qdbd: QuickDBD diagrams from the command line

qdbd turns QuickDBD-syntax .txt schemas, or a dbt project, into interactive ER diagrams with no table limit. The diagrams stay readable at 200+ tables.

▶ Live demo (200 tables) · ▶ dbt demo (jaffle_shop) · 📖 Docs · PyPI

Not affiliated with QuickDBD. It reads the same text syntax, so schemas can be copied back and forth between the two.

---

Contents

1. Install
2. Quick start: your first diagram in 5 minutes
3. Working with schema files
4. Exploring a diagram
5. dbt projects
6. Reviewing schema changes
7. CI and automation
8. Command reference
9. Development

---

1. Install

pipx install quickdbd-cli      # or: pip install quickdbd-cli / uv tool install quickdbd-cli
qdbd --version

• Requirements: Python 3.11+. Nothing else: no Graphviz, Node or browser. Layout uses ELK running in an embedded JavaScript engine.
• Try it without installing: prefix any command with uvx --from quickdbd-cli, e.g. uvx --from quickdbd-cli qdbd build schema.txt -o schema.html.
• qdbd: command not found? pip install put the script in a directory that isn't on your PATH. Use pipx, or run python -m pip show -f quickdbd-cli to find it.

2. Quick start: your first diagram in 5 minutes

2.1 Write (or paste) a schema

Save this as schema.txt. If you already have a schema in QuickDBD, copy it from the editor and paste it in unchanged.

Customer
-
CustomerId PK int
Name string
Email string UNIQUE

Order as o
----
OrderId PK int
CustomerId int FK >- Customer.CustomerId
OrderDate datetime
Total decimal(10,2)

A table is a name, a line of dashes, then one column per line. FK >- Customer.CustomerId means many orders → one customer. See the syntax cheat sheet.

2.2 Check it

qdbd check schema.txt

No errors found. Good job!

Errors print as schema.txt:12: error: … (QuickDBD's own messages) and exit with code 1.

2.3 Edit it live

qdbd serve schema.txt

A browser tab opens with the diagram and an editor pane.

• Edit in the pane (it autosaves after 1 s) or in your own editor. The diagram reloads on save.
• Drag tables to arrange them. Positions are saved to schema.layout.json; commit it next to the schema.
• Stop with Ctrl+C.

2.4 Build a page to share

qdbd build schema.txt -o schema.html

schema.html is a single self-contained file. Open it locally, email it, attach it to a ticket, or publish it on GitHub Pages (7.1).

2.5 Export an image

qdbd export schema.txt -o schema.png                       # whole diagram (.png, .svg or .drawio)
qdbd export schema.txt --focus Order --depth 1 -o order.svg   # one table and its neighbours

.drawio files open in diagrams.net as editable ER shapes.

3. Working with schema files

3.1 Syntax cheat sheet

# A comment above a table or column becomes its tooltip
"Sales Order" as so          # quote names with spaces or dots; "as" adds an alias
--------------
Id PK int IDENTITY
CustomerId int FK >- Customer.Id
Status varchar(20) default=new INDEX
Notes text NULL              # columns are NOT NULL unless marked NULL

Column keyword	Meaning
PK	primary key (several PK columns make a composite key)
FK >- Table.Column	foreign key with a relationship
NULL / NULLABLE	nullable
UNIQUE, INDEX	unique key, index
IDENTITY / AUTOINCREMENT	auto-generated
default=value	default value

Relationship	Meaning (left → right)
-	one to one
-< / >-	one to many / many to one
>-<	many to many
-0 / 0- / 0-0	involving zero-or-one
-0< / >0-	involving zero-or-many

Full reference: Schema Syntax.

3.2 One folder, many files

Big schemas are easier to manage split into files. Point any command at a folder and every *.txt under it (recursively) becomes one diagram:

schema/
├── sales.txt        # Order, OrderLine, Invoice …
├── customers.txt    # Customer, Address …
└── products.txt

qdbd build schema/ -o schema.html

• Foreign keys can point at tables in other files.
• Each file becomes a group in the sidebar and can be shaded as a region.
• --by-group arranges the tables so each file's tables sit together.

