CLI Reference

Run dbsprout --help for the live command list, or dbsprout <command> --help for any command’s options.

Global Options

Flag	Description
--verbose, -v	Show full tracebacks on error
--install-completion	Install shell completion
--show-completion	Print shell completion script
--help	Show help for any command

dbsprout init

Introspect a database schema and generate configuration.

Option	Description
--db	Database URL (env: DBSPROUT_TARGET_DB)
--file	Schema file (.sql, .dbml, .mmd, .puml, .prisma)
--django	Introspect Django models
--django-apps	Comma-separated Django app labels to include
--output-dir, -o	Output directory (default: .)
--dry-run	Preview without writing files

dbsprout generate

Generate seed data from the schema snapshot.

Option	Description	Default
--rows, -n	Rows per table (≥ 1)	100
--seed, -s	Deterministic seed (≥ 0)	42
--output-format, -f	sql, csv, json, jsonl, direct	sql
--output-dir, -o	Output directory	./seeds
--dialect, -d	SQL dialect	postgresql
--engine, -e	heuristic, spec, statistical, finetuned	heuristic
--privacy	Privacy tier (local, redacted, cloud)	local
--reference-data	Reference CSV (file or per-table dir) for --engine statistical	—
--db	Target DB URL for direct insertion (env: DBSPROUT_TARGET_DB)	—
--upsert	Emit insert-or-update SQL	false
--insert-method	Direct insertion: auto, copy, load_data, batch	auto
--file	Schema file for --incremental	—
--incremental	Apply only schema-diff-driven updates to existing seed data	false
--snapshot	Base snapshot hash prefix	latest
--lora	Path to a .gguf LoRA adapter (requires --engine spec)	—
--schema-snapshot	Explicit schema snapshot path	—
--config	Explicit config path	—

dbsprout validate

Validate integrity of generated seed data.

Option	Description	Default
--rows, -n	Rows per table	100
--seed, -s	Seed	42
--engine, -e	Generation engine to validate	heuristic
--format, -f	rich, json	rich
--reference-data	Reference CSV for fidelity comparison	—
--detection	Run C2ST detection metrics	false
--output	Write JSON output to file	—
--compact	Minified JSON output	false

dbsprout diff

Report schema changes since the last snapshot.

Option	Description	Default
--db	Database URL (env: DBSPROUT_TARGET_DB)	—
--file, -f	Schema file (SQL/DBML/Mermaid/PlantUML/Prisma)	—
--snapshot	Base snapshot hash prefix	latest
--format	rich, json	rich
--output-dir, -o	Project root containing .dbsprout/	.

dbsprout report

Generate an HTML report from recorded generation runs — no regeneration.

Option	Description	Default
--output, -o	Report destination path	./seeds/report.html
--run-id	Render a specific historical run id	most recent

dbsprout audit

Show the LLM interaction audit log.

dbsprout audit              # All audit entries
dbsprout audit --last 10    # 10 most recent entries

dbsprout doctor

Diagnose the local environment for common configuration issues (Python version, drivers, models, disk space, secrets, plugins, training accelerator).

Option	Description	Default
--db	Database URL to test (env: DBSPROUT_TARGET_DB)	—
--config	Config file to scan for secrets	dbsprout.toml

dbsprout models

List, download, and inspect embedded GGUF models.

dbsprout models list              # Registry + locally installed models
dbsprout models info <model-id>   # Details for one registry model
dbsprout models download <model-id>   # Download into .dbsprout/models/base/

dbsprout plugins

Inspect discovered DBSprout plugins.

dbsprout plugins list                  # Discovered plugins by entry-point group
dbsprout plugins check <group>:<name>  # Validate a plugin loads and conforms

dbsprout train

QLoRA fine-tuning pipeline. Subcommands run the stages individually; the top-level command runs them end to end.

Option	Description	Default
--db	Live database URL to sample (env: DBSPROUT_TARGET_DB)	—
--sample-rows	Rows to sample (≥ 1)	1000
--epochs	Training epochs	—
--output, -o	Base directory for artifacts	.dbsprout
--seed	Sampling seed	0
--no-pii-redaction	Disable PII redaction before serialization	false
--quiet	Suppress progress output	false

Subcommand	Description
extract	Extract a stratified sample from a live DB into Parquet
serialize	Serialize Parquet samples into GReaT-style JSONL
run	Fine-tune a QLoRA adapter on a serialized corpus