align 26.03 support matrix, release notes, and library quickstart with LanceDB default

[kheiss-uwzoo]

Summary

Closes gaps where extraction docs still read as Milvus-only or inconsistent with LanceDB as the default vector database (Milvus remains fully documented as an alternative).

Changes

• docs/docs/extraction/support-matrix.md — Retrieval bullet now states embedding/indexing target LanceDB (default) or Milvus, with a link to Data Upload.
• docs/docs/extraction/releasenotes-nv-ingest.md — Adds a 26.03 highlight that LanceDB is the default backend (Milvus still supported) and links to Data Upload; clarifies hybrid search for LanceDB; fixes “Lancedb” → LanceDB.
• docs/docs/extraction/quickstart-library-mode.md — States default library path is LanceDB (no milvus-lite required); describes optional Milvus; main example uses vdb_op="lancedb"; Step 3 is labeled Milvus retrieval (nvingest_retrieval) with a pointer to Data Upload for LanceDB retrieval/hybrid.

Notes

• Aligns with existing extraction pages (data-store.md, Compose quickstart, FAQ, Python API) that already describe LanceDB-first behavior.

[greptile-apps]

Greptile Summary

This PR aligns the 26.03 extraction docs with LanceDB as the default vector database backend, updating the support matrix, release notes, library quickstart, and the custom-metadata intro. Three of the four files are clean, but custom-metadata.md was only partially migrated.

• custom-metadata.md internal contradiction (P1): The intro was updated to reference LanceDB SQL predicates, but the "Use Custom Metadata to Filter Results During Retrieval" section (lines 109–172) still instructs users to write Milvus boolean expression syntax and links to the Milvus docs — directly contradicting the updated intro on the same page.

Confidence Score: 4/5

Safe to merge after resolving the internal contradiction in custom-metadata.md's filtering section.

A single P1 documentation inconsistency in custom-metadata.md (intro says LanceDB SQL, body says Milvus syntax) caps the score at 4. All other changes are clean and well-scoped.

docs/docs/extraction/custom-metadata.md — the retrieval/filtering section (lines 109–172) still needs to be updated or clearly labeled as Milvus-only to match the updated intro.

Important Files Changed

Filename	Overview
docs/docs/extraction/custom-metadata.md	Intro updated to reference LanceDB SQL, but the 'Filter Results During Retrieval' section (lines 109–172) still directs users to Milvus boolean expression syntax and uses Milvus-only code examples — leaving the page internally contradictory after this PR.
docs/docs/extraction/quickstart-library-mode.md	Main ingestion example updated to LanceDB; Milvus retrieval section relabeled as optional with a pointer to data-store.md. Minor: one reference to data-store.md uses backtick code format instead of a clickable hyperlink (line 120).
docs/docs/extraction/releasenotes-nv-ingest.md	26.03 highlights correctly updated: LanceDB-as-default bullet added, 'Lancedb' typo fixed to 'LanceDB', hybrid search description expanded.
docs/docs/extraction/support-matrix.md	Retrieval bullet updated from Milvus-only to 'LanceDB (default) or Milvus' with a link to Data Upload. Change is clean and consistent.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[User follows Library Mode Quickstart] --> B[.vdb_upload with vdb_op=lancedb]
    B --> C[Vectors stored in ./lancedb - default path]
    C --> D{Query path chosen}
    D -->|LanceDB default| E[See data-store.md for LanceDB API and hybrid search]
    D -->|Milvus alternative| F[nvingest_retrieval against Milvus collection]

    G[User follows custom-metadata.md] --> H[Intro: use LanceDB SQL predicates]
    H --> I{Filtering section}
    I -->|Line 7 intro| J[LanceDB SQL - correct]
    I -->|Lines 109-113 body| K[Milvus boolean expression syntax - contradicts intro]

Loading

Comments Outside Diff (1)

1. docs/docs/extraction/custom-metadata.md, line 109-113 (link)

   Filtering section contradicts the updated intro

   The intro (line 7) was updated to say "use predicates supported by LanceDB SQL", but the "Use Custom Metadata to Filter Results During Retrieval" section immediately below still says to use Milvus boolean expression syntax and links to the Milvus docs. This means a user following the LanceDB (default) path will encounter directly contradictory instructions on the same page. The filter expression examples (content_metadata["category"] == "technical") also use Milvus filter syntax, not LanceDB SQL.

   At minimum, the section heading and introductory sentence need to clarify this is Milvus-only guidance, mirroring the approach taken in quickstart-library-mode.md where the Milvus step is explicitly labeled as optional/alternative.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: docs/docs/extraction/custom-metadata.md
   Line: 109-113
   
   Comment:
   **Filtering section contradicts the updated intro**
   
   The intro (line 7) was updated to say "use predicates supported by [LanceDB SQL](https://lancedb.github.io/lancedb/sql/)", but the "Use Custom Metadata to Filter Results During Retrieval" section immediately below still says to use Milvus boolean expression syntax and links to the Milvus docs. This means a user following the LanceDB (default) path will encounter directly contradictory instructions on the same page. The filter expression examples (`content_metadata["category"] == "technical"`) also use Milvus filter syntax, not LanceDB SQL.
   
   At minimum, the section heading and introductory sentence need to clarify this is Milvus-only guidance, mirroring the approach taken in `quickstart-library-mode.md` where the Milvus step is explicitly labeled as optional/alternative.
   
   How can I resolve this? If you propose a fix, please make it concise.

Prompt To Fix All With AI

This is a comment left during a code review.
Path: docs/docs/extraction/custom-metadata.md
Line: 109-113

Comment:
**Filtering section contradicts the updated intro**

The intro (line 7) was updated to say "use predicates supported by [LanceDB SQL](https://lancedb.github.io/lancedb/sql/)", but the "Use Custom Metadata to Filter Results During Retrieval" section immediately below still says to use Milvus boolean expression syntax and links to the Milvus docs. This means a user following the LanceDB (default) path will encounter directly contradictory instructions on the same page. The filter expression examples (`content_metadata["category"] == "technical"`) also use Milvus filter syntax, not LanceDB SQL.

At minimum, the section heading and introductory sentence need to clarify this is Milvus-only guidance, mirroring the approach taken in `quickstart-library-mode.md` where the Milvus step is explicitly labeled as optional/alternative.

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

---

This is a comment left during a code review.
Path: docs/docs/extraction/quickstart-library-mode.md
Line: 120

Comment:
**Unlinked reference to `data-store.md`**

`data-store.md` is formatted as inline code rather than a markdown hyperlink, so it won't be clickable in the rendered docs. Every other reference to this file in the changed files uses the `[Data Upload](data-store.md)` link pattern.

```suggestion
This step is optional and shows how to query your ingested content from Milvus. If you are using the default LanceDB path, see [Data Upload](data-store.md) for LanceDB query examples.
```

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

_{Reviews (4): Last reviewed commit: "updated file per Greptile review" | Re-trigger Greptile}