Skip to main content
← Back to Releases
Oct2026

Ingest in Minutes, Backfill in Parallel, Load into Databricks & Extract MongoDB Faster with Arrow

New users can now sign up and ingest data in minutes. Ingestion is a first-class pipeline type with concurrent backfills and a new Streams tab, and Databricks is now a fully supported data store. This release also adds the tap-onspd UK postcode connector, Arrow batches for faster tap-mongodb extraction, and fixes for target-bigquery, tap-googleads, tap-instagram, target-snowflake, and more.

Features​

  • From new account to ingested data, in minutes: We've streamlined the process for new users to create an account, set up the essentials in a workspace and move data from source to warehouse. Head to Meltano.com, hit TRY MELTANO and start ingesting your data.
  • Ingestion as a first-class action: Ingestion is the simplest form of a pipeline and does what it says on the tin: it moves data from tap to target, fast. Stream selection is also improved, so it's easier to discover streams and use them to populate dbt sources.
  • Backfill improvements: Reloading history for an ingestion pipeline without affecting its incremental sync is now much simpler. An ingestion can have multiple backfills running at the same time, each listed in a new Backfills tab, and each backfill can be configured by stream and start date.
  • New Streams tab: Each pipeline now has its own Streams tab, so you can review discovered streams and export them in the format you need.
  • Databricks data store support: Databricks is now a fully supported data store, alongside Postgres, Snowflake and BigQuery. Run full-load and incremental pipelines with target-databricks, including Apache Arrow bulk loading for much faster loads on big tables, plus dbt transforms with dbt-databricks and pipeline state kept in Databricks.
  • New tap-onspd connector: Add the Office for National Statistics (ONS) Postcode Directory (ONSPD) as an extractor for comprehensive UK postcode data, including latitude, longitude, geographic classifications, and code-to-name lookups for those geographies.
  • Faster MongoDB extraction with Arrow: tap-mongodb v0.4.0 can now send incremental streams as Arrow batches for faster extraction. It also adds document_layout, which can load each top-level field as its own column, and change_stream_pipeline for change stream syncs.

Fixes​

  • Bumped target-bigquery to v1.1.3: Fixes a failed dedupe before upsert on a table with a float key. Timestamps with an offset like +0000 and fields with a period in their name now load.
  • Bumped tap-googleads to v0.15.1: Stops duplicate records when a custom query replaces a default stream, and corrects the setting docs.
  • Bumped dbt-tap-googleads to v0.6.1: The campaign performance by location model now builds on BigQuery.
  • Bumped tap-instagram to v0.5.4: Custom user insight reports no longer fail, and syncs over a long history are much faster because daily and time-series insights are requested in 30-day windows. A new sync writes one demographics snapshot per breakdown, not a copy for every past day.
  • Bumped target-snowflake to v0.20.4: Loaded data can be queried as soon as a run finishes, because the target now waits for its commit before closing. It also no longer pins certifi, which drops the distrusted eMudhra intermediate CAs.
  • Bumped tap-spreadsheets-anywhere to v0.7.2: skip_empty_rows now works for Excel files, and an Azure container is listed once per run instead of repeatedly.
  • Bumped tap-tiktok to v1.2.0: Now installs on Python 3.10.
  • Elementary fixes: Adding app-elementary now gives you its Data Quality Dashboard instead of opening Explore, and no longer resets your default data store to Warehouse. Elementary reports also run on Snowflake (files-dbt v1.6.2, app-elementary v0.4.2). Upgrade your files-dbt bundle to pick this up, and if your default data store was reset to Warehouse, set it back.