Demo Walkthrough

Run :GripStart to open the Softrear Inc. Analyst Portal. The demo database ships with seventeen base tables and fictional narrative data about a consumer goods company with some unusual patterns in the incident records.

The demo opens as a SQL notebook in the query pad. Place your cursor inside any sql block and press <C-CR> to run it. Results appear in the grid below. Press gn to open a notebook file picker for .md and .sql files in your project.

Setup

:GripStart

This creates a local DuckDB database when the duckdb client is available and falls back to SQLite through sqlite3. The main notebook requires either client, and its supplier federation section requires both. Dadbod Grip opens the Softrear Inc. notebook automatically, with seventeen base tables and an incident summary view available in the sidebar. The cursor starts in the first SQL block; opening the demo does not execute it.

Daily practice

Search the notebook for Daily practice to try this short loop with disposable data:

find -> inspect -> filter -> edit -> review SQL -> undo / apply

Find, inspect, and export a view

Press gb to focus schema, then H to collapse it. Press /, enter incident_summary, and confirm. Search reveals matching objects inside closed groups. Enter opens the view. To clear the sidebar search, press gb, then F to restore the previous folds. Press 3 from schema to return to the result. Both demo engines provide this aggregate view. Browse and export it rather than editing it.

In the result, move to max_severity and press s to sort. Press K to inspect a row vertically and q to close the float. Press gX to export, then choose the current page or all matching rows. Shortcut hints fit in two rows and remain visible when the table scrolls horizontally; ? shows the remaining keys.

Stage one reversible edit

Run :Grip consumer_incidents. Press <C-f>, enter id = 1, and confirm. Move to severity, press i, replace 10 with 9, and press Enter. This stages the edit; the database still contains 10.

Press gs to inspect the UPDATE, close the preview with q, and press u to undo. To try a write, repeat the edit, review it with gs, and close the preview with q. Then press a and confirm the write. Press r to refresh. Running :GripStart again restores the original demo data.

The notebook’s Stage a new lead from a pipe section also demonstrates :GripImport !command: preview stdout as rows, stage the batch, review its INSERT statements with gs, and undo the whole batch with one u or apply it with a. The pipe command runs only when you explicitly provide it.

Keep useful SQL and results

From results, press 2 to return to SQL. Press gz to zoom it and again to restore the splits. Inside SQL, 2 opens history; gh also opens history from the main panes. Press gw to return to the grid, then pin a result with gL and use gJ to switch retained results.

To save one query, use :GripQuery and replace the pad with just that SQL, then press <C-s> and name it. Saving captures the entire pad, so do not save the investigation when you intend to keep a single query. gq loads saved queries. Named queries and history persist across restarts; unsaved pad contents and the active result layout do not automatically restore.

Explore engine-specific objects

SQLite’s object list is flat. With DuckDB, expand reporting using l and open reporting.monthly_incidents under Views. To fold it afterward, press gb, move to reporting, and press h. Press 2 from schema to return to SQL and run these statements separately:

SELECT month, incident_count, highest_priority
FROM reporting.monthly_incidents
ORDER BY month

SELECT reporting.severity_label(4) AS low_priority,
       reporting.severity_label(8) AS high_priority

DuckDB macros are callable in SQL; Dadbod Grip does not list them as PostgreSQL functions. DuckDB file partitions are not PostgreSQL child-table partitions.

For an optional local PostgreSQL lab, follow the plugin’s demo guide. It requires a local server and psql, plus a new disposable database. The lab creates a fresh grip_demo schema and rejects an existing one. It exercises nested partition discovery, the open_incidents view, and function-source inspection. Enter on a function opens its definition without executing it; call the function explicitly from SQL when you want a result.

The investigation

Three questions:

1. Is there a product quality problem, and where does it originate?
2. Who knows what, and what has been done about it?
3. What decisions drove the company here, and how were they made?

1. Product catalog

DuckDB loads 505 rolls; the compact SQLite fallback has 35. Both include the named evidence used here. Sort by softness score ascending to see the edges:

SELECT sku, ply, softness_score, tensile_strength, discontinued
FROM rolls
ORDER BY softness_score ASC
LIMIT 20

ULTRA_BUDGET_XTRM surfaces at the top: softness_score = 0.0, discontinued = false. A product with a zero softness score exists in the catalog and it ships.

2. Consumer incidents

How bad are the incidents, and does a pattern emerge in the worst ones?

SELECT incident_type, severity, roll_sku, notes
FROM consumer_incidents
ORDER BY severity DESC
LIMIT 30

Every row with severity = 10 has incident_type = 'airplane', without exception. Airline bathrooms are a structural failure mode for this product category.

3. The supply chain in one query

Follow the foreign key chain from a consumer incident all the way to the supplier without leaving the query pad:

SELECT
  ci.roll_sku,
  ci.severity,
  pb.quality_score,
  pb.recall,
  f.name          AS facility_name,
  f.vibe_score,
  bc.alias        AS supplier_alias,
  bc.our_relationship
FROM consumer_incidents ci
JOIN rolls r               ON r.sku = ci.roll_sku
JOIN production_batches pb ON pb.id = r.batch_id
JOIN facilities f          ON f.id = pb.facility_id
JOIN bamboo_cartel_members bc ON bc.id = f.bamboo_supplier_id
WHERE ci.roll_sku = 'ULTRA_BUDGET_XTRM'
ORDER BY ci.severity DESC

The chain: recalled batch, Shanghai Liaison Office, supplier currently under embargo. The root cause is four foreign keys deep and visible in a single result set.

4. What they told themselves

SELECT directive, publicly_acknowledged, issued_by
FROM leadership_directives
ORDER BY publicly_acknowledged ASC

Three directives were never publicly acknowledged, and each one describes how to rename the problem in customer communications. The only acknowledged directive states: “Greg is authorized to self-certify all SKUs. This is intentional.”

What you just used

• Query pad with notebook mode: <C-CR> runs the SQL block under the cursor
• Notebook picker: gn opens .md and .sql files as notebooks
• FK trail: a single JOIN query traverses four hops instead of four separate gf presses
• Grid: each <C-CR> result opens as a sortable and filterable grid below. Press gO when a simple single-table result can be reopened for editing.

The same controls apply to your own database, subject to its adapter and permissions.

Go deeper

The full notebook lives at demo/softrear-internal.md inside the plugin directory. It includes the supplier intelligence cross-database join, the BambooKnows investigation, and the pricing arrangement that created the incentive to mislabel incoming shipments.

From a checkout of the plugin, gn finds it under demo/. In another project, :GripStart loads the bundled copy directly and recreates the disposable data.