A semantic model, not a copy
Semavex does not copy your tables into a store of its own. It keeps a semantic model: Vexon drafts the tables, relationships and measures from a plain-language description of your business, and your team reviews it.
Live database reporting
Reporting and analytics on your live database, without a data warehouse
Reporting projects usually start with a data warehouse. For many mid-market teams, the numbers they need already sit in one transactional database. Semavex reports on that database directly: no pipelines, no copy of the data, no SQL to write. Vexon, the AI agent in Semavex, builds the model, answers the questions and delivers the reports.
The usual path
Why the warehouse comes first, and what it costs
A warehouse brings data from many systems together and keeps heavy analytics away from the database your application uses. Getting there takes ETL pipelines, a modelling layer, a BI tool and people to maintain all of it. Each step adds a delay between a transaction and the report that shows it.
When most questions are about one operational database, such as a ledger, a billing system or a gaming platform, that stack is a long way round. The data team spends its time keeping pipelines running, and the people who need the numbers still wait for an analyst.
The Semavex way
How Semavex reports on live data safely
Plans, never generated SQL
Vexon turns each question into a structured query plan. Semavex validates the plan against your semantic model and a whitelist, then translates it into parameterized SQL with deterministic code.
Queries that respect production
Each query is a read, planned around your real indexes and stopped if it runs past its time limit. By default, reports never block your application's writes.
Current every time it runs
Saved and scheduled reports query the database when they run, so the numbers match what your application sees, with no sync lag.
Platform
What your team gets
- Saved reports with subtotals, grand totals, number formats and conditional formatting
- Scheduled email delivery as PDF, CSV or Excel
- Period comparisons, rolling periods, running totals, rankings and top-N
- AI explanations from exact statistics over the whole result
- Report chat: change a saved report by asking
- Data alerts on your thresholds, as often as every 15 minutes (Enterprise)
Being honest
When a warehouse still makes sense
Reporting on the live database is not the answer to everything. Keep or build a warehouse when:
- One report has to join data from several databases or SaaS tools. Semavex queries one database per question.
- You need history your transactional database does not keep, such as daily snapshots of balances that get overwritten.
- Your reporting load is heavy enough to compete with your application. Point Semavex at a read replica first: it is often all you need.
Semavex also runs next to a warehouse and the BI tools you already have, for the questions and recurring reports on operational data that would otherwise wait in an analyst's queue.
Guides by database
FAQ
Questions about reporting on a live database
Is it safe to report on a production database?
It is when the queries are controlled. Semavex only sends read queries, built by deterministic code from a validated plan, planned around your indexes and stopped at a time limit. By default they never block your application's writes, and we recommend connecting with a read-only account or to a read replica.
How current is the data in a Semavex report?
As current as your database. A report reads the data when it runs, whether you open it, refresh it or receive it on a schedule. There is no extract or nightly load in between.
Do we still need a data engineer?
Someone who knows your data should review the semantic model Vexon drafts and keep it current when the schema changes. There are no pipelines to build or repair, so the work is the model, not the plumbing.
Which databases does this work with?
PostgreSQL, SQL Server, MySQL and MariaDB. Snowflake is coming soon.
See it on your data
See Semavex on your own database
Book a demo and we will walk through the semantic model, a few reports and a schedule on data like yours.