Consumer ergonomics for the Morris cutover (hash / store / registry / preset / emulator)

[dratner]

Additive, app-neutral helpers the Morris team asked for to smooth their cutover. No new dependencies; no go.mod change. Each leaves policy with the app.

What's added

• content — SHA256HexBytes / SHA256HexString / SHA256HexReader(ctx) content-addressing primitives (the §10 hash helper). Documented as primitives only: the app decides whether a digest is raw-source / artifact / chunk / cache identity. SHA256HexReader streams and returns the byte count + honors ctx.
• store — IsNotFound(err) and DeleteIfExists(ctx, s, key). Keeps ObjectStore.Delete strict (the contract stays ErrObjectNotFound) while giving Morris a clean idempotent path for rollback/cleanup.
• extract — Registry.Supports + Registry.SupportedMediaTypes() (sorted, canonical — same normalization as Extract); and TextArtifacts / SingleTextArtifact. Per review, "text" is structural (Blob == nil && Text != ""), and SingleTextArtifact returns ErrNoTextArtifact / ErrMultipleTextArtifacts so multi-artifact resolution stays the app's call.
• extract/preset — opt-in document bundle: RegisterDocuments, NewDocumentRegistry(opts…), SupportedDocumentMediaTypes() (delegates to the registry so it can't drift), wiring text/html/pdf/docx/markdown. Added to the depguard deny-list; importing it pulls the format deps by design while core stays clean.
• store/gcs — NewEmulator(ctx, bucket, endpoint) convenience over WithEndpoint + WithoutAuthentication for dev/test wiring.

Tests

New unit tests for every helper (preset & store-helpers at 100%), plus NewWithClient panic coverage, NewEmulator validation, and an emulator round-trip in the integration suite.

Decisions carried from review

• text-artifact = structural (Blob == nil && Text != ""), not MediaType-based.
• preset lives at extract/preset with RegisterDocuments / NewDocumentRegistry / SupportedDocumentMediaTypes.

Verification

make lint, go test ./..., go vet, go test -race, and make test-integration all pass under go1.26.4. Core remains dependency-free (go list -deps ./store and ./extract show 0 cloud/adapter deps).

After merge this would be v0.2.0 (Morris bumps from v0.1.0 to pick these up).

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR adds small, additive helper APIs across content, store, and extract to improve consumer ergonomics (hashing primitives, idempotent delete helpers, extractor registry introspection + text-artifact helpers), plus a convenience “document bundle” registry preset and a GCS emulator constructor—keeping core packages dependency-free and pushing optional deps into subpackages.

Changes:

• Add SHA-256 lowercase-hex helpers in content (bytes/string/streaming reader with ctx support).
• Add store-level helpers (IsNotFound, DeleteIfExists) and GCS adapter convenience (NewEmulator) with accompanying unit/integration tests.
• Add extractor registry helpers (Supports, SupportedMediaTypes, TextArtifacts, SingleTextArtifact) and a extract/preset bundle, plus depguard enforcement for the new preset package.

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
store/helpers.go	Adds IsNotFound and DeleteIfExists helper APIs on top of store.ObjectStore semantics.
store/helpers_test.go	Unit tests covering the new store helper behaviors.
store/gcs/gcs.go	Adds NewEmulator convenience constructor for emulator endpoint + no-auth configuration.
store/gcs/gcs_test.go	Adds unit tests for NewWithClient panic behavior and NewEmulator validation.
store/gcs/gcs_integration_test.go	Adds an integration round-trip test using NewEmulator.
extract/preset/preset.go	Introduces opt-in document registry bundle wiring common extractors.
extract/preset/preset_test.go	Tests for preset registry support, extraction, sorting, and option pass-through.
extract/helpers.go	Adds Registry.Supports, Registry.SupportedMediaTypes, and text-artifact helpers + errors.
extract/helpers_test.go	Unit tests for new extract helpers.
content/hash.go	Adds SHA-256 lowercase-hex primitives, including streaming reader hashing with ctx awareness.
content/hash_test.go	Unit tests for hash helpers including known vectors and ctx cancellation.
.golangci.yaml	Extends depguard deny-list to prevent core packages importing extract/preset.

---

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