docs: sync main extraction docs with 26.05

[kheiss-uwzoo]

Summary

• Confirmed 14 documentation files under docs/ differed between main and 26.05.
• Synced main to match the 26.05 documentation content exactly (verified with git diff upstream/26.05 -- docs/ showing no remaining differences).
• Changes include release notes updates, support matrix cleanup, custom metadata doc refresh, notebooks page path consolidation (notebooks/index.md → notebooks.md), and mkdocs.yml nav updates.

Files changed

• docs/docs/extraction/audio-video.md
• docs/docs/extraction/concepts.md
• docs/docs/extraction/custom-metadata.md
• docs/docs/extraction/deployment-options.md
• docs/docs/extraction/faq.md
• docs/docs/extraction/getting-started-about.md
• docs/docs/extraction/integrations-langchain-llamaindex-haystack.md
• docs/docs/extraction/multimodal-extraction.md
• docs/docs/extraction/notebooks.md (renamed from notebooks/index.md)
• docs/docs/extraction/overview.md
• docs/docs/extraction/prerequisites-support-matrix.md
• docs/docs/extraction/releasenotes.md
• docs/docs/extraction/troubleshoot.md
• docs/mkdocs.yml

Test plan

• Verify git diff upstream/26.05 -- docs/ is empty on this branch
• Confirm MkDocs site builds without nav/link errors
• Spot-check release notes and support matrix pages render correctly

[greptile-apps]

Greptile Summary

This PR syncs 14 documentation files under docs/ from the 26.05 release branch to main, covering release notes, the support matrix, the notebooks page rename, and mkdocs.yml nav/redirect updates.

• Notebooks page rename: notebooks/index.md → notebooks.md; all internal cross-links and the mkdocs.yml nav entry updated consistently. The stale reverse-redirect is correctly removed.
• Support matrix & release notes: CUDA/Driver requirements and the default OCR NIM updated to 26.05 values; Nemotron Parse and Omni chart-captioning callouts removed across prerequisites-support-matrix.md, multimodal-extraction.md, faq.md, and troubleshoot.md.
• custom-metadata.md: Page rewritten with a simpler intro, but the new ingestion code example omits the three metadata sidecar parameters from .vdb_upload(), so users who copy it will ingest without metadata attached.

Confidence Score: 3/5

Safe to merge after addressing the incomplete metadata ingestion example and verifying the 26.03 docs URL.

The custom-metadata.md rewrite creates a code example that appears to demonstrate metadata ingestion but silently omits the required sidecar parameters, leaving users with a working-looking snippet that does nothing. The 26.03 release notes URL was changed to a format that does not match the versioning convention used by every other link on the same page and may resolve to a 404.

docs/docs/extraction/custom-metadata.md (incomplete vdb_upload example) and docs/docs/extraction/releasenotes.md (26.03 URL format mismatch).

Important Files Changed

Filename	Overview
docs/docs/extraction/custom-metadata.md	Page restructured and significantly shortened; the new vdb_upload code example is missing the three metadata sidecar parameters, so it does not actually attach metadata.
docs/mkdocs.yml	Nav updated for notebooks.md rename, stale redirect removed; exclude_docs guard weakened from /index.md to index.md with the explanatory comment deleted.
docs/docs/extraction/releasenotes.md	26.05 section condensed to a release-line summary; 26.03 section added inline; the 26.03 versioned docs URL format changed from 26.3.0 to 26.03, inconsistent with all other version links.
docs/docs/extraction/prerequisites-support-matrix.md	CUDA/Driver requirements updated to 12.2/535; OCR NIM updated to nemotron-ocr-v2; B200 limitation note and Nemotron Parse install prerequisites removed in line with 26.05.
docs/docs/extraction/notebooks.md	Renamed from notebooks/index.md; relative link to overview.md corrected; Related Topics section added.
docs/docs/extraction/audio-video.md	GPU pinning note converted to a proper !!! important admonition block; code example formatting corrected.
docs/docs/extraction/troubleshoot.md	Removed the open_clip / nemotron-parse troubleshooting entry, consistent with removing the [nemotron-parse] extra from prerequisites.
docs/docs/extraction/multimodal-extraction.md	Removed the Omni caption-scope callout for chart regions and simplified the OCR engine description to default to nemotron-ocr-v2.
docs/docs/extraction/faq.md	Removed the FAQ entry about PDF chart captioning with Omni, consistent with the multimodal-extraction.md cleanup.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["notebooks/index.md (old)"] -->|renamed to| B["notebooks.md"]
    B --> C["mkdocs.yml nav updated"]
    A -->|reverse redirect removed| D["redirect deleted"]
    C --> E["exclude_docs: /index.md changed to index.md"]

