Keboola Docs

LLM Export Command

Export project data to AI-optimized twin format directory structure.

kbc llm export [flags]

The command must be run in a directory initialized with kbc llm init.

Description

The twin format is designed for AI assistants to understand and work with Keboola projects directly from Git repositories. The export includes:

Bucket and table metadata with schema information
Transformation configurations with platform detection
Component configurations organized by type
Job execution history and statistics
Lineage graph showing data flow dependencies
Optional data samples (controlled by flags)

The export creates output files containing JSON with inline documentation (_comment, _purpose, _update_frequency fields) to help AI assistants understand the data structure.

Security Features

Public repository detection - Automatically detects if the directory is a public Git repository
Sample export disabled by default - Data samples must be explicitly enabled with --with-samples
Encrypted secrets - Fields starting with # are encrypted in the output

Options

-H, --storage-api-host <string> : Keboola instance URL, e.g., “connection.keboola.com”

-t, --storage-api-token <string> : Storage API token from your project

-f, --force : Skip confirmation when directory contains existing files

--with-samples : Include table data samples in the export

--sample-limit <int> : Maximum number of rows per table sample (default: 100, max: 1000)

--max-samples <int> : Maximum number of tables to sample (default: 50, max: 100)

Global Options

Output Structure

The export creates the following directory structure:

.
├── buckets/              # Bucket and table metadata
│   └── index.json
├── transformations/      # Transformation configurations
├── components/           # Component configurations by type
├── jobs/                 # Job execution history
│   ├── recent/
│   └── by-component/
├── indices/              # Query indices and lookups
│   └── queries/
├── ai/                   # AI assistant guides
├── samples/              # Table data samples (if --with-samples)
├── lineage.json          # Data flow dependencies
└── metadata.json         # Project metadata

Examples

Basic Export

➜ kbc llm export

[1/5] Getting default branch...
Using branch: Main (ID: 1234)
[2/5] Fetching project data from APIs...
Fetched: 5 buckets, 23 tables, 150 jobs
[3/5] Processing data (lineage, platforms, sources)...
Processed: 5 buckets, 23 tables, 8 transformations, 45 lineage edges
[4/5] Generating twin format output...
[5/5] Skipping samples (not requested)
Twin format exported to: /path/to/project
Export completed successfully.

Export with Data Samples

➜ kbc llm export --with-samples --sample-limit 50 --max-samples 20

[1/5] Getting default branch...
Using branch: Main (ID: 1234)
[2/5] Fetching project data from APIs...
Fetched: 5 buckets, 23 tables, 150 jobs
[3/5] Processing data (lineage, platforms, sources)...
Processed: 5 buckets, 23 tables, 8 transformations, 45 lineage edges
[4/5] Generating twin format output...
[5/5] Fetching and generating table samples...
Twin format exported to: /path/to/project
Export completed successfully.