ingestr v1.0.76 - DuckDB and BigQuery Ingestion Fixes

ingestr v1.0.76, published on July 13, 2026, focuses on database ingestion paths. The main change speeds up and fixes CSV ingestion into DuckDB, while BigQuery and DuckLake receive narrower correctness fixes.

The full release notes and downloads are on the GitHub release page.

CSV ingestion into DuckDB is the broadest runtime change in this release. The changelog says the path is both faster and fixed, but it does not identify the failure case, affected input shape, implementation detail, or benchmark.

That lack of detail limits the operational conclusion. Teams with CSV sources feeding DuckDB should test the same representative files on the current version and on v1.0.76. Compare elapsed load time, resulting row counts, and destination data before moving production jobs.

The combined wording also matters for regression tests. A successful load checks the correctness side, but it does not confirm the performance claim. Likewise, a shorter run does not establish that the unspecified ingestion bug is resolved for every input. The release gives operators a concrete path to inspect, not a universal performance number.

BigQuery column type changes are now batched into one operation. For a destination schema that needs several type adjustments, ingestr now groups that work instead of treating each change as a separate operation.

This is a useful execution change for pipelines that reconcile evolving source schemas with BigQuery. It narrows the schema update phase to one grouped operation. The notes do not state whether the change affects transaction scope, retry behavior, quotas, or error reporting, so those behaviors should not be inferred from the changelog.

During validation, inspect the final BigQuery schema and the job logs for a load that changes more than one column type. That directly exercises the released behavior without assuming anything about internal SQL or API calls that the notes do not document.

The DuckDB integration also fixes a DuckLake attachment problem. An attach alias could collide with the destination schema name, and v1.0.76 changes the path to avoid that collision.

For operators, the relevant test case is a workload where the DuckLake attachment alias and destination schema would otherwise overlap. The release notes do not name a new CLI flag, configuration key, or migration step. This appears to be an internal naming fix rather than a new interface.

The scope is deliberately narrow. The changelog does not claim broader DuckLake compatibility work or general DuckDB namespace changes. It identifies one alias collision and fixes that path.

The remaining changes add a pull request template and adapt it to repository conventions and SOC 2 change management. These entries affect contributor workflow, not ETL runtime behavior.

That distinction is useful for platform teams reviewing the tag. The repository now records more structure around proposed changes, while the deployable behavior remains concentrated in the three database paths above. The notes do not report breaking changes, API changes, or upgrade steps, and the release metadata marks v1.0.76 as a regular release rather than a prerelease.