feat: deterministic manifest schema v2 — the manifest hash is the content address

.github/workflows/ci.yml

@@ -189,6 +189,11 @@ jobs:
         run: cargo test --features test-utils
       - name: cargo test (no default features)
         run: cargo test --no-default-features
+      # The manifest crate embeds and resolves bundle paths, so its path
+      # handling is OS-sensitive; the dedicated ubuntu lane covers its
+      # feature matrix while this step covers the other OSes.
+      - name: cargo test -p ordvec-manifest
+        run: cargo test -p ordvec-manifest
       - name: cargo build --release --features bench-utils --example bench_rank
         run: cargo build --release --features bench-utils --example bench_rank
 

CHANGELOG.md

@@ -9,6 +9,56 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 _No unreleased changes._
 
+## 0.7.0 - 2026-07-13
+
+### Added
+
+- **`ordvec-manifest`: typed verification classification.** Every
+  verification issue code is now a named `pub const` in a `codes` module
+  (zero bare literals at emit sites), with a `#[non_exhaustive]`
+  `VerificationCode` enum and `ReportIssue::classification()` so
+  downstream integrity handling can branch on typed values instead of
+  comparing strings. A missing primary artifact or row-identity file is
+  reported with a dedicated `artifact_missing` / `row_identity_missing`
+  code (classified `ArtifactMissing` / `RowIdentityMissing`) only when the
+  file is genuinely absent; permission or I/O failures keep the generic
+  `*_path_unavailable` code and classify as `Unknown`, so a consumer never
+  mistakes an unreadable file for a missing one. `ReportIssue` gains optional structured mismatch
+  detail — artifact name plus expected/actual SHA-256 and sizes — at the
+  artifact, auxiliary, and row-identity mismatch sites, for lossless
+  downstream error construction. Reports without the new detail
+  serialize byte-identically to before (regression-tested).
+- **`ordvec-manifest`: shared hash helpers.** New `sha256_bytes` and
+  bounded `sha256_reader` share `sha256_file_bounded`'s bounded/EINTR
+  read core; the sqlite registry's private duplicate hasher is deduped
+  onto the public helper.
+
+### Changed
+
+- **BREAKING (`ordvec-manifest`): deterministic manifest schema v2.** The
+  manifest schema version is now `ordvec.index_manifest.v2`. `manifest_id`
+  and `created_at` are removed from `IndexManifest`, creation omits the
+  optional `build` field (serialized as absent, not `null`), and auxiliary
+  artifact entries are sorted by
+  `(name, path)`, so identical bundle content serializes to byte-identical
+  manifests and `sha256(manifest.json)` is the bundle's content address.
+  Existing v1 manifests no longer parse; loading one fails with an error
+  naming both schema versions (zero back-compat, pre-release). Embedded
+  paths must now be canonical — bundle-relative, forward slashes, no `.`,
+  `..`, or empty segments — enforced both at creation (non-embeddable
+  inputs fail `create` instead of minting a manifest that fails its own
+  verification) and at verification (`*_path_not_canonical` codes, now
+  also covering calibration and encoder-distortion profile refs). Absolute
+  paths and escaping `..` paths remain available behind the existing
+  `allow_absolute_paths` / `allow_path_escape` opt-ins. The
+  `write_manifest_file` serialization form is documented as the single
+  canonical byte form: hashing and signing operate on stored bytes, and any
+  serializer change is a schema-version event. The sqlite report registry
+  drops its `manifest_id` column: the cached `verification_reports` table is
+  migrated in place on open (rows preserved, under one atomic transaction),
+  while the rebuildable `active_manifest` pointer is reset and must be
+  re-activated.
+
 ## 0.6.0 - 2026-07-04
 
 ### Performance

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ordvec"
-version = "0.6.0"
+version = "0.7.0"
 edition = "2021"
 rust-version = "1.89" # AVX-512 intrinsics stabilized in 1.89.0; also clears the 1.87 floor from u64::is_multiple_of
 description = "Training-free ordinal & sign quantization for vector retrieval"

