Ask in English.
Run as Logic.
Self-service answers without the SQL. The AI Assistant turns plain-language questions into readable logic over your business concepts, runs it on your data where it lives and saves it, so people and agents get the same answer next time.
| customer | last order |
|---|---|
| Northwind | 112 days |
| Kestrel GmbH | 97 days |
| Alba SRL | 131 days |
Proven in production
Answers With
the Working Shown
Text-to-SQL is only as accurate as its grasp of your business. Generic tools guess at joins over a raw schema and hand back a number. Prometheux writes logic against your business concepts, so you can read it, rerun it and reuse it.
From Question
to Concept
Ask a question in the chat. The assistant drafts Vadalog against your ontology, saves it as a concept, executes it and streams its progress as it goes. The answer arrives with logic you can read, not just a result you have to trust.
Concept-Aware Drafting
Questions are answered in terms of your entities and concepts, not raw column names.
Clarify, Don't Assume
When a question is ambiguous, the assistant asks and offers options to pick from.
Self-Repairing Code
If generated code fails, the assistant fixes it and re-runs before handing back the result.
Safe by Default
It asks before making changes and always before deleting anything, stays scoped to your data and works with your own access rights.
Let analysts and business users ask directly. See the logic behind every number. Turn good questions into reusable concepts.
Bring the Query
Language You Know
Concepts can be written in Vadalog, SQL, Cypher or Python, and they all share the same concepts and lineage. The concept editor and the Vadalingo API translate existing SQL, RDF or OWL into Vadalog, and the concept editor drafts Vadalog from a plain-English description.
Four Concept Types
Write each concept as Logic (Vadalog), SQL, Cypher or Python, side by side.
Translate Existing Logic
Bring SQL, RDF or OWL and translate it into Vadalog from the concept editor or the Vadalingo API.
English to Vadalog
In the concept editor, describe what you want and generate Vadalog from the description.
Code to English
Generate a plain-English description from any concept's code, so reviewers know what it does.
Keep your team's SQL and graph skills. Move existing logic over without rewriting by hand.
From Question to
Verified Answer
Ask
Type a question in the AI Assistant, or click Ask AI on any concept.
Clarify
Pick from the options the assistant offers if your question could mean more than one thing.
Generate and Run
The assistant writes Vadalog over your concepts, saves it as a concept and runs it.
Review and Reuse
Read the code, edit if needed, and use the concept in apps, agents and reports.
Questions Without
the Ticket Queue
Self-Serve Answers
Ask questions in plain language and get results grounded in your business rules, without waiting on a data team.
Faster First Drafts
Start from generated logic, then refine it in the concept editor instead of writing from a blank page.
Migrate Existing Logic
Translate SQL, RDF or OWL into Vadalog and keep it in one governed model.
NL Query Answering
Our published work combines semantic awareness with LLMs to answer natural-language questions over enterprise knowledge graphs.
Browse the papers →“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
“Prometheux enabled us to build a system that can access our databases, process all the client's historical variables, and provide an intelligent, understandable explanation autonomously.”
Massimiliano MalinHead of Credit Risk · Banca Sella
Questions,
Answered
What does the AI Assistant generate?
Vadalog concepts over your connected data. It saves and runs them and repairs failed code. To translate SQL, RDF or OWL into Vadalog, use the concept editor or the Vadalingo API.
Do I need a semantic model before I can ask questions?
Questions are answered against your ontology, which plays that role and also holds executable rules. If you don't have one yet, the AI Assistant can help build it from your connected sources. Existing SQL, RDF or OWL can be translated into Vadalog with the concept editor or the Vadalingo API.
How do I know the answer is right?
The assistant saves the logic behind each answer as a concept you can read and edit. Once saved, it runs deterministically, so the same question over the same data gives the same result, traced back to the rules and source rows that produced it.
Can it change my rules without asking?
It asks before making changes and always before deleting anything, such as a schema change or an app.
What can the assistant see?
It is scoped to your data, Vadalog and the Prometheux platform, works with your own access rights and declines unrelated requests.
Which data can I ask questions over?
Files, relational databases, Snowflake, Databricks, Redshift, BigQuery, Hive, Presto, Neo4j and HTTP APIs, with the data left in place. One concept can combine several sources.
Can we put it inside our own tools?
Yes. The Agent API and the Bring Your Own Frontend guide in the docs cover embedding the assistant in your own interface, and agents can call saved concepts via APIs or MCP.
Related Capabilities
Ask Your Data
Anything
Bring a question your team keeps asking. We'll turn it into a concept over your own sources, live.
