Connecting a trace source

GigaFlow ingests traces from an observability platform (or directly via OTLP), maps each span to a GigaFlow primitive with a transform config, then runs Flow analysis. This directory has one guide per source.

For the five platforms below, the whole connection is driven by gigaflow setup: it signs you in with your waitlist email (same as gigaflow login), asks which platform you use, collects the connection details, then creates a project, applies a transform, registers the datasource, and runs the first sync. You never set a GigaFlow API key or backend URL by hand.

Source	What you provide	Source identifier	Bundled transform
Arize Phoenix	Postgres connection	table name (spans)	✅ works out of the box
Logfire	read token	project base URL	✅ works out of the box
Braintrust	API key	project name	✅ bundled (may need tailoring)
MLflow	token (or none)	server URL	✅ bundled (may need tailoring)
W&B Weave	WANDB_API_KEY	entity/project	⚠️ template — tailor to your ops

When the wizard asks for a transform, leave it blank to use the bundled one, or point it at your own transform.yml. Sync is append-only: re-running never deletes traces; already-imported ones are skipped.

Direct OTLP

Direct OTLP is the exception — the no-platform path where you point your app's OpenTelemetry exporter straight at GigaFlow. It isn't part of the gigaflow setup wizard and is a more advanced, API-driven setup; see its guide.