Setup: dbt Fusion Path

Use this when the project runs dbt Fusion.

Key insight

Adding --write-index (or setting DBT_USE_INDEX=1) to any Fusion command automatically generates the Parquet index as part of that command — there is no separate ingest step. The index is written after compilation, with ~2ms overhead.

Creating the index

Just add --write-index to your normal Fusion commands:

dbtf parse --write-index
dbtf build --write-index
dbtf run --write-index
dbtf test --write-index

Or set DBT_USE_INDEX=1 once so every Fusion command keeps the index fresh automatically:

export DBT_USE_INDEX=1

The index is written to target/index/ by default. Specifying a different location with --index-dir (or DBT_INDEX_DIR) is rarely needed — the default works in almost all cases:

dbtf parse --write-index --index-dir /path/to/index

Environment variables

Verify

dbt-index status

Keeping the index fresh

Since --write-index is additive to normal commands, the index stays fresh automatically as long as the flag (or env var) is set. Every dbtf build, dbtf run, etc. refreshes the index.

Differences from Core path

• No separate dbt-index ingest step needed — index is generated as part of the command
• Faster write path (Arrow → Parquet directly, ~2ms vs ~100ms)
• Column lineage: Fusion's compile-time static analysis populates dbt.column_lineage with richer data
• Everything else is identical — same commands, same schemas, same query surface