Vendor VSS source as files instead of a submodule (fix Windows build)

[phillipleblanc]

Summary

Replaces the extension/vss/upstream git submodule (→ duckdb-vss) with a committed copy of duckdb-vss's src/. Fixes the Windows build break introduced by the static-VSS work (#16).

Why

duckdb-vss has its own nested duckdb and extension-ci-tools submodules. Cargo recursively checks out all submodules of the duckdb-rs git dependency, so the vss submodule dragged in a full nested duckdb checkout — whose deeply-nested Swift example path exceeds Windows' MAX_PATH (260 chars):

failed to update submodule extension/vss/upstream
  failed to update submodule duckdb
  path too long: 'C:/Users/runneradmin/.cargo/git/checkouts/duckdb-rs-.../extension/vss/upstream/
    duckdb/tools/swift/duckdb-swift/Examples/SwiftUI/ExoplanetExplorer.xcodeproj/.../IDEWorkspaceChecks.plist'
  class=Filesystem (30)

Linux/macOS were unaffected, so it only surfaced on the Windows CI build. A vendored copy has no submodules for Cargo to recurse into.

Change

• Remove the extension/vss/upstream submodule and .gitmodules.
• Add extension/vss/src/ — a committed copy of duckdb/duckdb-vss@b833341c's src/ (vss_extension.cpp, hnsw/*.cpp, and the header-only usearch/fp16/simsimd headers + their LICENSEs). Same ABI-matched commit as before.
• vss_config.py: paths extension/vss/upstream/src → extension/vss/src.
• SKILL.md: documents the vendored approach and why it must stay vendored (the Windows MAX_PATH reason); upgrade procedure now re-vendors the source instead of bumping a submodule.

The generated duckdb.tar.gz content is unchanged (same vss source files); only how they're stored in the tree changes. This is step 1 of the re-chain (duckdb-rs → table-providers → spiceai duckdb#11107 follow).

[Copilot]

Pull request overview

This pull request removes the extension/vss/upstream git submodule and instead vendors the DuckDB VSS (HNSW) extension source directly under extension/vss/src/ to avoid Cargo recursively fetching nested submodules that break Windows builds due to MAX_PATH limitations.

Changes:

• Replace the VSS upstream submodule with a committed copy of duckdb-vss’s src/ tree (including header-only deps).
• Update extension/vss/vss_config.py to point to the vendored extension/vss/src paths.
• Update extension/vss/SKILL.md to document why vendoring is required and how to upgrade.

Reviewed changes

Copilot reviewed 38 out of 39 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
.gitmodules	Removes VSS submodule entry.
extension/vss/vss_config.py	Updates packaging paths to use vendored extension/vss/src.
extension/vss/SKILL.md	Documents the vendored approach + upgrade procedure.
extension/vss/src/vss_extension.cpp	Vendored VSS extension entrypoint/registration.
extension/vss/src/hnsw/CMakeLists.txt	Vendored build-source list for HNSW sources.
extension/vss/src/hnsw/hnsw_index.cpp	Vendored HNSW index implementation (usearch-backed).
extension/vss/src/hnsw/hnsw_index_macros.cpp	Vendored SQL macros registration for VSS helpers.
extension/vss/src/hnsw/hnsw_index_physical_create.cpp	Vendored physical operator to build HNSW indexes.
extension/vss/src/hnsw/hnsw_index_physical_create.hpp	Vendored header for physical create-index operator.
extension/vss/src/hnsw/hnsw_index_plan.cpp	Vendored index planning for CREATE INDEX (HNSW).
extension/vss/src/hnsw/hnsw_index_pragmas.cpp	Vendored pragmas (compact/info) for HNSW indexes.
extension/vss/src/hnsw/hnsw_index_scan.cpp	Vendored index scan table function implementation.
extension/vss/src/hnsw/hnsw_optimize_expr.cpp	Vendored expression optimizer rules for distance funcs.
extension/vss/src/hnsw/hnsw_optimize_join.cpp	Vendored join optimizer for HNSW-based join rewrite.
extension/vss/src/hnsw/hnsw_optimize_scan.cpp	Vendored TopN→index-scan optimizer rewrite.
extension/vss/src/hnsw/hnsw_optimize_topk.cpp	Vendored min_by→index-scan/top-k optimizer rewrite.
extension/vss/src/hnsw/hnsw_topk_operator.cpp	Vendored stub TopK operator registration hook.
extension/vss/src/include/aggregate_function_matcher.hpp	Vendored matcher helper header for optimizer rules.
extension/vss/src/include/vss_extension.hpp	Vendored extension class declaration.
extension/vss/src/include/hnsw/hnsw.hpp	Vendored HNSW module registration header.
extension/vss/src/include/hnsw/hnsw_index.hpp	Vendored HNSW index type header.
extension/vss/src/include/hnsw/hnsw_index_scan.hpp	Vendored HNSW index scan bind/function header.
extension/vss/src/include/fp16/LICENSE	Vendored fp16 license.
extension/vss/src/include/fp16/bitcasts.h	Vendored fp16 header-only dependency.
extension/vss/src/include/fp16/fp16.h	Vendored fp16 header-only dependency.
extension/vss/src/include/usearch/LICENSE	Vendored usearch license.
extension/vss/src/include/usearch/duckdb_usearch.hpp	Vendored DuckDB wrapper for usearch config macros.
extension/vss/src/include/usearch/index.hpp	Vendored usearch header-only dependency.
extension/vss/src/include/usearch/index_dense.hpp	Vendored usearch header-only dependency.
extension/vss/src/include/usearch/index_plugins.hpp	Vendored usearch header-only dependency.
extension/vss/src/include/simsimd/LICENSE	Vendored simsimd license.
extension/vss/src/include/simsimd/types.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/binary.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/geospatial.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/probability.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/dot.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/simsimd.h	Vendored simsimd header-only dependency.
extension/vss/src/include/simsimd/spatial.h	Vendored simsimd header-only dependency.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.