looker-fields 0.1.1

Extract field-level metadata from any Looker instance via the API

looker-fields

Extract field-level metadata from any Looker instance via the API.

Produces an accurate, de-duplicated inventory of every field (dimensions, measures, dimension_groups, filters, parameters) across all models, explores, and views — with correct model attribution.

Why

Parsing raw .lkml files cannot resolve include statements, leading to model misattribution and Cartesian duplication. This tool extracts the compiled truth directly from the Looker API — the final state after Looker's engine has resolved all includes, extensions, and refinements.

Install

pip install -e ".[dev]"

Setup

Create a .env file:

LOOKER_BASE_URL=https://your-instance.cloud.looker.com
LOOKER_CLIENT_ID=your_client_id
LOOKER_CLIENT_SECRET=your_client_secret

Usage

# Extract all fields (JSONL output)
looker-fields extract

# Extract as CSV
looker-fields extract --format csv --output fields.csv

# Extract specific model
looker-fields extract --model my_model

# Verify extraction against live API
looker-fields verify my_model my_explore

# Show instance info
looker-fields info

Output

Each row represents one field with grain (project, model, explore, field_name).

Supported formats: JSONL (default), CSV, Parquet, BigQuery.

See docs/FIELD_SPEC.md for the full output schema.

Architecture

src/looker_fields/
├── cli.py       # typer CLI
├── config.py    # .env + Settings
├── client.py    # Async httpx client
├── schema.py    # Output models + Swagger discovery
├── extract.py   # Core extraction logic
└── output.py    # Multi-sink writers

See gsd-lite/ARCHITECTURE.md for design decisions.

License

Apache 2.0