Setup

Step 1 — Create an Airflow source

In Prizm, go to Sources → Add Source → Airflow.

Identity fields

Field	Description
Name	The source’s identifier inside Prizm. For CLI and Plugin modes, the `connection_name` in every push must match this value exactly — character for character, including case. Choose a stable name; renaming requires updating your CLI command or Airflow Variable.
Description	Optional free text describing this connection.
Integration Type	`API Pull`, `CLI`, or `Plugin`. Locked once the source is saved — switching requires creating a new source.

Lineage & profiling fields

Field	Description
Lineage Method	`Query Tag` (default) or `Open Lineage`. Fixed to Query Tag for CLI sources. See the Lineage page for details.
Open Lineage Endpoint URL	Appears when Open Lineage is selected. Read-only, computed by Prizm per source — copy this URL into your Airflow OpenLineage provider configuration.
Event Automatic Profiling	Whether ingest events automatically trigger profiling of newly discovered assets.

Alert fields

Field	Description
DAG Alert Failure	Raise a Prizm alert when any DAG run ends in a failure state.
Task Alert Failure	Raise a Prizm alert when any task instance fails within a run.

These alert on run state directly — they are independent of the metric thresholds in What We Collect.

Step 2 — Configure your connection mode

API Pull
CLI

Prizm authenticates against your Airflow REST API and polls on a schedule. This requires a network path from Prizm to your Airflow webserver.Prizm auto-detects whether your Airflow is running REST API v1 (/api/v1) or v2 (/api/v2) based on the reported version, so upgrading Airflow between major API versions does not require changing the source config.

Additional fields (API Pull only)

Field	Description
Server	Your Airflow webserver’s base URL (e.g., `https://airflow.internal`)
Authentication Type	`Username/Password` or `Base Auth Key` (pre-built Basic auth header value)
Username / Password	Airflow account credentials — encrypted at rest in Prizm
Base Auth Key	Alternative to username/password — encrypted at rest
Use Vault	Source credentials from a configured vault at connection time instead of storing them in Prizm
No. of Runs / Run History Unit	How much run history each sync window pulls (e.g., last 30 runs, or last 7 days). Limits backfill so Prizm doesn’t pull Airflow’s entire history on every sync.

Schedule

Prizm polls daily at 02:30 UTC by default. On transient failures, Prizm retries automatically up to 3 times before skipping that sync window. You can adjust the schedule from the connector detail page after saving.

Required Airflow permissions

The account used for API Pull needs read access to DAG and Task resources in Airflow. In Airflow’s built-in RBAC, the Viewer role covers this.

After saving, trigger a manual sync immediately from the source page to confirm the connection before waiting for the scheduled window.

The CLI runs on your machine or in CI — it reads from Airflow and uploads to Prizm in a single command. Prizm never needs inbound network access to your Airflow, and never stores an Airflow credential.

Install

pip install prizm-airflow-cli

Step 1 — Dry run (recommended first)

A dry run confirms the CLI can reach and authenticate against Airflow and lets you inspect the extracted bundle — without uploading anything to Prizm or needing your Prizm API token. It is the fastest way to isolate an Airflow-side problem from a Prizm-side one.

export AIRFLOW_PASSWORD='your-airflow-password'

prizm-airflow extract \
  --prizm-url "https://your-workspace.prizmdata.ai" \
  --connection-name "<source name from Step 1>" \
  --hostname "https://your-airflow-webserver" \
  --username "<airflow-username>" \
  --dry-run

Inspect airflow_bundle.json — you should see your DAGs, their tasks, and recent runs.

Step 2 — Upload

Drop --dry-run and supply the Prizm API token via environment variable:

export AIRFLOW_PASSWORD='your-airflow-password'
export PRIZM_API_TOKEN='your-prizm-token'

prizm-airflow extract \
  --prizm-url "https://your-workspace.prizmdata.ai" \
  --connection-name "<source name from Step 1>" \
  --hostname "https://your-airflow-webserver" \
  --username "<airflow-username>" \
  --runs 30

Do not pass credentials as command-line flags (--password, --api-token). Anything on the command line lands in shell history and is visible in CI build logs. Always use environment variables for secrets.

