tallwright-core 0.1.0__py3-none-any.whl

tallwright_core/__init__.py

@@ -0,0 +1,93 @@
+# SPDX-License-Identifier: Apache-2.0
+"""tallwright_core — shared record schema, content-addressing, and Merkle primitives.
+
+The dependency-light foundation every other Tallwright package builds on. No
+networked or external-service code lives here (no TSA, transparency log, or storage).
+"""
+
+from __future__ import annotations
+
+from .document import AmbiguousDocument, DocumentTooDeep, load_document
+from .write import write_document
+from .canonical import (
+    MAX_CANONICAL_DEPTH,
+    MAX_CANONICAL_NODES,
+    CanonicalizationError,
+    canonicalize,
+)
+from .hashing import (
+    content_address,
+    decode_content_address,
+    decode_multihash,
+    multihash,
+    MalformedSalt,
+    SALT_LEN,
+    address_from_digest,
+    opens_address,
+    sha256,
+    verify,
+)
+from .merkle import (
+    MerkleTree,
+    consistency_proof,
+    inclusion_proof,
+    leaf_hash,
+    merkle_root,
+    node_hash,
+    verify_consistency,
+    verify_inclusion,
+)
+from .schema import (
+    CANONICALIZATION_METHOD,
+    HASH_ALGORITHM,
+    SCHEMA_VERSION,
+    AnchorReceipt,
+    EventType,
+    GapManifest,
+    Record,
+    link,
+    verify_chain,
+)
+
+__all__ = [
+    # reading a document, and replacing one
+    "load_document",
+    "AmbiguousDocument",
+    "DocumentTooDeep",
+    "write_document",
+    # canonicalization
+    "canonicalize",
+    "CanonicalizationError",
+    "MAX_CANONICAL_DEPTH",
+    "MAX_CANONICAL_NODES",
+    # hashing
+    "sha256",
+    "multihash",
+    "decode_multihash",
+    "content_address",
+    "MalformedSalt",
+    "SALT_LEN",
+    "address_from_digest",
+    "opens_address",
+    "decode_content_address",
+    "verify",
+    # merkle
+    "leaf_hash",
+    "node_hash",
+    "merkle_root",
+    "inclusion_proof",
+    "consistency_proof",
+    "verify_inclusion",
+    "verify_consistency",
+    "MerkleTree",
+    # schema
+    "EventType",
+    "GapManifest",
+    "AnchorReceipt",
+    "Record",
+    "link",
+    "verify_chain",
+    "SCHEMA_VERSION",
+    "HASH_ALGORITHM",
+    "CANONICALIZATION_METHOD",
+]

tallwright_core/canonical.py

@@ -0,0 +1,88 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Hardened RFC 8785 (JCS) canonicalization.
+
+Wraps ``rfc8785.dumps`` with fail-closed guards so a hostile or malformed input becomes
+a *typed, catchable* error instead of a crash, a silent drop, or non-deterministic bytes:
+
+  - **non-finite floats** (NaN / +-Infinity) are rejected — RFC 8785 has no representation
+    for them, and letting one through would make a hash irreproducible;
+  - **nesting depth is bounded** (``MAX_CANONICAL_DEPTH``) so a deeply-nested structure
+    cannot exhaust the interpreter stack (a cheap capture-path DoS);
+  - **total work is bounded** (``MAX_CANONICAL_NODES``) so a shallow but shared-reference
+    DAG cannot expand to exponentially many nodes and hang the serializer;
+  - any residual library-level failure (unserializable type, non-string key, recursion)
+    is re-raised as :class:`CanonicalizationError`.
+
+For every input **within the depth / total-work bounds**, the output is byte-for-byte
+identical to a raw ``rfc8785.dumps`` call — so record hashes are unchanged. Inputs that
+exceed those bounds (or that ``rfc8785`` itself rejects) raise a TYPED error rather than hanging or
+silently producing a non-reproducible value. What the capture layer then does depends on its mode:
+under the SHIPPED DEFAULT (``fail_closed=False``) the error is counted and sealed as an auditable
+rupture marker, and the agent keeps serving; with ``fail_closed=True`` it propagates and NOTHING is
+sealed — no gap record, no rupture marker, and no unrecorded action either.
