crawlkit v0.12.0

Changelog

• Add shared vector.Search backends, including the existing exact cosine scorer and an optional Python turbovec bridge for downstream semantic search over dimensions divisible by 8 up to 8,192. Thanks @vincentkoc.
• Add a remote.Client.UploadSQLite helper, SQLite object metadata in remote status payloads, and contract metadata for publisher-side raw SQLite archive uploads.
• Add gzip-compressed SQLite bundle manifests, chunk builders, and upload helpers for R2-backed remote archive bootstrap/fallback data.

Verification

• Current main at acb97d9e8937aa3d88582a5a1a97920db2603299 was clean and green before tagging.
• GitHub Actions on current main passed: Security Gate, CI, and CodeQL.
• Local release gate passed: GOWORK=off go mod tidy, git diff --exit-code -- go.mod go.sum, GOWORK=off go vet ./..., deadcode, and GOWORK=off go test -count=1 ./....
• Vector race proof passed: GOWORK=off go test -race -count=1 ./vector.
• Real TurboVec proof passed in an isolated temp venv with turbovec==0.7.0 and numpy.
• Downstream gitcrawl compatibility passed in a temp copy with a local crawlkit replace: GOWORK=off go test ./..., --help, --version, metadata --json, and status --json.

Release State Note

Existing signed git tags v0.8.0 through v0.11.0 are authoritative history. GitHub Release pages currently exist only through v0.7.0; this release does not rewrite or backfill older tags.