admindb 1.2.0 → 1.2.1

package/README.md

@@ -1,82 +1,101 @@
 # AdminDB
 
-
 [![Version](https://img.shields.io/npm/v/admindb.svg)](https://www.npmjs.com/package/admindb)
 [![License](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
 [![Node.js](https://img.shields.io/badge/node-%3E%3D20.0.0-brightgreen.svg)](https://nodejs.org/)
-[![Publish](https://github.com/MIbnEKhalid/admindb/actions/workflows/publish.yml/badge.svg?branch=main)](https://github.com/MIbnEKhalid/admindb/actions/workflows/publish.yml)
 [![Downloads](https://img.shields.io/npm/dm/admindb.svg)](https://www.npmjs.com/package/admindb)
+[![Publish](https://github.com/MIbnEKhalid/admindb/actions/workflows/publish.yml/badge.svg?branch=main)](https://github.com/MIbnEKhalid/admindb/actions/workflows/publish.yml)
 
+**A browser-based SQLite administration tool.** Manage a SQLite database
+entirely from the browser — browse and edit rows, run SQL, design schemas,
+import/export and seed data — with no separate frontend app to build or deploy.
 
-A browser-based SQLite database administration tool. Manage a SQLite database
-entirely from the browser: browse tables with full CRUD, inline
-spreadsheet-style grid editing, type-aware filters, bulk row operations, a
-visual table designer, an arbitrary SQL query runner, saved named queries,
-SQL-dump export, a "generate SQL" preview mode that never executes, a schema
-editor, and multi-database support.
+- **No frontend build step.** Handlebars server-rendered UI + vanilla JS +
+  Tailwind/DaisyUI; one Express process serves pages, static assets, and the
+  JSON API.
+- **Standalone or embeddable.** Run it from the CLI against a file or a folder
+  of databases, or mount it inside an existing Express app under any path on
+  the same port.
+- **Safe SQL by construction.** Every identifier is quoted and every literal is
+  escaped when SQL is generated, and the "Get query / preview" modes only
+  produce SQL strings — they never execute.
 
-- **Backend:** Node.js + TypeScript (compiled to plain JS) on Express.
-- **Frontend:** Handlebars server-rendered templates, Tailwind CSS + DaisyUI,
-  plain JavaScript (no frontend framework).
-- **Database:** SQLite via [better-sqlite3](https://github.com/WiseLibs/better-sqlite3). Requires
-  **Node.js ≥ 20**.
+![Home dashboard](docs/screenshots/home.png)
 
-> ## ⚠️ Security warning — read first
->
-> **This tool exposes full, unauthenticated database AND filesystem access.**
-> Every page and API route (browse, edit, delete, run arbitrary SQL, change the
-> schema, export the whole database, and — in manager mode — browse the
-> filesystem to open database files) is available to **anyone who can reach the
-> server**.
->
-> - **No authentication or authorization is built in.** The routes are **not
->   protected**.
-> - **It is your responsibility to protect access.** Do **not** expose
->   AdminDB to the public internet or to untrusted networks.
-> - Recommended ways to protect it:
->   - bind the standalone server to `127.0.0.1` (`HOST=127.0.0.1`) and use it
->     only from your own machine, and/or
->   - run it behind a reverse proxy that requires authentication (Basic auth,
->     OAuth, mTLS, …) or inside a VPN / private network.
->
-> Treat AdminDB as if it were a remote `sqlite3` shell with write access.
+> More screenshots in [`docs/screenshots/`](docs/screenshots/): table browser,
+> query editor, table designer, insert/edit form, schema editor, and the
+> databases landing page.
 