+The bounds are set far above any real record or tool payload, so this only ever trips a
+pathological/hostile input.
+"""
+
+from __future__ import annotations
+
+import math
+from collections.abc import Mapping, Sequence
+from typing import Any
+
+import rfc8785
+
+# JSON nesting deeper than this is rejected before serialization. Real records and tool
+# payloads are shallow; this only trips a pathological/hostile input.
+MAX_CANONICAL_DEPTH = 64
+# Upper bound on the total number of values visited during the guard walk. Bounds total
+# work so a shared-reference DAG (shallow but exponentially wide when traversed as a tree)
+# is rejected quickly instead of hanging. JSON from ``model_dump`` is a tree (node count ~=
+# payload size), so this only rejects a genuinely enormous or hostile-DAG payload — which
+# then becomes an auditable gap, never an unbounded hang.
+MAX_CANONICAL_NODES = 1_000_000
+
+
+class CanonicalizationError(ValueError):
+    """An input cannot be deterministically canonicalized under RFC 8785 (JCS)."""
+
+
+def canonicalize(obj: Any) -> bytes:
+    """Return the RFC 8785 (JCS) canonical bytes of ``obj``, or raise CanonicalizationError.
+
+    Guards run *before* serialization so depth/finiteness/total-work violations are reported
+    without ever entering the C serializer with a hostile input. For any already-canonicalizable
+    input the output is byte-identical to a raw ``rfc8785.dumps`` call.
+    """
+    remaining = MAX_CANONICAL_NODES
+
+    def guard(node: Any, depth: int) -> None:
+        nonlocal remaining
+        remaining -= 1
+        if remaining < 0:
+            raise CanonicalizationError(f"input exceeds MAX_CANONICAL_NODES={MAX_CANONICAL_NODES}")
+        if depth > MAX_CANONICAL_DEPTH:
+            raise CanonicalizationError(f"nesting exceeds MAX_CANONICAL_DEPTH={MAX_CANONICAL_DEPTH}")
+        if isinstance(node, bool):
+            return  # bool before int/float; nothing to recurse into
+        if isinstance(node, float):
+            if not math.isfinite(node):
+                raise CanonicalizationError(f"non-finite float {node!r} has no RFC 8785 representation")
+            return
+        if isinstance(node, Mapping):
+            for value in node.values():
+                guard(value, depth + 1)
+            return
+        if isinstance(node, (str, bytes)):
+            return  # sized but not a container to descend
+        if isinstance(node, Sequence):
+            for item in node:
+                guard(item, depth + 1)
+
+    guard(obj, 0)
+    try:
+        return rfc8785.dumps(obj)
+    except (ValueError, TypeError, RecursionError, OverflowError) as exc:
+        # rfc8785's own IntegerDomainError / etc. subclass ValueError; normalize them all
+        # (and any residual serializer failure) to one typed, catchable error.
+        raise CanonicalizationError(str(exc)) from exc

tallwright_core/document.py

@@ -0,0 +1,105 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Reading a document whose meaning does not depend on who reads it.
+
+JSON permits an object to carry the same key twice, and every parser resolves that its own way:
+``json.loads`` keeps the LAST, a person reading the file sees the FIRST, and another tool may see
+either. A file with two readings is not a document — and an exhibit is a document before it is
+anything else.
+
+What that buys a producer, measured against this project's own golden vector before this module
+existed: insert ``"action_type":"password_reset",`` in front of the real ``"action_type":"wire_transfer"``
+and the parsed object is unchanged, so the canonicalization is unchanged, so **every signature still
+verifies**. The court-facing tool prints its PASS lines for a file that reads as a different action
+than the one it authenticates. No key is forged; the ambiguity is in the format.
+
+So it is refused, at the door, wherever this project reads a document it will act on.
+
+**Nor is a document read past a depth nobody writes.** ``json.loads`` recurses once per level of
+nesting, and two hundred thousand ``[`` made it raise RecursionError — which is not a ValueError, so it
+passed every reader's "unreadable" branch: the verifier printed a traceback and exited 1, the code for
+evidence that was tampered with, and the same program in a browser lost its WebAssembly worker, whose
+stack runs out before Python's recursion check does. The depth is therefore counted BEFORE anything
+recurses, and a document past :data:`MAX_DOCUMENT_DEPTH` is refused as the ValueError it always should
+have been.
