feat(content): VSS-snapshot content-ingest pipeline for Docenta (UFI.0-UFI.6)

.github/workflows/release.yml

@@ -612,6 +612,47 @@ jobs:
             fi
           done
 
+      # Unsigned Rust binaries recurrently trip Defender's ML heuristic, which
+      # blocks the winget-pkgs PR (Validation-Defender-Error) hours after the
+      # release. Scan the freshly-built binaries here so the risk surfaces NOW,
+      # in the release summary, pointing at the remediation. EARLY WARNING ONLY:
+      # a clean result does NOT guarantee winget passes (its cloud/ML validation
+      # uses fresh defs this runner may lack), and a hit never fails the release.
+      - name: Defender false-positive early warning
+        if: contains(matrix.target, 'windows')
+        continue-on-error: true
+        shell: pwsh
+        run: |
+          $mp = (Get-ChildItem "$env:ProgramData\Microsoft\Windows Defender\Platform\*\MpCmdRun.exe" -ErrorAction SilentlyContinue |
+                 Sort-Object FullName -Descending | Select-Object -First 1).FullName
+          if (-not $mp) { $mp = "$env:ProgramFiles\Windows Defender\MpCmdRun.exe" }
+          if (-not (Test-Path $mp)) { "::notice::MpCmdRun not found; skipping Defender early-warning scan."; exit 0 }
+          $sig = (Get-MpComputerStatus -ErrorAction SilentlyContinue).AntivirusSignatureVersion
+          $dir = "target/${{ matrix.target }}/release"
+          $flagged = @()
+          Get-ChildItem "$dir/*.exe" -ErrorAction SilentlyContinue | ForEach-Object {
+            & $mp -Scan -ScanType 3 -File $_.FullName -DisableRemediation *> $null
+            if ($LASTEXITCODE -eq 2) { $flagged += $_.Name }
+          }
+          $tag = "${{ needs.release-preparation.outputs.tag }}"
+          if ($flagged.Count -gt 0) {
+            $list = $flagged -join ", "
+            "::warning::Defender flagged $($flagged.Count) binary(ies) [$list] (defs $sig). This release will likely hit the winget Validation-Defender-Error. Remediate with: just winget-av-submit $tag"
+            @(
+              "### 🛡️ Defender false-positive early warning",
+              "",
+              "**$($flagged.Count) binary(ies) flagged** (signature defs ``$sig``): $list",
+              "",
+              "This release will very likely trip ``Validation-Defender-Error`` on the winget-pkgs PR. Prep the WDSI false-positive submission:",
+              '```',
+              "just winget-av-submit $tag",
+              '```'
+            ) | Out-File -Append -Encoding utf8 $env:GITHUB_STEP_SUMMARY
+          } else {
+            "::notice::Defender early-warning scan clean (defs $sig). Note: winget cloud/ML validation may still differ."
+            "### 🛡️ Defender early-warning scan: clean (defs ``$sig``)" | Out-File -Append -Encoding utf8 $env:GITHUB_STEP_SUMMARY
+          }
+
       - name: Package binaries
         shell: bash
         run: |

.gitignore

@@ -39,7 +39,7 @@ build/
 *.rlib
 *.bin
 # `*.bin` is meant to keep transient generated blobs out of git, but the
-# workspace contains three INTENTIONAL committed binary assets that must
+# workspace contains INTENTIONAL committed binary assets that must
 # remain tracked.  Without these carve-outs, release-plz aborts with
 # "the working directory has uncommitted changes" because it sees these
 # files as both committed AND gitignored — the same R0-pattern issue
@@ -57,6 +57,11 @@ build/
 # `ipc_integration::real_mft_search` test in `uffs-daemon` cannot find
 # its fixture and silently no-ops.
 !tests/fixtures/**/*.bin
+# uffs-content-protocol golden wire-format fixtures (frozen manifest/frame
+# byte samples) — see `crates/uffs-content-protocol/tests/golden_fixtures.rs`.
+# These are the cross-language conformance corpus per the ingest-protocol
+# addendum §5.5; committing the exact bytes is the point, not incidental.
+!crates/uffs-content-protocol/tests/fixtures/*.bin
 /test_scf
 
