Xore Technologies

XoreQL · conversational analytics

Ask your data a question. Get an answer you can check

XoreQL turns plain English into SQL, runs it against your live database, and shows the result as a chart. It also shows the tables it used, the joins it made and the SQL it wrote — so the answer is auditable, not just plausible.

100%
Self-hostable
4
LLM providers supported
0
Lines of SQL required

The challenge

Your data is fine. Getting at it isn't.

The information people need already exists in the ERP. The bottleneck is the small number of people who can write the query.

SQL is the gate

Business users can't write it, so every question becomes a ticket for the two people who can.

Scattered sources

The ERP, a side database and a spreadsheet someone maintains. No single place to ask a question that spans them.

Days of turnaround

Report requests queue up. By the time the answer arrives the decision has already been made without it.

Cloud AI is a non-starter

Sending your schema and results to an external model is a compliance conversation most teams would rather not have.

Trust

The hard part isn't the SQL. It's believing the answer.

Any model can produce syntactically valid SQL that answers a subtly different question. XoreQL is built so a person can tell the difference — every answer carries the working out.

Lineage on every answer

The tables read, the joins made and the business terms that fired — shown under the result, not buried in a log.

A confidence signal

Derived from whether table selection needed a fallback and how well the question grounded in your knowledge pack. Low confidence says so.

View and edit the SQL

Power users open the generated query, adjust it and re-run — without leaving the conversation.

"This looks wrong" feeds the model

A correction becomes a taught term or join hint, so the same misreading doesn't happen twice.

Copilot-learner — teach a business term
When we say "net sales" we mean invoiced value minus credit notes

Learned mapping

Termnet sales
ExpressionSUM(inv) − SUM(cn)
TablesArTrnDetail
Scopeclient: acme
Term saved
Applies to every future question from this client
Taught terms, corrections and join hints compound over time

Capabilities

What's in the platform today

A conversational front end, a multi-stage query pipeline behind it, and the governance to put it in front of a whole organisation.

xoreql.com

Conversational interface

Ask, then refine. XoreQL keeps conversation context and narrows scope across follow-up questions, streaming the answer as it forms.

Multi-stage query pipeline

Intent detection, table selection, SQL generation and a focused refine pass — each stage narrowing the problem instead of one shot at the whole thing.

Schema browser & join builder

Browse tables, columns and row counts, drag them into the chat, or build joins visually on a canvas with INNER, LEFT, RIGHT and FULL — SQL generated for you.

Auto-charts & dashboards

Bar, line, pie or KPI chosen automatically from the result shape — then pin any answer as a dashboard tile that re-runs live each time it's opened.

Slack & Teams

Ask XoreQL where your team already works. A slash command or mention returns the answer and the chart in the channel, with a link through to the full result.

Databases and files

Point it at SQL Server, PostgreSQL or MySQL, or upload a CSV, Excel, Parquet or JSON file — the question you ask is the same either way.

Governed access

Denied tables, column restrictions and row-level policies are enforced before a query runs. Every request is written to an audit log.

Multi-tenant by design

Per-client isolation, per-client knowledge packs and per-client model configuration — one deployment serving many organisations cleanly.

The engine

Self-hosted models, grounded in your schema

Generic text-to-SQL fails on real enterprise schemas because it has never seen your table names. XoreQL solves that with retrieval over your own metadata, plus terms you teach it.

RAG schema intelligence

Extract

Tables, columns and foreign keys pulled from your live schema.

Embed

Indexed as vectors in Qdrant so schema search is semantic, not string matching.

Retrieve

Only the relevant slice of schema is injected into the prompt for each question.

Knowledge packs

Domain packs — SYSPRO among them — pre-load the vocabulary and relationships of a system so the model starts grounded rather than guessing from column names.

  • Business terms mapped to real expressions
  • Join hints for the relationships the schema doesn't declare
  • Corrections captured from users, applied per client

Bring your own model

Each client picks a provider. Keys are stored encrypted, provider errors are never leaked to end users, and switching providers is configuration — not a redeploy.

Ollama

Default

Fully self-hosted on your own hardware. No per-token cost and nothing leaves your network.

Claude

Anthropic's models where you want maximum reasoning quality on hard, ambiguous questions.

OpenAI

GPT models via API for teams already standardised on them.

Gemini

Google's models as an alternative managed option.

SQL Server PostgreSQL MySQL Qdrant SSE streaming Encrypted keys
Available now

Not everything lives in the database

Half the questions people actually ask involve a spreadsheet someone maintains on the side. XoreQL brings the same conversational interface to uploaded files — with the governance to keep that safe.

  • Upload CSV, Excel, Parquet or JSON and query it in plain English, exactly as you would a database
  • Schema detected automatically on ingest — no mapping step
  • Private by default; sharing across the organisation goes through an approval queue
  • Per-user and default upload size limits, configurable by admins
  • Admin preview of any source before it's approved for wider use

Document question-answering over PDFs and Word files, and joins that span a spreadsheet and the ERP, follow in the phases after.

Data sources
syspro-prod Shared · SQL Server
warehouse-pg Shared · PostgreSQL
Q3_freight_rates.csv Private · uploaded
supplier_scorecard.xlsx Pending share

Admin review queue

supplier_scorecard.xlsx · 2.4 MB Preview → approve
Uploads are private until an admin approves them for wider use

Roadmap

Where XoreQL is going

The direction is the same throughout: from "ask a question, get a table" to self-service analytics your whole organisation can be trusted with.

Available now

  • Natural language → SQL
  • Uploaded file data sources
  • Auto-charts & pin to dashboard
  • Confidence & lineage on answers
  • Editable SQL panel
  • Slack and Teams bot
  • Taught terms & join hints
  • RAG schema retrieval
  • Visual join builder
  • Per-client LLM provider
  • Access control & audit log

Shipping next

  • Result sanity checks
  • Disambiguation prompts
  • Searchable query history
  • Feedback analytics

On the roadmap

  • Multi-tile dashboards with filters
  • Saved & shared report library
  • Scheduled delivery
  • Threshold alerts
  • Document Q&A over PDFs
  • Cross-source joins

Later

  • Auto-insights & anomalies
  • Suggested questions
  • Data catalog browser
  • SSO, SAML and SCIM
  • Query cost governance
  • Embeddable widget & public API

Try it against your own schema

XoreQL is live at xoreql.com. Or send us a question your team asks every week and we'll show you what it looks like answered in seconds.