Docs / Configuration

Configuration

Conventions files

Lakecode reads conventions files to understand your team’s standards and preferences. There are two levels:

  • Global — ~/.lakecode/conventions.md applies to all projects on your machine.
  • Project — .lakecode/conventions.md in your project root applies only to that project. Project conventions override global ones.

Conventions are written in plain Markdown. Lakecode reads them before every command to tailor its behavior. Example:

# Team Conventions

## Default catalog and schema
Always use `analytics.prod` as the default catalog.schema unless specified.

## Naming patterns
- Tables: snake_case, prefixed with source (e.g., stripe_charges)
- Jobs: kebab-case with team prefix (e.g., data-eng-daily-refresh)

## Preferred warehouse
Use the `analytics-medium` SQL warehouse for ad-hoc queries.

## Standards
- All tables must have a `_loaded_at` timestamp column
- Partition by date for tables over 1B rows
- Tag all jobs with team name and SLA tier

Skills injection

Lakecode ships with a library of platform-specific skills (Unity Catalog, Delta Lake, Snowflake Streams, Tasks, Cortex, and more). When you run a command, Lakecode detects relevant keywords and injects the appropriate skills into the AI context. This means the AI always has deep, current knowledge about the platform features involved in your query — without you configuring anything.

Config files

Lakecode uses YAML config files at two levels:

  • Global — ~/.lakecode/config.yml applies to all projects.
  • Project — ./lakecode.yml in your project root. Project config overrides global.

You can also set LAKECODE_CONFIG_PATH to point to an external config file.

# lakecode.yml
platform: databricks          # or "snowflake"

databricks:
  profile: DEFAULT             # Databricks CLI profile name
  default_catalog: analytics
  default_schema: prod

snowflake:
  connection: default          # connection name from ~/.snowflake/config.toml
  warehouse: ANALYTICS_WH
  database: PROD
  role: DATA_ENG

Environment variables

Lakecode uses standard environment variables for workspace connectivity. Environment variables take precedence over config files.

Databricks

  • DATABRICKS_HOST — Your workspace URL (e.g., https://dbc-abc123.cloud.databricks.com)
  • DATABRICKS_TOKEN — A personal access token for your workspace
  • LAKECODE_PROFILE — Databricks CLI profile name (default: DEFAULT)
  • LAKECODE_TARGET — Target workspace for bundle operations

Snowflake

  • LAKECODE_SNOWFLAKE_CONNECTION — Connection name from ~/.snowflake/config.toml (default: default)
  • LAKECODE_SNOWFLAKE_ACCOUNT — Your Snowflake account identifier

General

  • LAKECODE_PLATFORM — Platform: databricks or snowflake
  • ANTHROPIC_API_KEY — (Optional) Your own Anthropic key for BYOK mode

If you have a Databricks CLI profile (~/.databrickscfg) or Snowflake connection (~/.snowflake/config.toml), Lakecode will use it automatically.

← Previous Commands