----
+## 30-Second Start
 
-## Install
+No install needed:
 
 ```bash
-npm install admindb
+npx admindb
+# → AdminDB is listening on http://localhost:3000
 ```
 
-Requires **Node.js ≥ 20**.
+With no arguments the server starts in **manager mode**: a **Databases** landing
+page where you can browse the filesystem and open any SQLite database file
+(`.db` / `.sqlite` / `.sqlite3`), or create new ones. Or point it straight at a
+file:
 
-### Use as a standalone CLI tool
+```bash
+npx admindb ./data/app.db          # open one database
+npx admindb -d ./dbs               # manage a folder of databases
+```
 
-AdminDB ships a command-line server. Install it globally (or just run it with
-`npx` — no install needed):
+From a clone:
 
 ```bash
-npm install -g admindb
-admindb                    # starts the server → open http://localhost:3000
+git clone https://github.com/MIbnEKhalid/admindb.git
+cd admindb
+npm install
+npm run build
+npm start                          # serves http://localhost:3000
+```
+
+> Requires **Node.js ≥ 20**. Runtime dependencies are `express`,
+> `express-handlebars`, and `better-sqlite3`; the rest is the application.
+
+## Why This Exists
+
+AdminDB exists because SQLite deserves a proper web admin UI, and desktop tools
+(DBeaver, DB Browser for SQLite) live outside both your browser and your stack.
+Web admin tools like phpMyAdmin target MySQL/Postgres, not SQLite. This one is
+server-rendered, so there is no frontend build step and nothing extra to deploy
+— the same Express process serves pages, assets, and the JSON API. Run it
+standalone from the CLI, or mount it under any path of an existing Express app
+on the same port.
+
+## Install
+
+### Try it (no install)
+
+```bash
+npx admindb -p 8080        # run on port 8080 without installing
+npx admindb ./app.db       # open a database file directly
 ```
 
-Run it without installing anything:
+### Install as a standalone CLI tool
+
+Requires **Node.js ≥ 20**:
 
 ```bash
-npx admindb -p 8080        # run on port 8080
+npm install -g admindb
+admindb                    # starts the server → open http://localhost:3000
 ```
 
-The `admindb` command starts the built-in server and opens the web UI in your
-browser. See [Run the built-in server](#run-the-built-in-server) for all the
-flags — `--port`, `--open <file>`, `--dir <folder>`, `--readonly`, and more.
+The `admindb` command starts the built-in server and prints the URL to open in
+your browser. See the [CLI reference](#cli-reference) for `--port`,
+`--open <file>`, `--dir <folder>`, `--readonly`, and more.
 
-## Using as an npm package
+### Embed AdminDB in your own app
 
-AdminDB is an Express app you can mount inside your own application, under
-your own path, on the same port as the rest of your server.
+AdminDB is an Express app you can mount inside your own application, under your
+own path, on the same port as the rest of your server.
 
-### Minimal example
+#### Minimal example
 
 ```ts
 import express from 'express';
@@ -95,7 +114,7 @@ app.listen(3000);
 `createRouter(options)` returns a fully wired Express app (pages + JSON API +
 static assets + view engine). Mounting it is just `app.use('/path', router)`.
 
-### Options
+#### Options
 
 | Option     | Type                 | Description                                                    |
 | ---------- | -------------------- | -------------------------------------------------------------- |
@@ -124,7 +143,7 @@ app.use('/tools/db', createRouter({
 app.listen(3000);
 ```
 
-### Multiple databases
+#### Multiple databases
 
 Pass a `DbManager` to manage several database files — either from a directory,
 from an explicit list of file paths, or both:
@@ -137,9 +156,9 @@ const app = express();
 app.use('/admin', createRouter({
   manager: new DbManager(
     {
-      dir: './data',                                  // scan a directory
+      dir: './data',                                   // scan a directory
       files: ['/srv/legacy/app.db', './shared.sqlite'], // and/or explicit paths
-      readonly: true,                                 // open all databases read-only
+      readonly: true,                                  // open all databases read-only
     },
     createLogger('info'),
   ),
@@ -154,97 +173,13 @@ In multi-db mode:
   `/admin/app.db/tables/users` and `/admin/app.db/api/tables`,
 - file names that collide across sources are deduped (`name__2.db`).
 