+"""
+
+from __future__ import annotations
+
+import json
+import re
+from typing import Any
+
+from .canonical import MAX_CANONICAL_DEPTH
+
+__all__ = ["AmbiguousDocument", "DocumentTooDeep", "MAX_DOCUMENT_DEPTH", "load_document"]
+
+#: The deepest a document read here may nest, its own outermost object or array counting as one. An
+#: exhibit nests 6 levels, a policy 4 and a disclosure 1; the deepest object this project canonicalises
+#: nests MAX_CANONICAL_DEPTH + 1, and this leaves room for it inside an envelope as deep again. CPython
+#: 3.12 raises RecursionError near 3,000 levels and a browser's WebAssembly stack gives out sooner, so
+#: the bound sits far below both, and far above anything real.
+MAX_DOCUMENT_DEPTH = 2 * MAX_CANONICAL_DEPTH
+
+#: One JSON string — closed, or running to the end of a text that never closes it — or one bracket. The
+#: possessive quantifiers consume a string's run once and never give it back, so no input makes the scan
+#: restart inside a string it has passed: the scan is linear in the text.
+_TOKENS = re.compile(r'"(?:[^"\\]++|\\.)*+(?:"|\\?\Z)|[\[\]{}]', re.DOTALL)
+
+
+class AmbiguousDocument(ValueError):
+    """A document that reads one way to one parser and another way to the next."""
+
+
+class DocumentTooDeep(ValueError):
+    """A document nested deeper than anything this project writes, refused before it is parsed."""
+
+
+def _no_repeats(pairs: list[tuple[str, Any]]) -> dict[str, Any]:
+    seen: dict[str, Any] = {}
+    for key, value in pairs:
+        if key in seen:
+            raise AmbiguousDocument(
+                f"the key {key!r} appears twice in one object. A reader of this file and a parser "
+                f"of it would not agree on what it says, so it is not read here."
+            )
+        seen[key] = value
+    return seen
+
+
+def _deeper_than(text: str, bound: int) -> bool:
+    """Whether any point of ``text`` lies inside more than ``bound`` objects and arrays.
+
+    Brackets inside strings are not nesting, and an escaped quote does not end a string. Malformed text
+    is not this function's to judge: it counts what it finds, and the parser refuses the rest.
+    """
+    if text.count("[") + text.count("{") <= bound:
+        return False  # a text cannot nest deeper than it has openings; nearly every document stops here
+    depth = 0
+    for token in _TOKENS.finditer(text):
+        first = text[token.start()]
+        if first in "[{":
+            depth += 1
+            if depth > bound:
+                return True
+        elif first in "]}":
+            depth -= 1
+    return False
+
+
+def load_document(text: str | bytes) -> Any:
+    """``json.loads``, refusing any object that carries a key twice and any document nested deeper than
+    :data:`MAX_DOCUMENT_DEPTH`.
+
+    Raises :class:`AmbiguousDocument` for a repeated key, :class:`DocumentTooDeep` for the nesting, and
+    ``json.JSONDecodeError`` — a ``ValueError`` like both — for anything else malformed, so a caller that
+    already handles unreadable input keeps handling it. Bytes are decoded the way ``json.loads`` decodes
+    them.
+    """
+    source = text.decode(json.detect_encoding(text), "surrogatepass") if isinstance(text, bytes) else text
+    if _deeper_than(source, MAX_DOCUMENT_DEPTH):
+        raise DocumentTooDeep(
+            f"it is nested deeper than {MAX_DOCUMENT_DEPTH} levels. Nothing this project writes comes "
+            f"near that, and a parser following it would run out of stack before it ran out of "
+            f"document, so it is not read here."
+        )
+    return json.loads(source, object_pairs_hook=_no_repeats)

tallwright_core/hashing.py

@@ -0,0 +1,207 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Content-addressing and self-describing multihash (SHA-256).