    F["prerequisites-support-matrix.md"] -->|CUDA 12.2 / Driver 535| G["Updated requirements"]
    F -->|nemotron-ocr-v1 to v2| H["OCR NIM updated"]
    F -->|Nemotron Parse removed| I["Affects troubleshoot.md, faq.md, multimodal-extraction.md"]

    J["releasenotes.md"] -->|26.05 condensed| K["26.03 section added inline"]
    K -->|URL 26.3.0 changed to 26.03| L["Possible broken link"]

    M["custom-metadata.md"] -->|Rewritten| N["vdb_upload example"]
    N -->|Missing meta_dataframe, meta_source_field, meta_fields| O["Metadata not attached"]

Loading

Comments Outside Diff (1)

1. docs/docs/extraction/custom-metadata.md, line 62-92 (link)

   Metadata parameters missing from the vdb_upload call

   The example creates meta_df and writes it to meta_file.csv, but the .vdb_upload(...) call on line 83–88 does not pass any of the three required metadata parameters (meta_dataframe, meta_source_field, meta_fields). As written, a user who copies this example will ingest documents without any custom metadata attached — the opposite of what the page advertises. The prose on line 92 says "merge values from meta_df (or file_path) into each document's content_metadata before vdb_upload" but provides no code showing how to do so.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: docs/docs/extraction/custom-metadata.md
   Line: 62-92
   
   Comment:
   **Metadata parameters missing from the `vdb_upload` call**
   
   The example creates `meta_df` and writes it to `meta_file.csv`, but the `.vdb_upload(...)` call on line 83–88 does not pass any of the three required metadata parameters (`meta_dataframe`, `meta_source_field`, `meta_fields`). As written, a user who copies this example will ingest documents **without** any custom metadata attached — the opposite of what the page advertises. The prose on line 92 says "merge values from `meta_df` (or `file_path`) into each document's `content_metadata` before `vdb_upload`" but provides no code showing how to do so.
   
   How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

Fix the following 3 code review issues. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 3
docs/docs/extraction/custom-metadata.md:62-92
**Metadata parameters missing from the `vdb_upload` call**

The example creates `meta_df` and writes it to `meta_file.csv`, but the `.vdb_upload(...)` call on line 83–88 does not pass any of the three required metadata parameters (`meta_dataframe`, `meta_source_field`, `meta_fields`). As written, a user who copies this example will ingest documents **without** any custom metadata attached — the opposite of what the page advertises. The prose on line 92 says "merge values from `meta_df` (or `file_path`) into each document's `content_metadata` before `vdb_upload`" but provides no code showing how to do so.

### Issue 2 of 3
docs/mkdocs.yml:209-212
The `exclude_docs` pattern was changed from `/index.md` to `index.md`, and the comment explaining the distinction was removed. In gitignore syntax, `index.md` (no leading slash) matches every `index.md` at any directory depth, while `/index.md` restricts the match to the docs root only. With `notebooks/index.md` now renamed there is currently only one `index.md` in the tree so the behavior is equivalent — but silently dropping the guard means any future `index.md` added in a subdirectory will be excluded without warning.

```suggestion
# MkDocs 1.6+: exclude suite landing and legacy duplicate pages (still in repo for parity).
# extraction/chunking.md — removed from nav; content is under concepts.md (redirect_maps keeps old URLs).
# Use /index.md (docs root only); bare index.md would exclude every index.md in any subdirectory.
exclude_docs: |
  /index.md
```

