remove custom-metadata.md; canonical path is vdbs + notebooks

[kheiss-uwzoo]

Summary

• Remove docs/docs/extraction/custom-metadata.md (duplicate of notebook + VDB README content added in align metadata docs with VDB filtering guide #2108).
• Expand Metadata and filtering on �dbs.md with a short overview and links to the worked notebooks.
• Drop the separate MkDocs nav entry; add redirect custom-metadata.md -> �dbs.md#metadata-and-filtering.
• Update cross-links and doc-snippet test registry.

Follows Julio's NVBugs 6205401 guidance: VDB/metadata facts live on �dbs.md; runnable walkthroughs stay in notebooks.

Notebooks (canonical examples)

• metadata_and_filtered_search.ipynb
• nemo_retriever_retriever_query_metadata_filter.ipynb

Operator/API reference remains in nemo_retriever/src/nemo_retriever/vdb/README.md.

Test plan

• MkDocs build; confirm redirect from old custom-metadata URL
• Nav no longer lists a separate Custom metadata page
• Notebook links resolve on GitHub

[greptile-apps]

Greptile Summary

Removes two thin doc pages (custom-metadata.md and integrations-langchain-llamaindex-haystack.md), consolidates their content into vdbs.md#metadata-and-filtering, and adds MkDocs redirects so existing URLs continue to resolve.

• vdbs.md gains a full prose overview of the sidecar metadata workflow (meta_dataframe/meta_source_field/meta_fields), retriever-service upload flow, and links to both worked notebooks; cross-links in six other doc files updated.
• notebooks/index.md now lists both canonical metadata notebooks (metadata_and_filtered_search.ipynb and nemo_retriever_retriever_query_metadata_filter.ipynb), addressing the gap noted in a prior review thread.
• The doc-snippet test registry swaps custom-metadata.md for vdbs.md; since vdbs.md contains no Python fenced blocks the entry is inert, though coverage is preserved via the .ipynb entry already in the tuple.

Confidence Score: 5/5

Safe to merge; all changes are documentation and redirects with no runtime code paths affected.

The PR deletes two doc pages, expands a section on vdbs.md, adds MkDocs redirects, and updates cross-links. No Python logic, APIs, or data paths are modified. The redirects are correctly wired, both canonical notebooks are now linked, and the previously flagged sparse metadata section has been filled in.

nemo_retriever/tests/test_src_documentation_snippets.py — the new _PUBLIC_RETRIEVER_DOCS entry for vdbs.md produces zero test blocks today; worth noting if Python examples are ever added to that page in the future.

Important Files Changed

Filename	Overview
docs/docs/extraction/vdbs.md	Metadata and filtering section expanded with overview prose, sidecar parameter details, service workflow, and links to the worked notebook; cross-links updated throughout.
nemo_retriever/tests/test_src_documentation_snippets.py	Swapped custom-metadata.md for vdbs.md in _PUBLIC_RETRIEVER_DOCS; vdbs.md currently has no Python fenced blocks, so the entry contributes zero coverage to the constructor-kwargs and syntax tests.
docs/docs/extraction/notebooks/index.md	Both canonical metadata notebooks are now listed; prior thread concern about missing metadata_and_filtered_search.ipynb link is addressed.
docs/mkdocs.yml	Removed custom-metadata and integrations nav entries; sections renumbered; two redirects added for deleted pages.
docs/docs/extraction/custom-metadata.md	Deleted; content consolidated into vdbs.md and notebooks; redirect configured in mkdocs.yml.
docs/docs/extraction/integrations-langchain-llamaindex-haystack.md	Deleted; thin wrapper page replaced by redirect to notebooks/index.md which already lists the LangChain and LlamaIndex examples.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    OLD_CM["custom-metadata.md\n(deleted)"]
    OLD_INT["integrations-langchain-llamaindex-haystack.md\n(deleted)"]
    REDIRECT1["redirect:\nextraction/custom-metadata.md\n→ vdbs.md#metadata-and-filtering"]
    REDIRECT2["redirect:\nextraction/integrations-…\n→ notebooks/index.md"]
    VDBS["vdbs.md\n#metadata-and-filtering\n(expanded)"]
    NB_INDEX["notebooks/index.md\n(both metadata notebooks listed)"]
    NB1["metadata_and_filtered_search.ipynb"]
    NB2["nemo_retriever_retriever_query_metadata_filter.ipynb"]
    VDB_README["nemo_retriever/src/nemo_retriever/vdb/README.md\n(canonical operator reference)"]

    OLD_CM -->|content moved to| VDBS
    OLD_CM -->|URL preserved by| REDIRECT1
    OLD_INT -->|URL preserved by| REDIRECT2
    VDBS -->|worked example link| NB2
    VDBS -->|operator reference link| VDB_README
    NB_INDEX --> NB1
    NB_INDEX --> NB2

Loading

_{Reviews (15): Last reviewed commit: "docs(extraction): address PR2195 review ..." | Re-trigger Greptile}

[jperez999]

You sure this is valid? I dont see this file, nor should it exist.

[kheiss-uwzoo]

Agreed — removed in 0c8878e2. examples/reindex_example.ipynb was deleted on main in #2163 and #2226 had already dropped this link from notebooks/index.md; re-adding it was a merge artifact from an older branch state.

[jperez999]

I cant find this ref either? Is this PR outdated, maybe this should be closed without merge?

[kheiss-uwzoo]

The target page exists at docs/docs/extraction/notebooks/index.md (mkdocs nav: Starter kits). The integrations page this PR removes only duplicated LangChain/LlamaIndex links already listed there.

Main still has both custom-metadata.md and integrations-langchain-llamaindex-haystack.md, so the consolidation is still needed — not planning to close without merge.

Renamed the link text to Starter kits in 0c8878e2 to match the nav label.

[jperez999]

same here

[kheiss-uwzoo]

Same as above — notebooks/index.md is the Starter kits page; link label updated to match nav in 0c8878e2.