SelectStar
Use cases

Three roles. One agent. Three different wins.

The same graph orchestrator serves an analyst asking "how many orders by status", an engineer debugging a production issue, and a data scientist running a quick clustering pass. Here's what each one gets.

For Analysts

Answer any data question without writing SQL

Analysts shouldn't have to context-switch between a BI tool, a SQL editor, and a charting app just to answer one question. SelectStar handles the whole pipeline: you ask in plain English, the agents decide what to query, how to profile, what to chart, and how to summarize. You read the answer and the canvas artifacts, then ask the next question. Drop a CSV in if you don't have a database handy — Classic Mode has you covered.

The pain today

  • ·Every data question means a SQL query, a chart, and a writeup — three tools
  • ·Self-serve BI dashboards never have the exact view you need
  • ·Spreadsheet analysis means a notebook, a pandas import, and a chart library

With SelectStar

  • Ask in plain English — get tables, charts, and prose in one turn
  • Iterate on questions without rewriting SQL or rebuilding charts
  • Drop a CSV/XLSX in Classic Mode when there's no database to connect
  • Export CSV/SVG/PNG straight from the canvas to drop into slides
localhost:3000/selectstar
● live
SelectStar in use — For Analysts
SelectStar in use by for analysts
For Engineers

Inspect and debug a database without leaving the chat

Engineers get paged for slow queries, schema drift, and unexpected data shapes. SelectStar gives them a single-pane view: ask 'what changed in the orders table' and the Schema agent answers from the cached snapshot; ask 'profile the last 1000 orders' and EDA returns stats and null percentages in seconds. Zen mode means you can investigate production writes safely — every write is gated, every resolution audit-logged.

The pain today

  • ·Investigating a production issue means juggling pgAdmin, a SQL editor, and a Python notebook
  • ·Read-only tools don't let you test a fix; write tools don't audit
  • ·Schema docs go stale the moment someone ships a migration

With SelectStar

  • Ask structural questions, get instant answers from the cached schema
  • Profile data and spot anomalies with one prompt
  • Zen mode lets you test writes safely — every action audit-logged
  • Dry-run a DELETE, see the affected row count, roll back — no commitment
localhost:3000/selectstar
● live
SelectStar in use — For Engineers
SelectStar in use by for engineers
For Data Scientists

Quick models and forecasts without spinning up a notebook

Data scientists often need a quick regression, a clustering pass, or a trend forecast — not a full notebook. SelectStar's ML agent runs OLS regression via normal equations, k-means with k-means++ initialization, and linear-trend forecasting, all in pure TypeScript. No Python runtime, no scikit-learn install, no kernel to restart. Get the metrics, decide whether to escalate to a real notebook.

The pain today

  • ·Spinning up a notebook for a 30-second clustering pass is overhead
  • ·Moving data between a SQL tool and pandas breaks the flow
  • ·Sharing results means exporting, naming, and uploading files

With SelectStar

  • Run OLS, k-means, and forecasting inline — pure TypeScript
  • Results land in the canvas with metrics and predictions
  • Share the session link — colleagues see the full agent trace
  • If the quick model is interesting, escalate to a real notebook
localhost:3000/selectstar
● live
SelectStar in use — For Data Scientists
SelectStar in use by for data scientists
Common patterns

Three ways teams wire SelectStar into their workflow

Whatever your stack looks like, there's a one-line way to drop SelectStar in. These are the patterns we see most often in production.

Self-serve analytics for support teams

Hand SelectStar to your support team with a read-only connection string. They can answer 'how many users hit this bug' without pinging engineering. Zen mode keeps them safe.

# Read-only Postgres role
postgres://support:***@db:5432/app?role=read_only

# Support team opens SelectStar,
# types a question, gets the answer.
# No SQL knowledge required.

On-call triage with Zen mode

On-call engineer gets paged at 3am. Opens SelectStar, asks 'profile the last 1000 failed orders', gets stats and a chart in 4 seconds. If they need to test a fix, Zen mode gates every write.

# Connect with write privileges
postgres://oncall:***@db:5432/app

# Toggle Zen mode ON
# → writes are gated
# → dry-run shows row count
# → rollback cancels
# → audit log records everything

Quick ML triage without a notebook

Data scientist wants a quick clustering pass on customer segments. Instead of spinning up a notebook, they ask SelectStar. If the result is interesting, escalate to a real notebook.

# Ask in plain English:
"Cluster customers into 5 groups
 based on order frequency
 and total spend"

# ML agent runs k-means++
# → canvas shows centroids
# → predictions table
# → R² and inertia metrics

Pick the pattern that fits your team

Open the live app, try the demo database, then connect your own. Zen mode keeps every write safe.

Try the appSee features