Ask your database anything. Get back a chart.
Datagini plugged into your Postgres database, turned plain-English questions into SQL, ran it, and handed back interactive charts. All on your own AI key.
GPT-4o
Revenue is up 22% over the last 6 months, with the sharpest jump in May.
SELECT date_trunc('month', created_at) AS month,
SUM(amount) AS revenue
FROM orders
GROUP BY 1
ORDER BY 1;[ 01 ] How it worked
From connection string to chart.
One question, followed all the way through. Scroll to watch it move.
[ 01 ]
Connect your database
Paste a Postgres connection string. Datagini tests it, encrypts it, and keeps it on your project. Nothing else to install.
Connect a database
Supabase, Neon, Railway, Render, or your own Postgres.
postgresql://user:password@host:5432/databaseThe connection string is encrypted before it's stored.
[ 02 ]
It reads the schema
Tables, columns, types and foreign keys come straight from information_schema, so the model knows what actually exists before it writes a line of SQL.
FROM information_schema.columns
orders.customer_id → customers.id
order_items.order_id → orders.id
order_items.product_id → products.id
Schema context ready · 6 tables · 36 columns · 3 keys
[ 03 ]
You ask in plain English
Type the question the way you'd ask a teammate. Leave the layout on Auto, or force one of 13 visualizations from the picker.
Ask anything about acme-prod.
Ask a question about your data...
Llama 3.3 70B[ 04 ]
An agent writes and runs the SQL
A small SQL agent calls run_sql against your database, reads the rows it gets back, and writes a short answer in plain English.
sql_analyst started
tool call · run_sql
SELECT c.city, SUM(o.amount) AS revenue
FROM orders o
JOIN customers c ON c.id = o.customer_id
WHERE o.created_at >= date_trunc('quarter', now()) - interval '3 months'
AND o.created_at < date_trunc('quarter', now())
GROUP BY c.city
ORDER BY revenue DESC
LIMIT 5;5 rows · 38 ms
Answer written, picking a layout
[ 05 ]
The answer renders itself
Instead of slow, fragile generated code, the model returns a tiny JSON spec. Prebuilt, themed components turn it into an interactive chart in milliseconds.
Model returns a tiny spec, not code
[ 02 ] Generative UI
13 ways to see an answer.
The model picked a layout for every result, or you forced one yourself. Hover the charts, sort the table, click through the layouts.
›Which products earned the most this quarter?
Ranked list with bars{ "layout": "leaderboard", "label": "product", "metric": "revenue", "highlightTop": 1 }[ 03 ] Schema explorer
Your schema, laid out clearly.
Every table, column and foreign key, mapped into a diagram the moment you connect.
Hover a key to trace the relationship
[ 04 ] Bring your own key
Your keys. Your data.
Plug in OpenAI, Claude, Gemini, or any OpenAI-compatible endpoint. A free Llama 3.3 70B through NVIDIA NIM came built in.
- API keys encrypted at rest, per project
- Connection strings encrypted before they're stored
- Queries run against your database, not a copy of it
- Switch models per project, no lock-in
[ -- ] Epilogue
This one goes on the wall.
A few months after I built it, AI assistants learned to do all of this right in the chat. That's how building feels right now: tools go out of date month by month. Datagini stays up as a snapshot of the moment it still needed solving.