Kanvaz AI Labs
All solutions

Your data middleware

GlossDB

Align definitions, mappings and unit conversions before data reaches planning. Keep your existing systems.

How it works

01

Agree the meaning

Keep product definitions, mappings and unit conversions in one place.

02

Approve once

GlossDB suggests mappings. Your experts confirm or correct them for reuse across loads and markets.

03

Check every load

Flag deviations by market with an accountable owner and data quality reports.

~/example

One market. One data domain.

Validate test data in your existing import format before scoping a wider pilot.

GlossDB

schema · declared aspects

DECLARE ASPECT unit WITH $${
  "type": "object",
  "properties": {
    "value": {"type": "string"},
    "source_column": {"type": "string"}
  }
}$$ AS FACT;

DECLARE ASPECT sell_out_units WITH $${
  "description": "Units sold to shoppers, by week",
  "indicators": ["sku_id", "market", "week"]
}$$ AS QUERY;

DECLARE ASPECT min_max WITH $${
  "type": "object",
  "properties": {"min": {}, "max": {}}
}$$ AS MEASUREMENT ON COLUMN;

The vocabulary is declared up front. Nothing can be written against an aspect that was not declared.

data contract · grounding

GLOSS sell_out_units ON pos.weekly_sales AS $${
  "sql": "SELECT sku_id, market, week,
          SUM(qty * pack_size) AS units
          FROM weekly_sales GROUP BY 1, 2, 3",
  "assumptions": [
    {"dimension": "unit",
     "assumption": "qty is in cases; pack_size converts to each",
     "basis": "column_stats", "confidence": 0.75}
  ]
}$$;

Each metric carries its SQL and every assumption, with how sure the agent is. Unsure assumptions go to your team to confirm or correct.

replay · commit history

  1. Mar 02
  2. Mar 02
  3. Mar 09
  4. Mar 10
  5. Mar 16
commit 5 · agent
New load checked against the ruling
snapshot
s-0458
definition in force
SUM(qty × pack_size)
status
holds · 0 difference

Every statement is one commit: who wrote it, person or agent, and the data snapshot it was measured on. Scrub back to see what planning was told at any point.

Written in glossQL, the open-source language behind GlossDB. Sample retail data.

Open core. Your infrastructure.

deployment · inside your VPC

enterprise VPC / on-premise · data stays here

sources

  • SAP ERP
  • POS & retailer feeds
  • EDI orders

kanvaz

  • GlossDB · semantic layer
  • Causal Engine

integration

  • Your integration layer

planning

  • SAP IBP
  • OMP
  • Kinaxis

only outside call → your identity provider (sign-in)

GlossDB and the Causal Engine run self-hosted next to your systems. Planning tools keep receiving data through your existing integration. Illustrative setup.

Faster and far less expensive than typical alternatives.

Time to value
6–8 weeks
vs 12–24 months
Cost
Open-source core
vs multi-million licences
System impact
Non-invasive
vs rip-and-replace
Upkeep
Declarative contracts
vs brittle pipelines

~/spec

Technical specification

The glossQL language specification, the normative reference for how GlossDB works.

Read SPEC.md

~/security

Security posture

Self-hosted on your own infrastructure. Your data stays there; the only outside call is to your identity provider.

Read the FAQ