-## Features
-
-- Browse every table and its rows (paginated, with primary-key awareness).
-- **Filter rows by column** — type-aware, per-column filter controls: foreign-key
-  dropdowns, boolean toggles, date and numeric range inputs, plus the exact
-  (`=value`), comparison (`>5`, `<=10`), prefix (`pre*`) and substring (plain
-  text) operators on text columns. Filters survive sorting and pagination.
-- **Inline (spreadsheet-style) editing** — double-click any cell to edit it in
-  place with a type-aware control (FK dropdown, boolean toggle, date picker,
-  number/text input). Changes are staged and highlighted in the grid, then
-  applied all at once in a single transaction, or discarded.
-- Insert and delete rows (full CRUD). FK columns become dropdowns.
-- **Bulk row operations** — select rows with checkboxes (or "select all"), then
-  **delete** them in one transaction (with a warning listing how many rows in
-  other tables reference them) or **export** only the selected rows as CSV/JSON.
-- **Export** table rows or query results as **CSV or JSON**, and **import a CSV
-  file** (or pasted CSV) into a table.
-- Create tables visually: name, type, primary key, not-null / unique,
-  default value, and foreign-key references.
-- Write and run arbitrary SQL — SELECTs render as a table, `COUNT` queries show
-  a readable summary, write statements execute and report affected rows.
-- Save named queries and reload them from a dropdown.
-- Export the whole database as a downloadable SQL dump (`CREATE` + `INSERT`).
-- **Get query / preview mode:** generate `CREATE` / `INSERT` / `UPDATE` SQL from
-  the UI without executing it.
-- **Edit schema:** rename the table, add / rename / drop columns (with
-  relationship-safety checks), and drop tables.
-- **Indexes:** create indexes (plain or unique, on one or many columns — pick
-  columns in order, with a live `CREATE INDEX` SQL preview) and drop them from
-  the schema editor; automatic SQLite indexes are protected.
-- **Related rows:** every table that has a foreign key pointing at a table gets
-  its own column at the end of that table's grid — each cell shows how many of
-  its rows reference that record; click it to open a nested table with all of
-  the referencing table's columns and data for that specific record.
-- **Read-only mode:** open the database(s) without write access — the file is
-  opened `SQLITE_OPEN_READONLY` + `query_only`, every write route returns `403`,
-  and the UI hides/disables all write controls and shows a banner.
-- **Standalone CLI / file browser:** `admindb` runs as a full web app — with no
-  arguments it opens a **Databases** landing page where you can browse the
-  filesystem and open any SQLite database file, or create new ones. Flags set
-  the port, open a file, manage a folder, and more (`admindb --help`).
-- **Multiple databases:** directory scanning and/or explicit file lists, each
-  with its own workspace under `/{db}/…`.
-
-## Screenshots
-
-Home dashboard — stat cards and the table list.
-
-![Home dashboard](docs/screenshots/home.png)
-
-Table browser — sortable columns, sticky header, and per-row actions.
-
-![Table browser](docs/screenshots/table.png)
-
-Query editor — line-numbered editor with a results table.
-
-![Query editor](docs/screenshots/query.png)
-
-Table designer — visual columns with a live SQL preview.
-
-![Table designer](docs/screenshots/designer.png)
+### CLI reference
 
-Insert / edit form — column defaults are pre-filled.
-
-![Insert form](docs/screenshots/form.png)
-
-Schema editor — rename the table, add / rename / drop columns safely.
-
-![Schema editor](docs/screenshots/schema.png)
-
-Databases landing page (multi-db mode) — manage multiple SQLite files.
-
-![Databases](docs/screenshots/databases.png)
-
-## Run the built-in server
-
-For convenience a standalone server is included. From a clone of the repo:
-
-```bash
-npm install
-npm run build
-npm start        # open http://localhost:3000
-```
-
-The server is a full standalone web app. With no arguments it runs in
+The standalone server is a full web app. With no arguments it runs in
 **manager mode**: a **Databases** landing page where you can browse the
 filesystem and open any SQLite database file (`.db` / `.sqlite` / `.sqlite3`),
 or create new ones.
 
