dsh-db-visualizer

DeepSeek Harness plugin: offline SQL DDL schema visualization and analysis.

Paste CREATE TABLE text (MySQL, PostgreSQL, or SQLite dialect) into the tools; they parse it in-process and return structure, a Mermaid ER diagram, and schema-hygiene findings. No database connection, no subprocesses, no network — every tool is a pure DDL text analysis.

Install

npx -y @deepseek-ai/dsh plugin --profile web add @qingshanjiluo/dsh-db-visualizer

Tools

Tool	Input	Output
db_parse_schema	ddlText — raw DDL script	Tables with columns (type, nullability, defaults, comments, ENUM/CHECK values), primary keys, indexes, and foreign keys (target, ON DELETE/UPDATE)
db_er_mermaid	ddlText	Mermaid erDiagram source with typed attributes (PK/FK/UK) and FK-derived relationship lines, plus table/relation counts
db_analyze	ddlText	Findings: missing-index on foreign keys (with a suggested CREATE INDEX), missing-primary-key, nullable-fk-column, nullable-unique-column, nullable-without-default, dangling-fk

Configuration

Field	Type	Default	Meaning
includeComments	boolean	true	Expose COMMENT text declared in the DDL
indexNamePrefix	string	idx	Prefix for index names suggested by db_analyze

Parsing notes

The parser is string-aware and line/regex based: it understands quoted identifiers (`x`, "x", [x]), schema-qualified names, inline and table-level PRIMARY KEY / UNIQUE / FOREIGN KEY ... REFERENCES / KEY|INDEX items, ENUM('...') and CHECK (col IN (...)) value sets, and COMMENT= / COMMENT ON TABLE|COLUMN comments. CREATE TABLE ... AS SELECT is skipped; structural problems are reported in the errors array instead of throwing.

Development

npm install
npm run typecheck   # tsc --noEmit
npm run build       # tsc + tsdown → lib/
npx vitest run      # behavior tests
node scripts/load-smoke.mjs   # loads lib/index.js, asserts the tool face

License

MIT