+
+Zero external dependencies. SHA-256 is carried as a self-describing *multihash*
+so a digest records *which* algorithm produced it — the algorithm-agility
+property the seal spec needs for long-term verifiability. Only sha2-256 is supported in V1;
+the structure leaves room for a future migration (e.g. to an SHA-3 / longer hash)
+without breaking stored proofs.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import hmac
+
+# multicodec code for sha2-256 is 0x12; its digest length is 32 bytes (0x20).
+SHA2_256_CODE = 0x12
+SHA2_256_LENGTH = 32
+# multibase prefix for base16, lower-case (RFC 4648). Keeps content addresses
+# trivially inspectable and verifiable.
+_MULTIBASE_BASE16_LOWER = "f"
+
+
+def sha256(data: bytes) -> bytes:
+    """Return the raw 32-byte SHA-256 digest of ``data``."""
+    return hashlib.sha256(data).digest()
+
+
+def _uvarint_encode(value: int) -> bytes:
+    if value < 0:
+        raise ValueError("uvarint cannot encode a negative value")
+    out = bytearray()
+    while True:
+        byte = value & 0x7F
+        value >>= 7
+        if value:
+            out.append(byte | 0x80)
+        else:
+            out.append(byte)
+            return bytes(out)
+
+
+def _uvarint_decode(buf: bytes, offset: int) -> tuple[int, int]:
+    result = 0
+    shift = 0
+    pos = offset
+    while True:
+        if pos >= len(buf):
+            raise ValueError("truncated uvarint")
+        byte = buf[pos]
+        pos += 1
+        result |= (byte & 0x7F) << shift
+        if not byte & 0x80:
+            return result, pos
+        shift += 7
+        if shift > 63:
+            raise ValueError("uvarint too long")
+
+
+def multihash(data: bytes, *, algo: str = "sha2-256") -> bytes:
+    """Return the self-describing multihash of ``data``.
+
+    Layout: ``uvarint(code) || uvarint(length) || digest``.
+    """
+    if algo != "sha2-256":
+        raise ValueError(f"unsupported hash algorithm: {algo!r}")
+    digest = sha256(data)
+    return _uvarint_encode(SHA2_256_CODE) + _uvarint_encode(len(digest)) + digest
+
+
+def decode_multihash(mh: bytes) -> tuple[int, int, bytes]:
+    """Return ``(code, length, digest)`` from a multihash, validating length."""
+    code, pos = _uvarint_decode(mh, 0)
+    length, pos = _uvarint_decode(mh, pos)
+    digest = mh[pos:]
+    if len(digest) != length:
+        raise ValueError("multihash length does not match its declared length")
+    return code, length, digest
+
+
+def content_address(data: bytes) -> str:
+    """Return a multibase (base16-lower) content address for ``data``.
+
+    Used for ``payload_hash`` and for the record chain address. NOT for ``record_id``, which the
+    recorder mints as a UUID4 and the reference pipeline sets to a human label.
+    """
+    return _MULTIBASE_BASE16_LOWER + multihash(data).hex()
+
+
+def address_from_digest(digest: bytes) -> str:
+    """The same content address as :func:`content_address`, from a digest already computed.
+
+    For data that arrives in pieces and must not be held whole. A streamed response body is the
+    case this exists for: the transport adapter folds each chunk into a ``hashlib`` object and keeps
+    only the prefix it is allowed to seal, so a multi-gigabyte download costs the recorder a
+    constant amount of memory instead of twice the body.
+
+    The knowledge of what an address LOOKS like stays in this module, which is the point. A caller
+    assembling the multibase prefix and the multicodec header itself would be a second place that
+    knows the format, and a second place is a place that one day answers differently.
+    """
+    if len(digest) != SHA2_256_LENGTH:
+        raise ValueError(
+            f"expected a {SHA2_256_LENGTH}-byte sha2-256 digest, got {len(digest)} bytes"
+        )
+    header = _uvarint_encode(SHA2_256_CODE) + _uvarint_encode(SHA2_256_LENGTH)
+    return _MULTIBASE_BASE16_LOWER + (header + digest).hex()
+
+
+#: How long a per-record salt is, everywhere, without exception.