-### CLI flags
-
 | Flag                 | Description                                              |
 | -------------------- | -------------------------------------------------------- |
 | `-p, --port <port>`  | Port to listen on (default `3000`)                       |
@@ -255,11 +190,14 @@ or create new ones.
 | `-b, --base-path <p>`| URL prefix to serve under (default `/`)                  |
 | `-r, --readonly`     | Open databases read-only (all writes disabled)           |
 | `-l, --log-level <l>`| `debug` \| `info` \| `warn` \| `error` (default `info`)  |
-| `-h, --help`         | Show help                                               |
+| `-h, --help`         | Show help                                                |
 | `-v, --version`      | Show the version                                         |
 
 A positional `path` argument opens a database file directly, or manages a
-folder when it is a directory. Flags override the environment variables below:
+folder when it is a directory.
+
+Every flag has a matching environment variable; **flags override the
+environment**:
 
 | Variable    | Default | Description                          |
 | ----------- | ------- | ------------------------------------ |
@@ -293,7 +231,8 @@ page lists the managed databases and includes an **"Open an existing
 database"** file browser: navigate folders, pick a database file, and open it.
 Opened files are added to the list so you can switch between databases freely.
 
-File-browser policy:
+**File-browser policy:**
+
 - When a **folder** is given (`--dir`, a directory path, or `DB_DIR`), browsing
   is limited to that folder — it cannot navigate above it and only databases
   inside it can be opened.
@@ -302,135 +241,161 @@ File-browser policy:
   listed.
 - With no folder or files, browsing is unrestricted.
 