3.3 Table positions

Situation	What happens
First build	Tables are auto-arranged by their relationships
You drag tables in qdbd serve	Positions are saved to the layout file
You add a new table later	It's placed next to the tables it's related to. Nothing else moves
Schema pasted from QuickDBD with a [layout] section	Those positions are used for tables the layout file doesn't cover
You want a fresh arrangement	qdbd layout PATH --reset

Where the layout file goes:

• schema.txt → schema.layout.json
• a folder → <folder>/qdbd.layout.json
• anywhere else: --layout FILE

qdbd build never writes the layout file, so CI builds leave your working tree clean. More: Table Positions.

3.4 Differences from QuickDBD

qdbd accepts a few things QuickDBD rejects, and prints a warning for each:

• tab characters
• a space inside a type: decimal(10, 2)
• an explicit NOT NULL
• indented table names
• [layout] entries for tables that don't exist

qdbd check --strict turns these warnings into errors, so a file that passes will paste into QuickDBD cleanly. Also, default= values keep their case, where QuickDBD lowercases them.

4. Exploring a diagram

Every page from build or serve works the same way. A full tour is in Reading the Diagram.

4.1 Page tour

Part	What you can do
Explorer (left)	Tables grouped by file. Press / to search; Enter jumps to the best match
Canvas	Drag tables, drag empty space to pan, scroll to zoom. Hover a column to highlight its relationships
Details panel (right)	Click a table: columns, doc comments, References / Referenced by, path finder
Toolbar (top right)	Collapse tables (Full / Keys / Names), Regions, Link, SVG / PNG / draw.io download, zoom
Minimap (bottom right)	Click or drag to move around

Link copies a URL to exactly what you're looking at: the filter, selected table, path, collapse mode and zoom. Send it to a teammate.

4.2 Filter with selectors

The explorer's filter box takes dbt-style selectors that follow relationships (not lineage). The same selectors work as -s/--select and --exclude on build and export.

Selector	Tables
orders, stg_*	by name or alias (glob, case-insensitive)
+orders / orders+	orders and what it references / what references it
+orders+1	one hop in both directions
group:marts, file:core.txt	a folder or a file
col:customer_id	tables with a matching column
a b / a,b / !a	union / intersection / exclude

qdbd build erd/ -s 'group:marts' --exclude '*_tmp' -o marts.html
qdbd export erd/ -s '+orders+1' -o orders.png

More: Selecting Tables.

4.3 Find how two tables join

In a table's details panel, type another table's name. The shortest join path is highlighted on the canvas, with a ready-to-paste SQL JOIN. From the terminal:

qdbd path schema/ --from Invoice --to Campaign

5. dbt projects

5.1 Generate the schema from dbt

cd your-dbt-project
dbt docs generate                                # writes target/manifest.json + catalog.json
qdbd import dbt target/ -o erd/ --infer-fks      # one .txt per dbt folder
qdbd build erd/ --by-group -o erd.html           # open erd.html

dbt docs generate is what supplies column names and types for models without documented columns. Without it, many tables show up with no columns and qdbd prints a hint.

See it on a real project: the jaffle_shop demo and the step-by-step walkthrough.

5.2 What gets imported

In the diagram	Comes from
Tables	your project's models, seeds and snapshots, plus sources
Primary keys	primary_key constraints, or a column with both unique and not_null tests
Relationships	relationships tests (incl. dbt_utils relationships_where) and foreign_key constraints
Column types	catalog.json
Tooltips	model and column descriptions

Useful options: --select 'marts.*' (repeatable), --no-sources, --package all, -o schema.txt for a single file.

5.3 Infer missing relationships

Most dbt projects don't declare every foreign key. --infer-fks links columns like customer_id to the table they name (dim_customers, stg_shop__customers, …) when no relationship is declared. Ambiguous matches are skipped, never guessed.

5.4 Keep it in sync

Re-run the import whenever models change. It overwrites the generated .txt files. Your dragged positions survive, because the layout file is keyed by table name. In CI, --check fails if the committed files are out of date (7.2).

More: dbt Import.

6. Reviewing schema changes

6.1 See what changed

qdbd diff old.txt new.txt                       # two versions
qdbd diff --base origin/main erd/               # a git branch vs your working tree
qdbd diff --base origin/main erd/ --format markdown -o diff.md   # for a PR comment

It reports tables, columns (added, removed, type and flag changes) and relationships added or removed. --format json is also available, and --exit-code exits 1 when anything changed.