+#:
+#: It is not a tuning knob. ``SHA-256(salt ‖ plaintext)`` commits to the CONCATENATION and not to
+#: the SPLIT, so if the length were free, every ``k`` would give a second valid reading:
+#: ``(salt ‖ plaintext[:k], plaintext[k:])`` opens the very same address. An operator could then
+#: disclose a TRUNCATED payload — pushing the amount, the account, the instruction into what they
+#: call the salt — and a court's own verifier would confirm it opens. Because RFC 8785 sorts the
+#: keys, the choppable prefix is deterministic and starts with the fields most worth withholding.
+#:
+#: A fixed length removes the second reading: there is exactly one way to cut 32 bytes off the front.
+#: Found 2026-08-25 by adversarial review, demonstrated against a real record before it was fixed.
+SALT_LEN = 32
+
+
+class MalformedSalt(ValueError):
+    """A salt that is not exactly :data:`SALT_LEN` bytes. Not a mismatch — an unreadable disclosure.
+
+    Kept distinct because the two say completely different things to a court. "This disclosure does
+    not open the address" accuses the operator of giving a false account of their own record. "This
+    file is not in the format" says nothing about the record at all, and rendering the second as the
+    first is exactly the kind of overstatement this codebase refuses everywhere else.
+    """
+
+
+def opens_address(address: str, salt: bytes, plaintext: bytes) -> bool:
+    """Does this salt and this plaintext produce that committed address?
+
+    The whole check a court performs when the archive holds no content — the operator discloses the
+    two inputs, anyone recomputes, and no state belonging to the recorder is involved at any point.
+    That is why it keeps working after the recorder is gone, and why it lives in the spine
+    rather than in the service.
+
+    The salt is why an address is not a search key. Without it, a short payload ("Termin abgesagt")
+    could be recovered from its hash by trying candidates; with it, the operator holds the one thing
+    that opens the commitment. Lose the salt and the record can never be opened again — the chain
+    still proves it existed unaltered at that position, and nothing proves what it said.
+
+    Raises :class:`MalformedSalt` if the salt is not exactly :data:`SALT_LEN` bytes. That check is
+    what makes the split unique and therefore what makes this a commitment to the PAIR rather than
+    to their concatenation — see the note on :data:`SALT_LEN`.
+    """
+    if len(salt) != SALT_LEN:
+        raise MalformedSalt(
+            f"a salt is exactly {SALT_LEN} bytes; this one is {len(salt)}. Without a fixed length "
+            "the same address opens under many different splits, so a truncated payload would "
+            "verify as the whole one."
+        )
+    return content_address(salt + plaintext) == address
+
+
+def is_content_address(value: str) -> bool:
+    """Is this string an address this module could have produced?
+
+    Three places asked `value.startswith("f")` instead, so `'f'` was a well-formed
+    address, and so were `'frei'`, `'f' + 'z' * 68`, a truncated digest and an
+    upper-case one. An archive reporting OK over a record whose payload_hash commits
+    to nothing leaves a court with nothing to recompute against, which is the whole of
+    the disclosure story when the archive holds no content.
+
+    It lives here for the reason `address_from_digest` gives: a second place that knows
+    the format is a place that one day answers differently. Deliberately NOT a validator
+    on `Record.payload_hash` — the spine and verifier suites build link-only records with
+    stub hashes, and a schema validator would turn an existing bad log into a parse
+    failure instead of a named finding.
+    """
+    if not value.startswith(_MULTIBASE_BASE16_LOWER):
+        return False
+    body = value[1:]
+    # Upper-case hex is a DIFFERENT multibase ('F'), so a body that is not its own
+    # lower-case did not come from here even when it decodes.
+    if len(body) % 2 or body != body.lower():
+        return False
+    try:
+        code, length, digest = decode_multihash(bytes.fromhex(body))
+    except ValueError:
+        return False
+    return code == SHA2_256_CODE and length == SHA2_256_LENGTH and len(digest) == SHA2_256_LENGTH
+
+
+def decode_content_address(addr: str) -> bytes:
+    """Inverse of :func:`content_address`; returns the raw multihash bytes."""
+    if not addr or addr[0] != _MULTIBASE_BASE16_LOWER:
+        raise ValueError("unsupported multibase prefix (expected base16 'f')")
+    return bytes.fromhex(addr[1:])
+
+
+def verify(data: bytes, mh: bytes) -> bool:
+    """Return True iff ``mh`` is the multihash of ``data`` (constant-time)."""