 # =============================================================================

Cargo.toml

@@ -24,22 +24,27 @@ cargo-features = [
 resolver = "3"
 members = [
   # ── Foundation ──
-  "crates/uffs-polars",          # 🚀 Polars facade (compilation isolation)
-  "crates/uffs-security",        # 🔒 Crypto, key storage, secure FS ops
-  "crates/uffs-text",            # 📝 Unicode text processing, i18n foundation
-  "crates/uffs-time",            # ⏱️ NTFS FILETIME arithmetic (pure, zero deps)
-  "crates/uffs-version",         # 🏷️ Shared --version strings + build-metadata stamp (leaf)
-  "crates/uffs-statusfmt",       # 🎨 Shared operator-status styling (color, glyphs, fields) (leaf)
-  "crates/uffs-broker-protocol", # 📟 Cross-platform broker wire-protocol types (F5)
-  "crates/uffs-winsvc",          # 🪟 Native Windows service control + broker-pipe probe (leaf)
-  "crates/uffs-mft",             # 📦 MFT reading → Polars DataFrame
-  "crates/uffs-format",          # 🧾 Shared CSV formatter (daemon + thin CLI)
-  "crates/uffs-core",            # 🎯 Query engine + compact search engine
+  "crates/uffs-polars",                  # 🚀 Polars facade (compilation isolation)
+  "crates/uffs-security",                # 🔒 Crypto, key storage, secure FS ops
+  "crates/uffs-text",                    # 📝 Unicode text processing, i18n foundation
+  "crates/uffs-time",                    # ⏱️ NTFS FILETIME arithmetic (pure, zero deps)
+  "crates/uffs-version",                 # 🏷️ Shared --version strings + build-metadata stamp (leaf)
+  "crates/uffs-statusfmt",               # 🎨 Shared operator-status styling (color, glyphs, fields) (leaf)
+  "crates/uffs-broker-protocol",         # 📟 Cross-platform broker wire-protocol types (F5)
+  "crates/uffs-content-protocol",        # 📨 Cross-platform Content Service wire-protocol types
+  "crates/uffs-content-reader-protocol", # 📡 Private Coordinator<->Snapshot Reader wire-protocol types
+  "crates/uffs-winsvc",                  # 🪟 Native Windows service control + broker-pipe probe (leaf)
+  "crates/uffs-mft",                     # 📦 MFT reading → Polars DataFrame
+  "crates/uffs-format",                  # 🧾 Shared CSV formatter (daemon + thin CLI)
+  "crates/uffs-core",                    # 🎯 Query engine + compact search engine
   # ── Daemon Architecture ──
-  "crates/uffs-daemon", # 🛡️ Background service process
-  "crates/uffs-client", # 📡 Thin client library
-  "crates/uffs-mcp",    # 🤖 MCP stdio adapter for AI agents
-  "crates/uffs-broker", # 🔑 Windows elevated handle broker (optional)
+  "crates/uffs-daemon",         # 🛡️ Background service process
+  "crates/uffs-client",         # 📡 Thin client library
+  "crates/uffs-mcp",            # 🤖 MCP stdio adapter for AI agents
+  "crates/uffs-broker",         # 🔑 Windows elevated handle broker (optional)
+  "crates/uffs-vss-requestor",  # 🩹 Per-run native VSS snapshot helper, spawned by uffs-broker (optional)
+  "crates/uffs-content",        # 📦 Content Service — VSS-snapshot-scoped file content export (optional)
+  "crates/uffs-content-reader", # 📖 Privileged narrow Snapshot Reader, spawned by uffs-content (optional)
   # ── Surfaces ──
   "crates/uffs-cli",    # 🖥️  Command-line interface
   "crates/uffs-update", # ⬆️  Self-update acquire helper (HTTP/TLS isolated from the CLI)
@@ -149,6 +154,23 @@ uffs-client = { path = "crates/uffs-client", version = "0.6.27" }
 # F5 (issue #205) so neither side duplicates `BROKER_PIPE_NAME` /
 # wire-format byte literals.
 uffs-broker-protocol = { path = "crates/uffs-broker-protocol", version = "0.6.27" }
+# `uffs-content-protocol` carries the wire-protocol types shared between
+# `uffs-content` (the unprivileged content-coordinator producer,
+# Windows-only binary) and any downstream consumer (e.g. Docenta).
+# Pure-logic Layer-0 lib — cross-platform tests run on every CI lane,
+# matching the `uffs-broker-protocol` pattern above. Design references
+# (docs/dev/architecture/, local-only, not tracked): the original
+# `content-stream-tool-design.md` sketch, its
+# `uffs-content-stream-enterprise-design-review.md` replacement-design
+# review, and Docenta's `uffs-ingest-protocol-v2-vss.md`.
+uffs-content-protocol = { path = "crates/uffs-content-protocol", version = "0.6.27" }
+# `uffs-content-reader-protocol` — the private wire format between
+# `uffs-content` (Coordinator) and the privileged Snapshot Reader process
+# (addendum §2.1-§2.4). Deliberately does NOT depend on
+# `uffs-content-protocol` — both are Layer 0, and Layer-0-to-Layer-0
+# internal deps are disallowed (see crate-graph.md); see this crate's
+# Cargo.toml for the full rationale.
+uffs-content-reader-protocol = { path = "crates/uffs-content-reader-protocol", version = "0.6.27" }
 # `uffs-winsvc` — native Windows service control (SCM query/start/stop) +
 # the non-connecting broker-pipe readiness probe.  Layer-0 leaf: its only
 # dependency is the `windows` crate (windows-target), with non-Windows
@@ -203,6 +225,10 @@ windows = { version = "0.62.2", features = [
   "Win32_System_Console",
   "Win32_System_IO",
   "Win32_System_Ioctl",
+  # Job Object kill-on-close (`uffs-broker`'s Snapshot Manager spawns
+  # `uffs-vss-requestor` per lease, assigned to a Job Object so it's
+  # killed if the Broker dies — `broker/snapshot_manager/vss_helper.rs`).
+  "Win32_System_JobObjects",
   "Win32_System_Memory",
   "Win32_System_Pipes",
   "Win32_System_ProcessStatus",
@@ -305,6 +331,11 @@ rustc-hash = "2.1.3"
 itoa = "1.0.18"
 sha2 = "0.11.0"
 hex = "0.4.3"
+# BLAKE3 is the content-integrity digest mandated by the UFFS Content-Ingest
+# protocol (uffs-content-protocol's manifest/frame checksums and the
+# FILE_END content digest) — chosen there for speed over sha2, which stays
+# in place for its existing unrelated consumers.
+blake3 = "1.8.5"
 
 # ───── Network (self-update acquire helper only) ─────
 # Blocking HTTP with rustls + the system trust store. `rustls-tls-native-roots`

crates/uffs-broker-protocol/src/lib.rs

@@ -48,6 +48,8 @@
 
 use thiserror::Error;
 
+pub mod snapshot_manager;
+
 /// Named-pipe path the broker listens on.
 ///
 /// Both the broker server (`uffs-broker::broker`) and the daemon client