datumara

Open-source analytics intelligence

The faster path
from question to insight.

Datumara is the open analytics model that turns business questions into schema-aware SQL and decision-ready answers.

Try Datumara Get the model
MIT licensedLocal-firstOllama ready
datumara / query local
01schema → retail_warehouse
orders(id, customer_id, total, created_at)
customers(id, segment, region)
04question → Which regions grew fastest last quarter?
-- generated SQL
SELECT region, SUM(total) AS total_revenue
FROM orders
JOIN customers ON orders.customer_id = customers.id
WHERE created_at >= '2026-04-01'
GROUP BY region
ORDER BY total_revenue DESC;
✓ schema-aware response local inference

Ready to try it?

Install in one command

$curl -fsSL https://raw.githubusercontent.com/achagani/datumara/main/install.sh | bash

Then run: ollama run datumara-local

01Ask in plain language
02Generate grounded SQL
03Move from data to action
04Keep control of your data

An analytics model, not a chatbot

Less hunting.
More knowing.

01 / ground

Knows the shape of your data

Pass a schema, ask a question, and get a query grounded in the tables and relationships that actually exist.

schema → context → SQL
02 / answer

Built for real analysis

Go beyond lookup questions. Explore segments, trends, cohorts, anomalies, and the metrics your team uses every day.

questions → signals → decisions
03 / own

Your data stays yours

Run locally with Ollama or deploy on infrastructure you control. The model fits into your workflow instead of replacing it.

private / portable / open

Choose your starting point

Small enough
to start now.

Use the local model to explore the experience, then move to a higher-capacity checkpoint when your workload demands it.

ProfileUse caseRuntimeStatus
Datumara Local 1.1BPrivate local analyticsOllama / 4 GB class Available
Datumara SQL 3B classProduction SQL workflowsHosted GPU Coming next
Datumara Scale 7B+Deep analytical workloadsHigh-memory GPU Roadmap

Prefer to install manually?

Download from
GitHub Releases.

Or skip the script and pull directly from Ollama. Either way, you're one command away from your first insight.

Browse releases
$ollama pull datumara-local
$ollama run datumara-local

For builders

Built in the open.
Worth looking under the hood.

Training recipes, hardware detection, progress reports, and evaluation tooling are available for people who want to adapt Datumara to their own data.

Read the training notes

Early access

Shape the next
generation of analytics models.

Join the waitlist for upcoming Datumara SQL (3B) and Scale (7B+) models. Get notified when new capabilities launch and help prioritize features for your workflow.

Join the waitlist

Be first to know when new models launch. No spam, unsubscribe anytime.

Built for people who ask
the next useful question.

Datumara / 2026github.com/achagani/datumara ↗