning over knowledge graphs with hundreds of millions of nodes and billions of edges.
Related Capabilities
Query Your Own Tables
as a Graph
Bring a schema with relationships in it. We'll query it as a graph in place, with paths, PageRank and connected components, and no ETL.
