bikoret/vav_pipeline/lib/vav_pipeline/doorpost.ex

defmodule VavPipeline.Doorpost do
  @moduledoc """
  Category-grade specification identity verification — the doorpost
  text identity check.

  Vehaya — Devarim 11:20: וּכְתַבְתָּם עַל מְזוּזוֹת בֵּיתֶךָ
  וּבִשְׁעָרֶיךָ. The doorpost-text verse appears near-identically
  in the Shema's first paragraph (Devarim 6:9, singular pole) and
  V'haya's second paragraph (Devarim 11:20, oscillating pole). The
  two are identical in pronunciation and meaning; the Masoretic
  text preserves a single-letter spelling difference (*mezuzot*
  written **מזוזת** chasser at 6:9 vs. **מזוזות** male at 11:20).
  See specs/threshold_inscription.md for why the precise reading
  matters — the system's own gilui_arayot halt rule defends ketiv
  preservation and would treat a male/chasser collapse as a
  category-2 boundary violation. What is constant across paragraphs
  is the FUNCTION (write the corporate text at every threshold),
  not the consonantal text strictly.

  This module is the companion to `L1Reader.verify_integrity/0`.
  L1Reader defends the Torah scroll's identity (the SOURCE text).
  This module defends the *am*'s covenant-content identity (the
  CATEGORY-GRADE specifications the system was built to hold).

  WHY THIS IS LOAD-BEARING:

  Devarim 4:2 — לֹא תֹסִפוּ עַל־הַדָּבָר אֲשֶׁר אָנֹכִי מְצַוֶּה
  אֶתְכֶם וְלֹא תִגְרְעוּ מִמֶּנּוּ. The injunction is plural
  (etkhem) — corporate. Lo tosif is a property the CATEGORY must
  hold, not just a discipline each individual practices. Without
  an enforcement mechanism, lo tosif is moral exhortation, not
  structural commitment — different instances could silently drift,
  running different halt rules, different compiler rules, or
  different disclosure text, and no one would notice.

  Devarim 11:16-17 — the failure mechanism operates whether or not
  anyone notices. Corporate breach (specifications drift) produces
  corporate consequence (loss of what the category was built to
  carry). The detection mechanism — verifying threshold-text
  identity against category-source — is what catches drift before
  the consequence accumulates.

  Devarim 13:1 — the keeping is plural (corporate); the actual
  not-adding/not-subtracting happens at every individual threshold
  (singular `lo tosef`, `lo tigra`). Each threshold COULD add or
  subtract; only the corporate detection mechanism catches it.

  Vehaya §3.3 — "the same specifications must be inscribed at every
  threshold. Drift is breach. Implement a verification mechanism
  for category-text identity."

  THE MECHANISM:

  1. The category-grade specification files are listed in
     `@spec_files`. These are the DECLARATIVE artifacts — halt rules,
     compiler rules (13 middot of R. Yishmael), foundational reading
     docs, the architectural specs, and the BUILD canon. Operations
     files (pekudei, terumah) are excluded because they are
     execution-grade rather than category-grade orientation.

  2. `compute_identity/0` produces a stable SHA-256 over the
     canonical concatenation of per-file hashes (sorted by
     repo-relative path). The hash is independent of filesystem
     listing order and tolerant of single-file drift detection
     (the manifest exposes per-file hashes for diagnosis).

  3. `@expected_hash` is BAKED IN. When specs change, the operator
     must update both the spec file AND this constant. The constant
     is the witness; the file is what is witnessed.

  4. `verify_identity/0` returns :ok or {:error, reason} — same
     shape as L1Reader.verify_integrity/0. The HaltMonitor is the
     natural caller (so spec drift halts the system the same way
     L1 corruption halts it); commit-9-style integration to follow.

  5. `manifest/0` returns the per-file hash so operators can
     diagnose WHICH file drifted when verify fails. Operator-side
     queryable via /kodesh/specification_identity (see
     bigdei_kodesh/router.ex).