+    try:
+        expected = multihash(data, algo="sha2-256")
+    except ValueError:
+        # FAIL CLOSED. Unreachable while sha2-256 is the only algorithm, but this module
+        # documents an algorithm migration as the point of the multihash envelope, and the
+        # migration is exactly when this branch starts to matter. `return True` here would
+        # make verify() answer True for ANY mh; the mutation lane found nothing pinning the
+        # direction, and `# pragma: no cover` meant coverage never would.
+        return False
+    return hmac.compare_digest(expected, mh)

tallwright_core/merkle.py

@@ -0,0 +1,186 @@
+# SPDX-License-Identifier: Apache-2.0
+"""RFC 6962 Merkle tree: root, inclusion proofs, consistency proofs, verifiers.
+
+Hashing is domain-separated exactly per RFC 6962 (leaf = ``SHA-256(0x00 || entry)``,
+interior = ``SHA-256(0x01 || left || right)``) so proofs are portable and
+interoperable with Certificate Transparency / Trillian-Tessera tooling — which is
+what makes an export verifiable years later with no live service.
+
+The prover (:func:`inclusion_proof`, :func:`consistency_proof`) follows the
+recursive RFC 6962 definitions; the verifiers follow the iterative RFC 9162
+algorithms. They are deliberately separate code paths so their agreement is
+strong evidence of correctness.
+"""
+
+from __future__ import annotations
+
+from .hashing import sha256
+
+LEAF_PREFIX = b"\x00"
+NODE_PREFIX = b"\x01"
+
+
+def leaf_hash(entry: bytes) -> bytes:
+    """RFC 6962 leaf hash of a raw entry."""
+    return sha256(LEAF_PREFIX + entry)
+
+
+def node_hash(left: bytes, right: bytes) -> bytes:
+    """RFC 6962 interior node hash of two child hashes."""
+    return sha256(NODE_PREFIX + left + right)
+
+
+def _largest_power_of_two_less_than(n: int) -> int:
+    # Largest k that is a power of two with k < n (requires n >= 2).
+    k = 1
+    while (k << 1) < n:
+        k <<= 1
+    return k
+
+
+def merkle_root(entries: list[bytes]) -> bytes:
+    """Return the RFC 6962 Merkle Tree Hash (MTH) over the ordered entries.
+
+    The hash of an empty list is the hash of the empty string.
+    """
+    if not entries:
+        return sha256(b"")
+    if len(entries) == 1:
+        return leaf_hash(entries[0])
+    k = _largest_power_of_two_less_than(len(entries))
+    return node_hash(merkle_root(entries[:k]), merkle_root(entries[k:]))
+
+
+def inclusion_proof(entries: list[bytes], index: int) -> list[bytes]:
+    """Return the audit PATH for the entry at ``index`` (RFC 6962)."""
+    n = len(entries)
+    if not 0 <= index < n:
+        raise IndexError("leaf index out of range")
+    if n == 1:
+        return []
+    k = _largest_power_of_two_less_than(n)
+    if index < k:
+        return inclusion_proof(entries[:k], index) + [merkle_root(entries[k:])]
+    return inclusion_proof(entries[k:], index - k) + [merkle_root(entries[:k])]
+
+
+def _subproof(m: int, entries: list[bytes], at_root: bool) -> list[bytes]:
+    n = len(entries)
+    if m == n:
+        return [] if at_root else [merkle_root(entries)]
+    k = _largest_power_of_two_less_than(n)
+    if m <= k:
+        return _subproof(m, entries[:k], at_root) + [merkle_root(entries[k:])]
+    return _subproof(m - k, entries[k:], False) + [merkle_root(entries[:k])]
+
+
+def consistency_proof(entries: list[bytes], first_size: int) -> list[bytes]:
+    """Return the consistency PROOF between the tree of ``first_size`` and the full tree."""
+    n = len(entries)
+    if not 0 <= first_size <= n:
+        raise ValueError("first_size out of range")
+    if first_size == 0 or first_size == n:
+        return []
+    return _subproof(first_size, entries, True)
+
+
+def verify_inclusion(
+    entry: bytes, index: int, tree_size: int, root: bytes, proof: list[bytes]
+) -> bool:
+    """Verify an inclusion proof for a raw ``entry`` (RFC 9162 algorithm)."""