README.md

@@ -212,7 +212,7 @@ Details in [`docs/RANK_MODES.md`](docs/RANK_MODES.md).
 
 ```toml
 [dependencies]
-ordvec = "0.6"
+ordvec = "0.7"
 
 # Or, to track unreleased `main`, use a git dependency instead:
 # ordvec = { git = "https://github.com/Project-Navi/ordvec" }

THREAT_MODEL.md

@@ -1,6 +1,6 @@
 # Threat Model — `ordvec`
 
-> **Status:** v0.6.0 (pre-1.0), 2026-06-15. This is the maintained threat model
+> **Status:** v0.7.0 (pre-1.0), 2026-06-15. This is the maintained threat model
 > for the `ordvec` Rust crate, C ABI, Go wrapper, PyO3/maturin Python bindings,
 > and the `ordvec-manifest` sidecar verifier. It is reviewed when the
 > attack surface changes (new persistence formats, new `unsafe` kernels, new

ordvec-ffi/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ordvec-ffi"
-version = "0.6.0"
+version = "0.7.0"
 edition = "2021"
 rust-version = "1.89"
 publish = false

ordvec-manifest-python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ordvec-manifest-python"
-version = "0.6.0"
+version = "0.7.0"
 edition = "2021"
 rust-version = "1.89"
 description = "Python bindings for ordvec-manifest index provenance verification"

ordvec-manifest-python/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "maturin"
 
 [project]
 name = "ordvec-manifest"
-version = "0.6.0"
+version = "0.7.0"
 description = "Python bindings for ordvec index manifest verification"
 readme = "README.md"
 requires-python = ">=3.10"

ordvec-manifest-python/python/ordvec_manifest/__init__.py

@@ -50,4 +50,4 @@
     "create_manifest",
 ]
 
-__version__ = "0.6.0"
+__version__ = "0.7.0"

ordvec-manifest-python/tests/test_manifest_bindings.py

@@ -26,8 +26,6 @@ def write_unloadable_manifest(tmp_path):
     digest = hashlib.sha256(artifact.read_bytes()).hexdigest()
     manifest = {
         "schema_version": ordvec_manifest.SCHEMA_VERSION,
-        "manifest_id": "urn:uuid:7c66ad6e-bdde-49a8-b420-f1136d04f5bd",
-        "created_at": "2026-06-09T00:00:00Z",
         "artifact": {
             "path": artifact.name,
             "sha256": digest,

ordvec-manifest/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "ordvec-manifest"
-version = "0.6.0"
+version = "0.7.0"
 edition = "2021"
 rust-version = "1.89"
 license = "MIT OR Apache-2.0"
@@ -29,7 +29,7 @@ required-features = ["cli"]
 chrono = { version = "0.4.44", default-features = false, features = ["clock", "std"] }
 clap = { version = "4.6.1", features = ["derive"], optional = true }
 hex = "0.4.3"
-ordvec = { version = "0.6.0", path = ".." }
+ordvec = { version = "0.7.0", path = ".." }
 rusqlite = { version = "0.40.0", optional = true }
 serde = { version = "1.0", features = ["derive"] }
 serde_json = "1.0"

ordvec-manifest/README.md

@@ -30,7 +30,12 @@ ordvec-manifest verify --manifest path/to/index.manifest.json
 From a workspace checkout, prefix the same commands with
 `cargo run -p ordvec-manifest --`.
 
-The schema version is `ordvec.index_manifest.v1`. Relative paths resolve from
+The schema version is `ordvec.index_manifest.v2`. The v2 schema is
+deterministic: identical bundle content serializes to identical manifest
+bytes, so `sha256(manifest.json)` is the bundle's content address. Manifests
+carry no `manifest_id` or `created_at`, auxiliary artifact entries are sorted
+by `(name, path)`, and embedded paths must be canonical (bundle-relative,
+forward slashes, no `.`, `..`, or empty segments). Relative paths resolve from
 the manifest file's directory, absolute paths are rejected by default, and
 relative paths may not escape the manifest directory unless explicitly allowed.
 `create` follows the same policy: by default it emits only paths that should
@@ -229,11 +234,11 @@ A consuming database can keep the ordvec row identity as
 `RowIdentity::RowIdIdentity { row_count }` and declare its ID sidecar file as a
 required auxiliary artifact (e.g. `app.ids`). That makes the vector row count an
 ordvec invariant while leaving the caller's `u64` document IDs as caller-owned
-sidecar bytes. Do not encode the ID sidecar as `RowIdentity::Jsonl`: v1 JSONL
+sidecar bytes. Do not encode the ID sidecar as `RowIdentity::Jsonl`: JSONL
 row identity is UUID-oriented (`id_kind = "uuid"`), and generic row-map ID
 formats are intentionally deferred to
 [#145](https://github.com/Project-Navi/ordvec/issues/145). The reserved
-`row_identity.db` metadata block is rejected in v1 because it is not byte-bound
+`row_identity.db` metadata block is rejected because it is not byte-bound
 or path-checked.
 
 Stable row-identity boundary codes:
@@ -273,6 +278,7 @@ Stable sidecar states:
 | `failed` | Code-specific | Path policy, hashing, size, digest, or limit validation failed. |
 
 Common `failed` reason codes include `auxiliary_artifact_path_empty`,
+`auxiliary_artifact_path_not_canonical`,
 `auxiliary_artifact_base_dir_unavailable`,
 `auxiliary_artifact_path_unavailable`,
 `auxiliary_artifact_path_escape_rejected`,
@@ -287,7 +293,6 @@ and records checks that were intentionally not run, such as
 {
   "ok": true,
   "checked_at": "2026-06-03T17:20:00Z",
-  "manifest_id": "urn:uuid:11111111-1111-4111-8111-111111111111",
   "artifact": {
     "manifest_path": "index.ovrq",
     "observed_path": "index.ovrq",
@@ -346,7 +351,6 @@ read and absent when the file is missing:
 {
   "ok": false,
   "checked_at": "2026-06-03T17:21:00Z",
-  "manifest_id": "urn:uuid:11111111-1111-4111-8111-111111111111",
   "artifact": {
     "manifest_path": "index.ovrq",
     "observed_path": "index.ovrq",