-## The JSON REST API (under `basePath`)
-
-| Method | Path                                     | Purpose                          |
-| ------ | ---------------------------------------- | -------------------------------- |
-| GET    | `/api/tables`                            | List tables                      |
-| GET    | `/api/tables/:table/info`                | Column + FK metadata, PK columns |
-| GET    | `/api/tables/:table/fk-options`          | Values for FK dropdowns          |
-| GET    | `/api/tables/:table/rows?page&limit&f`   | Paginated rows (`f` = URL-encoded JSON filters — legacy strings like `{"age":">35"}` or structured conditions like `{"balance":{"op":"gte","value":"100"}}`) |
-| GET    | `/api/tables/:table/row/:id`             | Single row by (encoded) PK       |
-| POST   | `/api/tables/:table/rows`                | Insert row                       |
-| POST   | `/api/tables/:table/rows/generate`       | Generate INSERT SQL (no execute) |
-| POST   | `/api/tables/:table/rows/import`         | Import CSV (`{ csv }`, header row must match columns) |
-| GET    | `/api/tables/:table/export?format=`      | Download all rows as `csv` or `json` |
-| PUT    | `/api/tables/:table/row/:id`             | Update row (`{ values, nulls? }` — `nulls` explicitly sets columns to NULL) |
-| PUT    | `/api/tables/:table/row/:id/generate`    | Generate UPDATE SQL (no execute) |
-| DELETE | `/api/tables/:table/row/:id`             | Delete row                       |
-| POST   | `/api/tables/:table/rows/bulk-impact`    | FK-impact preview: how many rows in other tables reference the selected rows |
-| POST   | `/api/tables/:table/rows/bulk-delete`    | Delete selected rows (`{ ids, confirmImpact }`; transactional) |
-| POST   | `/api/tables/:table/rows/bulk-export`    | Export only the selected rows as `csv`/`json` (`{ ids, format }`) |
-| POST   | `/api/tables/:table/rows/bulk-update`    | Apply staged inline edits (`{ updates: [{ id, values?, nulls? }] }`; transactional) |
-| POST   | `/api/tables`                            | Create table                     |
-| POST   | `/api/tables/generate`                   | Generate CREATE SQL (no execute) |
-| GET    | `/api/tables/:table/schema`              | Full schema (constraints, indexes, FK refs) |
-| POST   | `/api/tables/:table/rename`              | Rename the table                 |
-| POST   | `/api/tables/:table/columns`             | Add a column                     |
-| PUT    | `/api/tables/:table/columns/:column`     | Rename a column                  |
-| DELETE | `/api/tables/:table/columns/:column`     | Drop a column (safety-checked)   |
-| DELETE | `/api/tables/:table`                     | Drop the table (safety-checked)  |
-| POST   | `/api/tables/:table/indexes`             | Create an index (`{ name?, columns[], unique? }`) |
-| DELETE | `/api/tables/:table/indexes/:index`      | Drop an index (auto indexes refused) |
-| POST   | `/api/query`                             | Run arbitrary SQL                |
-| POST   | `/api/query/export`                      | Run a SELECT and download as `csv`/`json` |
-| GET    | `/api/queries`                           | List saved queries               |
-| POST   | `/api/queries`                           | Save a named query               |
-| DELETE | `/api/queries/:id`                       | Delete a saved query             |
-
-In multi-db mode, every route is scoped under the database, e.g.
-`/api/app.db/tables`. Every API response uses the consistent shape
-`{ success, data?, error? }`.
-
-In manager mode (the databases landing page), these extra endpoints manage
-database files and power the filesystem browser:
-
-| Method | Path                    | Purpose                                     |
-| ------ | ----------------------- | ------------------------------------------- |
-| GET    | `/api/databases`        | List managed databases                      |
-| POST   | `/api/databases`        | Create a new database (`{ name }`)          |
-| DELETE | `/api/databases/:id`    | Delete a database                           |
-| GET    | `/api/fs/list?path=`    | List subfolders + SQLite files under a path (file browser) |
-| POST   | `/api/databases/open`   | Open/register an existing database file by path (`{ path }`) |
-
-## Behaviour notes
+## Features
+
+### 🔍 Browse & edit rows
+
+- **Table browser** — every table with pagination, sorting (click a column
+  header), a sticky header, and per-row actions; composite primary keys are
+  supported (values are URL-encoded and comma-joined in the row endpoints).
+- **Type-aware filters** — per-column filter controls: foreign-key dropdowns,
+  boolean toggles, date and numeric range inputs, plus the exact (`=value`),
+  comparison (`>5`, `<=10`), prefix (`pre*`) and substring (plain text)
+  operators on text columns. Filters survive sorting and pagination.
+- **Inline (spreadsheet-style) editing** — double-click any cell to edit it in
+  place with a type-aware control (FK dropdown, boolean toggle, date picker,
+  number/text input). Changes are staged and highlighted in the grid, then
+  applied all at once in a single transaction, or discarded.
+- **Insert and delete rows** (full CRUD). FK columns become dropdowns; insert
+  forms pre-fill column defaults from the schema.
+- **Bulk row operations** — select rows with checkboxes (or "select all"), then
+  **delete** them in one transaction (with a warning listing how many rows in