+    # BOTH bounds, and a positive tree_size. A NEGATIVE index must be rejected: Python's saturating
+    # >> and & on negatives pin `fn` to a fixed parity, so every leaf would otherwise have an infinite
+    # family of negative aliases that verify against the same signed root — a forged, producer-
+    # controlled leaf_index reading as VALID. (The prover, inclusion_proof, always checked both
+    # bounds; only the verifier did not. Found in review.)
+    if not 0 <= index < tree_size:
+        return False
+    fn, sn = index, tree_size - 1
+    r = leaf_hash(entry)
+    for p in proof:
+        if sn == 0:
+            return False
+        if fn & 1 or fn == sn:
+            r = node_hash(p, r)
+            if not fn & 1:
+                while fn and not fn & 1:
+                    fn >>= 1
+                    sn >>= 1
+        else:
+            r = node_hash(r, p)
+        fn >>= 1
+        sn >>= 1
+    return sn == 0 and r == root
+
+
+def verify_consistency(
+    first_size: int,
+    second_size: int,
+    first_root: bytes,
+    second_root: bytes,
+    proof: list[bytes],
+) -> bool:
+    """Verify a consistency proof between two tree sizes (RFC 9162 algorithm)."""
+    # Reject negative sizes for the same reason as verify_inclusion: a negative operand would drive
+    # the shift/parity loop below into meaningless state rather than failing closed.
+    if first_size < 0 or second_size < 0 or first_size > second_size:
+        return False
+    if first_size == second_size:
+        return len(proof) == 0 and first_root == second_root
+    if first_size == 0:
+        return len(proof) == 0
+    fn, sn = first_size - 1, second_size - 1
+    while fn & 1:
+        fn >>= 1
+        sn >>= 1
+    it = iter(proof)
+    if fn == 0:
+        fr = sr = first_root
+    else:
+        try:
+            fr = sr = next(it)
+        except StopIteration:
+            return False
+    for c in it:
+        if sn == 0:
+            return False
+        if fn & 1 or fn == sn:
+            fr = node_hash(c, fr)
+            sr = node_hash(c, sr)
+            if not fn & 1:
+                while fn and not fn & 1:
+                    fn >>= 1
+                    sn >>= 1
+        else:
+            sr = node_hash(sr, c)
+        fn >>= 1
+        sn >>= 1
+    return sn == 0 and fr == first_root and sr == second_root
+
+
+class MerkleTree:
+    """A small append-only RFC 6962 tree over in-memory entries."""
+
+    def __init__(self, entries: list[bytes] | None = None) -> None:
+        self._entries: list[bytes] = list(entries) if entries else []
+
+    def add(self, entry: bytes) -> int:
+        """Append a raw entry; return the new tree size."""
+        self._entries.append(entry)
+        return len(self._entries)
+
+    @property
+    def size(self) -> int:
+        return len(self._entries)
+
+    def entry(self, index: int) -> bytes:
+        return self._entries[index]
+
+    def root(self) -> bytes:
+        return merkle_root(self._entries)
+
+    def inclusion_proof(self, index: int) -> list[bytes]:
+        return inclusion_proof(self._entries, index)
+
+    def consistency_proof(self, first_size: int) -> list[bytes]:
+        return consistency_proof(self._entries, first_size)

tallwright_core/schema.py

@@ -0,0 +1,176 @@
+# SPDX-License-Identifier: Apache-2.0
+"""The court-defensible record envelope, canonicalization, and hash-chain.
+
+Realizes the 14-field record schema. Design rule:
+the *record* holds hashes + metadata; the full payload (the context value, or the
+action parameters) is sealed **off-record** and referenced by ``payload_hash``.
+
+Canonicalization is RFC 8785 (JCS) so the record hash is reproducible by a court
+expert years later. ``anchor_receipt`` and ``certification_pointer`` are attached
+*after* sealing (anchoring is batched; certification is added later), so they are
+EXCLUDED from the hashed and hash-chained content — otherwise a record could not
+be anchored after its hash was computed.