  WHAT THIS DOES NOT CLAIM:

  - That the specifications are correct (only that they are
    identifiable and unchanged from the witnessed state).
  - That all instances will run the same code (only that all
    instances running this code witness the same specifications).
  - That hash equality across deployments proves no drift outside
    this spec set (it does NOT — other artifacts not listed here
    are not covered).

  WHAT TO DO ON DRIFT:

  If verify_identity returns {:error, ...}, the operator must:
    1. Run `manifest/0` to identify which file's hash mismatches.
    2. Inspect the diff against git to determine whether the change
       was deliberate (an authorized update) or unauthorized
       (a drift incident).
    3. If deliberate: update @expected_hash in this file, commit
       both changes together with a message naming the spec change.
    4. If unauthorized: revert the file, investigate how the change
       was introduced, fire chatzotzrot :assembly_leaders.

  The discipline: every change to @expected_hash should be traceable
  in git to a corresponding spec change in the SAME COMMIT. A bare
  hash-update commit with no spec-file change is itself a drift
  signal — the operator updated the witness without updating what
  is witnessed, which is exactly the failure mode this module
  defends against.
  """

  # Repo-relative paths to the category-grade specification files.
  # Sorted alphabetically so the canonical hash is independent of
  # filesystem listing order.
  #
  # WHAT'S INCLUDED:
  #   - Halt rules (the cardinal-sin specifications — yehareg
  #     v'al ya'avor)
  #   - Compiler rules (the 13 middot of R. Yishmael — the
  #     inference protocol)
  #   - Foundational readings (eish_zarah, structure_not_content,
  #     lo_taaseh_lekha)
  #   - Architectural specs (PRD, architecture, vayakhel /
  #     construction, tamid / operations including the EXTERNAL
  #     NEDARIM section)
  #   - Build canon (BUILD_KORBAN, BUILD_AZAZEL)
  #
  # WHAT'S EXCLUDED:
  #   - bikoret_pekudei (economic model — operations)
  #   - bikoret_terumah (materials — operations)
  #   - Implementation code (execution-grade; the spec is what
  #     the execution implements)
  #   - Test files (these verify the spec; they are not the spec)
  @spec_files [
    "BUILD_AZAZEL.md",
    "BUILD_KORBAN.md",
    # Pass 7 — compiler rules (operational hermeneutics derived from
    # Torah, sibling to talmudic_engine/src/middot/) brought in per
    # user's instruction with verbatim preservation and accurate
    # source labeling. Devarim 17:11 — al pi haTorah asher yorukha.
    "specs/bikoret_architecture.md",
    "specs/bikoret_prd.md",
    "specs/bikoret_tamid.md",
    "specs/bikoret_vayakhel.md",
    "specs/bishearecha.md",
    "specs/compiler_rules.md",
    "specs/eish_zarah_reading.md",
    "specs/lo_taaseh_lekha.md",
    # Vehaya §6 items 4, 5, 8 — added by the closure pass:
    "specs/lo_tosif.md",
    "specs/return_specification.md",
    "specs/structure_not_content.md",
    "specs/threshold_inscription.md",
    "talmudic_engine/src/halt/anchors.pl",
    "talmudic_engine/src/halt/avodah_zarah.pl",
    "talmudic_engine/src/halt/gilui_arayot.pl",
    "talmudic_engine/src/halt/lo_taaseh_lekha.pl",
    "talmudic_engine/src/halt/mishkal.pl",
    "talmudic_engine/src/halt/shfichut_damim.pl",
    "talmudic_engine/src/halt/structure.pl",
    "talmudic_engine/src/middot/binyan_av.pl",
    "talmudic_engine/src/middot/davar_yatza.pl",
    "talmudic_engine/src/middot/gezerah_shavah.pl",
    "talmudic_engine/src/middot/kal_vachomer.pl",
    "talmudic_engine/src/middot/klal_ufrat.pl",
    "talmudic_engine/src/middot/lamed_inyan.pl",
    "talmudic_engine/src/middot/middot.pl",
    "talmudic_engine/src/middot/shnei_ketuvim.pl"
  ]

