Model
Define dimensions, measures, joins, calendars, hierarchies, and glossary terms once — versioned and deployed like software, so what is served is always what was approved.
Governed analytics for BI, Excel, and AI
One trusted answer for every tool — at a fraction of the cost.
Tessallite is the governed layer between your tools and your data. Define every metric once, and dashboards, Excel, SQL clients, APIs, and AI agents all receive the same approved number — served in moments from automatically maintained summaries instead of full-table scans. On the published TPC-DS SF1000 benchmark, the simple everyday question dropped from $330 to 5 cents per 1,000 runs (accelerated vs unaccelerated on the same source; independent BigQuery tie-out was not measured).
- Use AI to build and maintain data dictionaries and business glossaries.
- Build intelligent aggregation layers that deliver sub-second analytical performance over datasets with billions of records — running on conventional relational database technology.
- Create governed semantic models through an intuitive web interface.
- Interact with your data through a trusted conversational AI agent built on semantic generation — not direct SQL generation, indiscriminate context stuffing, or other hard-to-govern techniques.
- Access and analyse large-scale data directly from Microsoft Excel, putting the technology in front of business users in the tool they already use.
The hosted trial at cloud.tessallite.io is the SF1 demo (about 1 GB). The published terabyte / SF1000 result is a separate one-off test, not what you get when you open the trial.
Published SF1000 simple question: $330 → 5¢ per 1,000 runs (ON vs OFF)
66 GB → a few KB scanned per question
Identical answers, verified against the source
Excel, BI, SQL, API, and AI — one governed model
What Tessallite is
The place your numbers get their meaning.
Tessallite is a semantic layer and query acceleration platform. To your tools it looks like a database. To your business it is where every important number has one approved definition, one access policy, and one fast path — modelled once, inherited by every dashboard, workbook, and AI agent.
Govern
Roles, personas, row security, and column controls are written once and enforced everywhere automatically — never re-implemented per tool.
Route
Every query takes the fastest path that is provably exact — aggregate, pocket table, or source. Speed never changes the answer.
Explain
Owners, freshness, lineage, and route detail travel with every result — so when someone asks where a number came from, there is an answer.
See it working
Real screens. Real answers.
This is the product, not a mock-up: the modelling canvas, live query routing, and Excel running on the governed model. The MCP surface is Beta (unit-tested; acceptance still in progress). The Product Tour walks through every screen.
Conversational agent
Ask in plain business language and get a written answer, a chart, the underlying rows, and the exact query it ran. The agent answers only through the governed model — it cannot invent numbers.
Query routing
Watch a query get parsed, bound, and routed to a maintained summary — and see exactly why it came back fast.
Excel workflow
Analysts keep their workbooks and PivotTables — now running on governed definitions over terabyte-scale sources, no SQL required.
Enterprise control
Security rules travel with every answer.
One set of access rules, enforced at one chokepoint. Whether the question arrives from Excel, a BI tool, SQL, an API, or the AI agent, the same row, column, and persona policies decide what comes back.
Audience-aware models
Personas shape exactly which measures, dimensions, hierarchies, and filters each audience sees — and administrators can preview any audience’s view before it ships.
Row and column controls
Row security, data tags, and column restrictions hold across every protocol. A field hidden in the dashboard is hidden in Excel, SQL, and chat too.
Audit and trace
Every answer carries its evidence — query log, route, owner, freshness, and lineage — ready for any reviewer who asks.
Performance, measured
A terabyte benchmark. Answers for pennies on the simple question.
Tessallite watches your query patterns, builds the right summary tables automatically, and keeps them fresh. On the published TPC-DS SF1000 run (BigQuery), the simple everyday question that cost $330 per 1,000 unaccelerated runs came back for 5 cents when accelerated — identical to the unaccelerated Tessallite route, not an independent BigQuery baseline. Medium and heavy questions in the same mix cost more; do not annualise the simple bar across every dashboard. The hosted trial at cloud.tessallite.io is the SF1 demo, not the terabyte dataset.
Aggregates
Rollups for the questions your teams actually ask — created, refreshed, and retired automatically. Nobody hand-maintains a summary table, ever.
Pocket tables
Hot slices of big fact tables, cached close to the query, so repeated filtered analysis stops re-scanning the source.
Exactness, guaranteed
If a prepared answer cannot be exact for a query, Tessallite goes to the source and gets the right one — then learns from the miss. Fast never means wrong.
Connected tools
Your tools already speak its language.
Each deployed model is simultaneously an OLAP cube for Excel and Power BI, a PostgreSQL database for any SQL client, a REST API for your applications, and grounded context for AI — defined once, identical everywhere, nothing to install on the client.
Excel and Power BI
Connect natively over XMLA: browsable cubes, friendly field names, drillable hierarchies, and refreshable workbooks — no add-ins required.
SQL and BI tools
Any PostgreSQL-compatible client just works — DBeaver, Tableau, pgAdmin, psycopg2, or your own drivers — and sees only the governed model.
Apps and agents
Headless APIs, embeddable chat, and AI-assistant integration put governed answers inside your own products and workflows.
PostgreSQL
BigQuery
Spark (preview)
Redshift (preview)
Day two and beyond
Built to be operated, not just admired.
A semantic layer earns trust in month six, not day one. Model health, query logs, usage analytics, impact analysis, schema-change review, refresh history, alerting, and export/import are all in the product — so the layer keeps its promises long after launch.
Model health
Catch invalid dimensions, broken measures, failed aggregates, and refresh issues before a single user hits them.
Usage and impact
Know which questions, reports, and columns are actually used, so every model change is reviewed with evidence instead of guesswork.
Export and import
Promote whole models and projects between environments — credentials never travel, connections rebind cleanly on arrival.
Evaluation to production
Prove it on your own infrastructure. Scale when it wins.
Community is free and self-hosted, with the core modelling, BI access, governance, and acceleration features shared with Enterprise. Enterprise lifts the scale limits and adds SSO (Beta), support, and production deployment options. Some connectors and integrations are in beta or preview — see the capability maturity notes.
Community
The core platform on your own servers, free — delivered as a signed bundle you can verify before you run it.
Install guide
Docker Compose, one installer, and minutes from download to a running stack. Add the built-in demo model with one flag, or connect your own source.
Enterprise
Unlimited tenants, models, and users; SSO (Beta); SLA-backed support; and guided production deployment.
Give every team the same trusted numbers.
See it on the SF1 hosted trial — or install Community today and prove it yourself.
The hosted trial at cloud.tessallite.io is the SF1 demo, not the published terabyte / SF1000 benchmark.