6.2 Highlight changes in the diagram

qdbd build erd/ --diff-base origin/main -o erd.html

New tables are green and changed tables amber, with a bar on each changed column. A banner summarises the changes, and Show changes filters to them plus their neighbours. The selector diff:modified+1 does the same from the CLI.

More: Schema Diffs.

7. CI and automation

7.1 GitHub Action

Checks the schema and builds the diagram. On pull requests it also posts one schema-diff comment (updated on every push) and uploads the diagram with the changes highlighted.

# .github/workflows/erd.yml
name: ERD
on: [pull_request, push]
jobs:
  erd:
    runs-on: ubuntu-latest
    permissions: { contents: read, pull-requests: write }
    steps:
      - uses: actions/checkout@v4
      - uses: MarkPhamm/quickdbd-cli@v0.1.7
        with:
          path: erd/                  # schema file(s) or folder(s)
          output: site/index.html

Input	Default
path	erd/	schema files or folders
output	erd/index.html	where to write the HTML
check / strict	true / false	fail on errors / also on QuickDBD-compatibility warnings
comment	true	post the diff comment on PRs
select		selector for the diagram, e.g. group:marts
qdbd-version	latest	PyPI version

Outputs: html, changed, diff-file. To publish the diagram on GitHub Pages, add actions/upload-pages-artifact + actions/deploy-pages; this repo's erd.yml does that for the live demo. All inputs: GitHub Action.

7.2 dbt projects in CI

Run your dbt setup first, then let the action generate the schema:

      # … dbt deps && dbt build && dbt docs generate
      - uses: MarkPhamm/quickdbd-cli@v0.1.7
        with:
          dbt-manifest: target/
          dbt-args: --infer-fks
          path: erd/
          output: site/index.html

Commit erd/ so pull requests can be compared with main. Set dbt-check: true to fail the job when the committed files are out of date instead of regenerating them. Walkthrough: Automate in CI.

7.3 pre-commit

repos:
  - repo: https://github.com/MarkPhamm/quickdbd-cli
    rev: v0.1.7
    hooks:
      - id: qdbd-check
        files: ^erd/.*\.txt$

7.4 GitLab CI (or any CI)

erd:
  image: python:3.12
  script:
    - pip install quickdbd-cli
    - qdbd check erd/
    - qdbd build erd/ -o public/index.html
  artifacts:
    paths: [public/index.html]

Builds are deterministic (no timestamps or randomness), so the output only changes when the schema does.

8. Command reference

qdbd check  PATH... [--strict]                      # validate
qdbd serve  PATH... [--port 8765]                   # live editor + drag to arrange
qdbd build  PATH... -o diagram.html                 # interactive HTML
qdbd export PATH... -o out.png|.svg|.drawio [--focus T --depth 1]
qdbd layout PATH... [--reset]                       # (re)write the layout file
qdbd path   PATH... --from A --to B                 # shortest join path + SQL JOIN
qdbd diff   OLD NEW | --base REF PATH...            # schema changes
qdbd import dbt target/ -o erd/ [--infer-fks]       # schema files from dbt

PATH is a .txt file or a folder (read recursively). Run qdbd COMMAND --help for every option, or see The qdbd CLI.

9. Development

uv venv && uv pip install -e ".[dev,docs]"
.venv/bin/pytest                                      # full suite
.venv/bin/ruff check src tests scripts && .venv/bin/ruff format src tests scripts
.venv/bin/mkdocs serve                                # docs site at localhost:8000
.venv/bin/qdbd serve examples/synthetic               # 200-table playground

More: Local Development. The design notes are in docs/superpowers/specs/.

9.1 Releasing

1. Put PyPI credentials in .env (gitignored): TWINE_USERNAME=__token__ and TWINE_PASSWORD=pypi-….
2. Bump version in pyproject.toml and src/quickdbd_cli/__init__.py, then merge.
3. Run PATH="$PWD/.venv/bin:$PATH" scripts/release.sh. It builds, uploads to PyPI, tags vX.Y.Z and creates the GitHub release.

License

Apache-2.0. Bundled third-party files (ELK, Inter) are listed in NOTICE.