  # Repo root — directory containing the spec files. Configurable
  # via :vav_pipeline application env so tests can point at
  # fixtures.
  @repo_root Application.compile_env(
               :vav_pipeline,
               :doorpost_repo_root,
               Path.expand("../../..", __DIR__)
             )

  # The category-grade-identity hash. Updated when any spec file
  # changes; the update is a deliberate operator commit. A bare
  # hash-update commit (without a corresponding spec change in the
  # same commit) is itself a drift signal.
  @expected_hash "c963c4fe5076ec713ae12f07618e2606788b1a6b706a426800fab6a0fdd1bcfa"

  @doc "Return the list of files that constitute the category-grade specification."
  @spec files() :: [String.t()]
  def files, do: @spec_files

  @doc "Return the baked-in expected hash."
  @spec expected_hash() :: String.t()
  def expected_hash, do: @expected_hash

  @doc "Return the configured repo root."
  @spec repo_root() :: String.t()
  def repo_root, do: @repo_root

  @doc """
  Compute the canonical hash over the current spec files.

  The hash combines per-file SHA-256s in canonical (sorted by path)
  order. If any spec file is missing from the filesystem, returns
  {:error, "missing spec file: <path>"} — the spec set is
  incomplete by definition; no partial hash is meaningful.
  """
  @spec compute_identity() :: String.t() | {:error, String.t()}
  def compute_identity do
    items = manifest_items()

    case Enum.find(items, fn {_, h} -> h == :missing end) do
      nil ->
        combined =
          items
          |> Enum.map(fn {path, hash} -> "#{path}\t#{hash}" end)
          |> Enum.join("\n")

        :crypto.hash(:sha256, combined) |> Base.encode16(case: :lower)

      {missing_path, _} ->
        {:error, "missing spec file: #{missing_path}"}
    end
  end

  @doc """
  Return the manifest as a list of %{path, hash} maps. When a file
  is missing, hash is the atom :missing. Used by the
  /kodesh/specification_identity endpoint and by operators
  diagnosing drift.
  """
  @spec manifest() :: [%{path: String.t(), hash: String.t() | :missing}]
  def manifest do
    Enum.map(manifest_items(), fn {path, hash} ->
      %{path: path, hash: hash}
    end)
  end

  defp manifest_items do
    Enum.map(@spec_files, fn path ->
      full_path = Path.join(@repo_root, path)

      case File.read(full_path) do
        {:ok, contents} ->
          {path, :crypto.hash(:sha256, contents) |> Base.encode16(case: :lower)}

        {:error, _} ->
          {path, :missing}
      end
    end)
  end

  @doc """
  Verify the current spec set matches the expected hash.

  Returns :ok when the hashes match. Returns {:error, reason} when
  any spec file is missing OR the computed hash does not match the
  baked-in @expected_hash. Same shape as L1Reader.verify_integrity/0.

  Lo tosif (Devarim 4:2) — addition or subtraction is detected
  here, BEFORE it propagates into reception. When the hash matches,
  the halt rules' anchors are verified too (`VavPipeline.Anchors`):
  a rule without a verse, or a verse not in L1, fails the same way.
  """
  @spec verify_identity() :: :ok | {:error, String.t()}
  def verify_identity do
    case compute_identity() do
      {:error, reason} ->
        {:error, "SPEC_INTEGRITY_FAILED: #{reason}"}

      hash when is_binary(hash) ->
        if hash == @expected_hash do
          # The rules grounded as the outputs are (bishearecha.md §X):
          # every halt rule anchored to a verse, every verse in L1.
          VavPipeline.Anchors.verify(@repo_root)
        else
          {:error,
           "SPEC_INTEGRITY_FAILED: hash mismatch — computed #{hash}, " <>
             "expected #{@expected_hash}. Run " <>
             "VavPipeline.Doorpost.manifest/0 to identify which file " <>
             "drifted; inspect git for whether the change was deliberate."}
        end
    end
  end
end