Problem
Connector documentation under v2.0.x-SNAPSHOT/connectors/ has fallen behind Collate's ai-2.0 docs in several ways:
1. Overview pages use an outdated template
Every connector overview page is built from a generic MetadataIngestionUi/Steps component pattern, rather than the fully-written-out single-page format used in Collate's ai-2.0 docs. This makes the OSS docs visually and structurally inconsistent with the reference implementation, and harder to maintain since content is split across shared generic snippets instead of being explicit per connector.
2. Four connectors are missing entirely
IOMETE, Google Drive, Omni, and Prefect have no documentation under v2.0.x-SNAPSHOT/connectors/ at all — no overview page, no yaml.mdx, no troubleshooting.mdx, no docs.json entry, and no card on the connectors landing page or homepage integration grid.
3. Per-connector screenshots are stale
Metadata Ingestion screenshots (Select Service / Add New Service Name / Configure Service Connection) for existing connectors were carried over from pre-2.0 assets rather than the current ai-2.0 reference images. Affects 256 images across all 10 connector categories.
4. Collate-only content has leaked into OSS docs
20 database connector pages contain a "Hybrid Ingestion Runner" tip block referencing docs.getcollate.io — a Collate-only concept with a link to Collate's own docs site, inapplicable to OpenMetadata OSS.
5. Ingestion guide pages are out of date with actual product behavior
- Lineage/Usage workflow pages document an old wizard-style UI instead of the current Agent-based UI.
- Spark Lineage only documents the deprecated custom
openmetadata-spark-agent.jar, with no mention of the native OpenLineage Spark listener integration now supported (POST /api/v1/openlineage/lineage).
- Export Lineage only documents CSV export; PNG and PDF export are undocumented.
- Unity Catalog's incremental extraction page states an explicit
GRANT SELECT is required on information_schema.tables, which contradicts Databricks' own documentation.
6. Writing-style checklist violations
Latin abbreviations (i.e./e.g./etc.), arrow characters in UI navigation paths (→ instead of **X** > **Y**), and headings with no body sentence before the next heading/image/code block, scattered across connector pages.
7. No connector search/filter on the v2.0.x-SNAPSHOT connectors page
Unlike Collate's ai-2.0 connectors page, v2.0.x-SNAPSHOT/connectors.mdx has no way to search or filter the full connector list.
Tracking
Addressed by #389.
Problem
Connector documentation under
v2.0.x-SNAPSHOT/connectors/has fallen behind Collate's ai-2.0 docs in several ways:1. Overview pages use an outdated template
Every connector overview page is built from a generic
MetadataIngestionUi/Stepscomponent pattern, rather than the fully-written-out single-page format used in Collate's ai-2.0 docs. This makes the OSS docs visually and structurally inconsistent with the reference implementation, and harder to maintain since content is split across shared generic snippets instead of being explicit per connector.2. Four connectors are missing entirely
IOMETE, Google Drive, Omni, and Prefect have no documentation under
v2.0.x-SNAPSHOT/connectors/at all — no overview page, noyaml.mdx, notroubleshooting.mdx, nodocs.jsonentry, and no card on the connectors landing page or homepage integration grid.3. Per-connector screenshots are stale
Metadata Ingestion screenshots (Select Service / Add New Service Name / Configure Service Connection) for existing connectors were carried over from pre-2.0 assets rather than the current ai-2.0 reference images. Affects 256 images across all 10 connector categories.
4. Collate-only content has leaked into OSS docs
20 database connector pages contain a "Hybrid Ingestion Runner" tip block referencing
docs.getcollate.io— a Collate-only concept with a link to Collate's own docs site, inapplicable to OpenMetadata OSS.5. Ingestion guide pages are out of date with actual product behavior
openmetadata-spark-agent.jar, with no mention of the native OpenLineage Spark listener integration now supported (POST /api/v1/openlineage/lineage).GRANT SELECTis required oninformation_schema.tables, which contradicts Databricks' own documentation.6. Writing-style checklist violations
Latin abbreviations (i.e./e.g./etc.), arrow characters in UI navigation paths (
→instead of**X** > **Y**), and headings with no body sentence before the next heading/image/code block, scattered across connector pages.7. No connector search/filter on the v2.0.x-SNAPSHOT connectors page
Unlike Collate's ai-2.0 connectors page,
v2.0.x-SNAPSHOT/connectors.mdxhas no way to search or filter the full connector list.Tracking
Addressed by #389.