Known issue — upload endpoint: In the current release, the CLI’s derived upload path may not match Prizm’s ingest route, causing a 404. If you see a 404 on upload, add the endpoint explicitly:

--endpoint-url "https://your-workspace.prizmdata.ai/api/v1/schedule/airflow/ingest"

CLI flag reference

Flag	Required	Default	Notes
`--prizm-url`	Yes	—	Your Prizm workspace base URL
`--endpoint-url`	See note above	Derived from `--prizm-url`	Full ingest URL; overrides the derived default
`--api-token`	Yes (upload)	—	Prizm API access token. Prefer `PRIZM_API_TOKEN` env var.
`--client-secret`	Yes (upload)	—	Alias for `--api-token`
`--connection-name`	Yes	—	Must match the source Name in Prizm exactly
`--hostname`	No	`http://localhost:8080`	Airflow webserver URL
`--username`	Yes	—	Airflow account username
`--password`	Yes	—	Airflow password. Prefer `AIRFLOW_PASSWORD` env var.
`--runs`	No	`30`	Number of recent DAG runs to extract
`--tasks`	No	`true`	Include task-level metadata
`--extract-source-code`	No	`true`	Include each DAG’s Python source. Required for cross-DAG lineage and PR impact analysis.
`--dry-run`	No	off	Write the bundle to `airflow_bundle.json` instead of uploading
`--out`	No	`airflow_bundle.json`	Output path when using `--dry-run`

Keeping it fresh

Nothing runs on a schedule for CLI mode — Prizm receives data only when you run the command. Wire it into a cron job or CI pipeline step to automate it, and monitor the exit code. A silently failing cron job looks identical to an Airflow environment that stopped changing.

Step 3 — Verify your connection

After setup, confirm the following. If any step fails, see the troubleshooting section below.

Check	Where to look	Expected
Source is connected	Sources → your Airflow source	No error state on the source record
DAGs cataloged	Assets, filtered to the source	One asset per active DAG
Tasks cataloged	Any DAG asset → children	One asset per task, parented to its DAG
Run history present	DAG asset → Metric tab	Recent runs with states and durations
Metrics evaluating	DAG asset → Metric tab	Freshness (DAG) and Execution Time (task) populated
Lineage edges	Lineage graph	Edges to warehouse tables (Snowflake only — see Lineage)

Troubleshooting

Symptom	Likely cause	Fix
CLI upload fails with 404	Ingest endpoint mismatch	Pass `--endpoint-url` explicitly — see the note in Step 2.
CLI or Plugin push lands nowhere	`connection_name` doesn’t match source Name	Compare character for character, including case.
401 / 403 from Airflow	Insufficient credentials or role	Verify the account has read access to DAG and Task resources.
Source connects but shows zero DAGs	Sync hasn’t run yet	Trigger a manual sync, or run the CLI with `--dry-run` and inspect the bundle.
No lineage edges appear	Warehouse not supported, or query tags absent	Check the warehouse support matrix.

Next Steps

Plugin Setup

Set up real-time push-based metadata with the Prizm Airflow Plugin.

What We Collect

Full field-level breakdown of every metadata object Prizm extracts from Airflow.

Get Started

Core Concepts

Platform

Architecture

Data Sources

Integrations

Security & Access Control

Deployment

Help

Settings

MCP

Release Notes

API Reference

Step 1 — Create an Airflow source

Identity fields

Lineage & profiling fields

Alert fields

Step 2 — Configure your connection mode

Additional fields (API Pull only)

Schedule

Required Airflow permissions

Install

Step 1 — Dry run (recommended first)

Step 2 — Upload

CLI flag reference

Keeping it fresh

Step 3 — Verify your connection

Troubleshooting

Next Steps

Plugin Setup

What We Collect

​Step 1 — Create an Airflow source

​Identity fields

​Lineage & profiling fields

​Alert fields

​Step 2 — Configure your connection mode

​Additional fields (API Pull only)

​Schedule

​Required Airflow permissions

​Install

​Step 1 — Dry run (recommended first)

​Step 2 — Upload

​CLI flag reference

​Keeping it fresh

​Step 3 — Verify your connection

​Troubleshooting

​Next Steps

Plugin Setup

What We Collect

Step 1 — Create an Airflow source

Identity fields

Lineage & profiling fields

Alert fields

Step 2 — Configure your connection mode

Additional fields (API Pull only)

Schedule

Required Airflow permissions

Install

Step 1 — Dry run (recommended first)

Step 2 — Upload

CLI flag reference

Keeping it fresh

Step 3 — Verify your connection

Troubleshooting

Next Steps