add queriable search index for ducklake docs

[guillesd]

DuckLake Docs Search Index

Adds a queryable full-text search index of the DuckLake documentation as a .duckdb file, built automatically during site deployment.

What it does

scripts/generate_search_index.py parses all markdown files under docs/stable/ and docs/preview/, chunks them (one chunk per H2 section), and writes them to a DuckDB database with a pre-built FTS index using porter stemming.

The resulting data/docs-search.duckdb contains a single docs_chunks table with ~344 rows across both versions.

How to query

ATTACH 'https://ducklake.select/data/docs-search.duckdb' AS docs (READ_ONLY);
LOAD fts;
USE docs;

SELECT chunk_id, page_title, section, url, score
FROM (
    SELECT *,
           fts_main_docs_chunks.match_bm25(chunk_id, 'attach catalog') AS score
    FROM docs_chunks
    WHERE version = 'stable'
)
WHERE score IS NOT NULL
ORDER BY score DESC
LIMIT 10;

Schema

Column	Type	Description
chunk_id	VARCHAR (PK)	Unique identifier, e.g. stable/duckdb/usage/connecting#examples
page_title	VARCHAR	Page title from front matter
section	VARCHAR	Section heading (null for page intros)
breadcrumb	VARCHAR	Path-derived breadcrumb, e.g. DuckDB > Usage > Connecting
url	VARCHAR	URL path with anchor, e.g. /docs/stable/duckdb/usage/connecting#examples
version	VARCHAR	stable or preview
text	TEXT	Full preserved markdown text of the chunk