+  other tables reference them) or **export** only the selected rows as CSV/JSON.
+- **Related rows** — every table that has a foreign key pointing at a table gets
+  its own column at the end of that table's grid; each cell shows how many of its
+  rows reference that record. Click it to open a nested table with all of the
+  referencing table's columns and data for that specific record.
+
+### ⚡ Query & run SQL
+
+- **Arbitrary SQL runner** — SELECTs render as a table, `COUNT` queries show a
+  readable summary, and write statements execute and report affected rows.
+- **Saved named queries** — save a query and reload it from a dropdown.
+- **"Get query" / preview mode** — generate `CREATE` / `INSERT` / `UPDATE` SQL
+  from the UI without executing it. The preview modes only return the SQL
+  string; they never touch the database.
+- **SQL dump** — export the whole database as a downloadable `CREATE` + `INSERT`
+  SQL file.
+
+### 🗂️ Design & manage schema
+
+- **Visual table designer** — create a table with name, type, primary key,
+  not-null / unique, default value, and foreign-key references, with a live
+  `CREATE TABLE` SQL preview.
+- **Schema editor** — rename the table, add / rename / drop columns, and drop
+  tables, with relationship-safety checks: drops are refused when the column is
+  a primary key, has a UNIQUE constraint, is used by an index, is part of a
+  foreign key, or is referenced by another table's foreign key. Tables
+  referenced by other tables cannot be dropped. Internal (`_`-prefixed) tables
+  cannot be renamed or dropped.
+- **Indexes** — create indexes (plain or unique, on one or many columns — pick
+  columns in order, with a live `CREATE INDEX` SQL preview) and drop them from
+  the schema editor; automatic SQLite (primary-key/unique) indexes are
+  protected.
+
+### 🔄 Import, export & seed data
+
+- **Export** — table rows or query results as **CSV or JSON** (a whole table or
+  only selected rows).
+- **CSV import** — paste or upload CSV into a table; the header must match
+  existing columns, and the import runs in a single transaction (a failed row
+  rolls everything back).
+- **Seed data generator** — fill a table with realistic rows in one go. Each
+  column gets an auto-detected strategy (first/last/full name, email, phone,
+  city, country, UUID, random integer/decimal/date/datetime/boolean/bytes, a few
+  words, a sentence, a fixed value, a random value from a list, or "skip — let
+  the DB default apply"); foreign-key columns can sample real values from the
+  referenced table. Insert up to 5,000 rows transactionally, or preview the
+  generated `INSERT` SQL without executing it.
+
+### 🚀 Deploy
+
+- **Read-only mode** — open the database(s) without write access: the file is
+  opened `SQLITE_OPEN_READONLY` + `query_only`, every write route returns `403`,
+  and the UI hides/disables all write controls and shows a banner.
+- **Standalone CLI / file browser** — `admindb` runs as a full web app; with no
+  arguments it opens a **Databases** landing page where you can browse the
+  filesystem and open any SQLite database file, or create new ones. Flags set
+  the port, open a file, manage a folder, and more (`admindb --help`).
+- **Multiple databases** — directory scanning and/or explicit file lists, each
+  with its own workspace under `/{db}/…`.
+
+## Security warning
+
+> ⚠️ **This tool exposes full, unauthenticated database AND filesystem access.**
+> Every page and API route (browse, edit, delete, run arbitrary SQL, change the
+> schema, export the whole database, and — in manager mode — browse the
+> filesystem to open database files) is available to **anyone who can reach the
+> server**.
+>
+> - **No authentication or authorization is built in.** The routes are **not
+>   protected**.
+> - **It is your responsibility to protect access.** Do **not** expose
+>   AdminDB to the public internet or to untrusted networks.
+> - Recommended ways to protect it:
+>   - bind the standalone server to `127.0.0.1` (`HOST=127.0.0.1`) and use it
+>     only from your own machine, and/or
+>   - run it behind a reverse proxy that requires authentication (Basic auth,
+>     OAuth, mTLS, …) or inside a VPN / private network.
+>
+> Treat AdminDB as if it were a remote `sqlite3` shell with write access.
+
+## API
+
+All routes live under `basePath` and return the consistent shape
+`{ success, data?, error? }`. In multi-db mode, every route is scoped under the
+database, e.g. `/api/app.db/tables`.
+
+```text
+GET    /api/tables                          List tables
+GET    /api/tables/:table/rows              Paginated rows (filters, sorting)
+POST   /api/tables/:table/rows              Insert row
+PUT    /api/tables/:table/row/:id           Update row
+DELETE /api/tables/:table/row/:id           Delete row
+POST   /api/query                           Run arbitrary SQL
+POST   /api/tables                          Create table
+GET    /api/tables/:table/export            Download all rows as csv|json
+POST   /api/tables/:table/seed              Generate and insert seed rows
+```
+
+Plus bulk row operations, CSV import, schema and index management, saved
+queries, a data generator — and, in manager mode, database-file management and
+the filesystem browser.
+
+> **Full reference:** every endpoint (including seed config/generate, bulk
+> update/delete/export, and the multi-db manager endpoints) is documented in
+> **[`docs/API.md`](docs/API.md)**.
+
+**Behaviour notes:**
 