### Issue 3 of 3
docs/docs/extraction/releasenotes.md:57
The 26.03 versioned docs URL was changed to `26.03` format, but every other link in this table uses the `YY.M.P` semver-style path (e.g., `25.9.0`, `26.1.2`, `26.1.1`). The old value `26.3.0` was consistent with that pattern; `26.03` is not and may resolve to a 404 on the NVIDIA docs CDN.

```suggestion
| [26.03](https://docs.nvidia.com/nemo/retriever/26.3.0/extraction/releasenotes/)
```

_{Reviews (1): Last reviewed commit: "Update faq.md" | Re-trigger Greptile}

[greptile-apps]

The exclude_docs pattern was changed from /index.md to index.md, and the comment explaining the distinction was removed. In gitignore syntax, index.md (no leading slash) matches every index.md at any directory depth, while /index.md restricts the match to the docs root only. With notebooks/index.md now renamed there is currently only one index.md in the tree so the behavior is equivalent — but silently dropping the guard means any future index.md added in a subdirectory will be excluded without warning.

Suggested change

	# MkDocs 1.6+: exclude suite landing and legacy duplicate pages (still in repo for parity).
	# extraction/chunking.md — removed from nav; content is under concepts.md (redirect_maps keeps old URLs).
	# Use /index.md (docs root only); bare index.md would exclude every index.md (e.g. extraction/notebooks/index.md).
	exclude_docs: |
	/index.md
	index.md
	# MkDocs 1.6+: exclude suite landing and legacy duplicate pages (still in repo for parity).
	# extraction/chunking.md — removed from nav; content is under concepts.md (redirect_maps keeps old URLs).
	# Use /index.md (docs root only); bare index.md would exclude every index.md in any subdirectory.
	exclude_docs: |
	/index.md

Prompt To Fix With AI

This is a comment left during a code review.
Path: docs/mkdocs.yml
Line: 209-212

Comment:
The `exclude_docs` pattern was changed from `/index.md` to `index.md`, and the comment explaining the distinction was removed. In gitignore syntax, `index.md` (no leading slash) matches every `index.md` at any directory depth, while `/index.md` restricts the match to the docs root only. With `notebooks/index.md` now renamed there is currently only one `index.md` in the tree so the behavior is equivalent — but silently dropping the guard means any future `index.md` added in a subdirectory will be excluded without warning.

```suggestion
# MkDocs 1.6+: exclude suite landing and legacy duplicate pages (still in repo for parity).
# extraction/chunking.md — removed from nav; content is under concepts.md (redirect_maps keeps old URLs).
# Use /index.md (docs root only); bare index.md would exclude every index.md in any subdirectory.
exclude_docs: |
  /index.md
```

How can I resolve this? If you propose a fix, please make it concise.

[greptile-apps]

The 26.03 versioned docs URL was changed to 26.03 format, but every other link in this table uses the YY.M.P semver-style path (e.g., 25.9.0, 26.1.2, 26.1.1). The old value 26.3.0 was consistent with that pattern; 26.03 is not and may resolve to a 404 on the NVIDIA docs CDN.

Suggested change

	| [26.03](https://docs.nvidia.com/nemo/retriever/26.03/extraction/releasenotes/)
	| [26.03](https://docs.nvidia.com/nemo/retriever/26.3.0/extraction/releasenotes/)

Prompt To Fix With AI

This is a comment left during a code review.
Path: docs/docs/extraction/releasenotes.md
Line: 57

Comment:
The 26.03 versioned docs URL was changed to `26.03` format, but every other link in this table uses the `YY.M.P` semver-style path (e.g., `25.9.0`, `26.1.2`, `26.1.1`). The old value `26.3.0` was consistent with that pattern; `26.03` is not and may resolve to a 404 on the NVIDIA docs CDN.

```suggestion
| [26.03](https://docs.nvidia.com/nemo/retriever/26.3.0/extraction/releasenotes/)
```

How can I resolve this? If you propose a fix, please make it concise.

[kheiss-uwzoo]

Closing: sync direction was incorrect. main/docs/docs already has the authoritative GA content (especially releasenotes.md). Copying from 26.05 regressed main. Follow-up PR will sync 26.05 to match main instead.