 - **Empty input = "not set".** Empty form fields are omitted so DB defaults
-  apply; `0` is a valid value and is never treated as empty.
-- **Insert forms pre-fill defaults.** On the "new row" form, columns that have
-  a schema default are pre-filled (string/number/boolean literals and
-  `CURRENT_TIMESTAMP`-style defaults) so you can see and adjust them.
-- **Row filters.** The filter panel adapts to each column's type: foreign keys
-  become dropdowns (with a "not set / NULL" option), booleans become
-  any/true/false toggles, and dates and numbers become min/max range inputs.
-  Text columns keep the operator syntax: exact match (`=value`), comparison
-  (`>5`, `>=5`, `<5`, `<=5`, `!=value`), prefix (`pre*`) and case-insensitive
-  substring (plain text). Filters are carried in the URL (as legacy strings or
-  structured conditions) and survive sorting and pagination.
+  apply; `0` is a valid value and is never treated as empty. Insert forms
+  pre-fill column defaults from the schema (string/number/boolean literals and
+  `CURRENT_TIMESTAMP`-style defaults).
 - **Inline editing is staged, not instant.** Double-click a cell to edit it in
   place; edits are buffered locally and highlighted in the grid rather than
-  written immediately. Press **Apply** to write every pending change to the
-  database in a single transaction (the page then reloads so related-row counts
-  stay accurate), or **Discard** to revert everything back to the saved values.
-  Clearing a text/date/number field stages a `NULL`.
-- **Bulk operations.** Checkboxes select rows on the current page; "select all"
-  checks every visible row. **Delete** first shows a warning listing each table
-  that references the selected rows and how many rows point at them (these may
-  be cascaded away or orphaned depending on the foreign-key action, and the
-  delete can fail if a constraint blocks it). **Export CSV / JSON** downloads
-  only the selected rows. Both delete and export are capped at 1000 rows per
-  batch.
-- **CSV import.** The first row must be a header whose names match existing
-  columns (unknown or duplicate names are rejected). Empty cells are treated as
-  "not set" so database defaults apply. The whole import runs in a single
-  transaction — a failed row rolls everything back.
-- **Indexes.** Creating an index takes one or more columns and an optional
-  unique flag (the index name is optional too). Only explicitly created indexes
-  (`origin = 'c'`) can be dropped from the UI — automatic primary-key/unique
-  indexes are protected.
-- **Schema editor.** Drops are refused when the column is a primary key, has a
-  UNIQUE constraint, is used by an index, is part of a foreign key, or is
-  referenced by another table's foreign key. Tables referenced by other tables
-  cannot be dropped. Internal (`_`-prefixed) tables cannot be renamed or dropped.
+  written immediately. Press **Apply** to write every pending change in a single
+  transaction (the page then reloads so related-row counts stay accurate), or
+  **Discard** to revert. Clearing a text/date/number field stages a `NULL`.
+- **Bulk operations** run per page with "select all"; delete first shows an
+  FK-impact warning (rows may be cascaded away or orphaned depending on the
+  foreign-key action, and the delete can fail if a constraint blocks it). Both
+  delete and export are capped at 1000 rows per batch.
+- **CSV import** runs in a single transaction — a failed row rolls everything
+  back.
 - **Identifiers are quoted and string values escaped** everywhere SQL is built,
-  so generated SQL is correct and safe.
+  so generated SQL is correct and safe; **"Get query" never executes** — it only
+  returns the generated SQL string.
 - **`COUNT` queries** return a readable summary message instead of a table.
-- **"Get query" never executes** — it only returns the generated SQL string.
 - **Internal table** `_saved_queries` stores saved queries and is kept out of
   user-facing FK pickers. Schema initialization is idempotent and safe to run
   repeatedly.
-- Composite primary keys are supported (values are URL-encoded and comma-joined
-  in the row endpoints).
-
-## Development
-
-```bash
-npm install
-npm run build      # TypeScript → dist + Tailwind CSS
-npm run dev        # tsx watch server + Tailwind watch
-npm test           # builds and runs the unit tests
-```
-
-`npm test` compiles TypeScript to `dist/` and runs the `node:test` suite
-covering the SQL generator (INSERT / UPDATE / CREATE output, type mapping,
-quoting, rejection of unsupported types), the SQL classifier, CSV parsing and
-serialization, row filters (legacy and structured conditions), the database
-layer (pagination, transactions, bulk update/delete, read-only enforcement),
-and the database manager.
 
-## Publishing to npm
-
-The package ships the compiled `dist/` (with type declarations), this `README`,
-and the `LICENSE`. When you are ready to publish:
-
-```bash
-npm run build     # happens automatically via the prepack script
-npm login
-npm publish
-```
+## Contributing
 
-Update `name`/`version` in `package.json` to match your intended package name
-and add an `author`/`repository` if desired.
+PRs are welcome. Development workflow (`npm run dev` for live reload, `npm test`
+to build and run the unit suite), project layout, and contribution guidelines
+live in **[`CONTRIBUTING.md`](CONTRIBUTING.md)**.
 
 ## License