+
+Framing discipline (non-negotiable): we record the context received and the action
+triggered — never the agent's reasoning or intent. Integrity is from the moment of
+capture, not the truth of the inputs.
+"""
+
+from __future__ import annotations
+
+from enum import Enum
+from typing import Any
+
+from pydantic import BaseModel, ConfigDict, Field, model_validator
+
+from .canonical import canonicalize
+from .hashing import content_address, multihash
+
+SCHEMA_VERSION = "0.1.0"
+HASH_ALGORITHM = "sha2-256"
+CANONICALIZATION_METHOD = "RFC8785-JCS"
+
+# Attached post-sealing -> never part of the record hash / chain.
+_UNHASHED_FIELDS = ("anchor_receipt", "certification_pointer")
+
+
+class EventType(str, Enum):
+    """We seal what the agent received and what it triggered — nothing else."""
+
+    CONTEXT_RECEIVED = "CONTEXT_RECEIVED"
+    ACTION_TRIGGERED = "ACTION_TRIGGERED"
+
+
+#: How a capture gap announces itself in ``GapManifest.notes``.
+#:
+#: Defined HERE because two packages have to agree on it and neither owns the other: the capture
+#: middleware writes it when a capture is dropped, and the alarm reads it to tell the operator.
+#: Two spellings of one marker is a gap that is recorded and never reported — which is the same as
+#: not recording it, on the day it matters.
+RUPTURE_NOTE_PREFIX = "rupture: "
+
+
+class GapManifest(BaseModel):
+    """Declares which capture tiers were active, scoping any completeness claim.
+
+    Every tier defaults to FALSE, including the strongest one. A record built without saying how it
+    was captured must not claim we watched it happen — and until 2026-08-25 it did, because
+    ``tier1_runtime_hooks`` defaulted to True and ``Record.gap_manifest`` has a default factory.
+    Both archive paths of the product's closed service build records without a manifest, so every
+    archived record asserted boundary capture; on the hashes-only path, where the operator sends an
+    address and we see nothing by construction, that assertion was flatly false.
+
+    The adapter that genuinely installs runtime hooks says so explicitly and still does. Everything
+    else now under-claims, which is the only direction this may ever fail in.
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    tier1_runtime_hooks: bool = False
+    tier2_transport_bytes: bool = False
+    tier3_mcp_proxy: bool = False
+    notes: str | None = None
+
+
+class AnchorReceipt(BaseModel):
+    """External-anchor evidence (the moat field); attached after batched anchoring."""
+
+    model_config = ConfigDict(extra="forbid")
+
+    merkle_root: str
+    tree_size: int
+    anchored_at: str
+    timestamp_ref: str | None = None  # always set; honest about the token
+    timestamp_algorithm: str | None = None  # e.g. "local-ed25519-ts" (non-qualified) vs "rfc3161-cms"
+    qualified_timestamp_ref: str | None = None  # ONLY set for a genuinely qualified token
+    witness_checkpoint_ref: str | None = None
+    public_anchor_ref: str | None = None
+
+
+class Record(BaseModel):
+    """A single sealed context/action record.
+
+    The field numbers in comments number the record's fourteen fields.
+    """
+
+    model_config = ConfigDict(extra="forbid")
+
+    # 13 — re-verifiability + reproducible hashing
+    schema_version: str = SCHEMA_VERSION
+    hash_algorithm: str = HASH_ALGORITHM
+    canonicalization_method: str = CANONICALIZATION_METHOD
+
+    # 1 record id, 2 chain link, 3 content-address of the sealed payload
+    record_id: str
+    prev_record_hash: str | None = None
+    payload_hash: str
+
+    # 4 trusted time (RFC 3339 string; upgrade to a qualified TS where weight matters)
+    timestamp: str
+
+    # 5 runtime + version, 6 accountable operator
+    agent_system_id: str
+    agent_version: str
+    deployer_operator_id: str
+
+    # 7 event type; 8/9 the payload itself is off-record (payload_hash), action_type
+    #   is the small inline metadata for ACTION_TRIGGERED
+    event_type: EventType
+    action_type: str | None = None
+
+    # 10 accountable authority
+    actor_authority: str | None = None
+
+    # 11 retention + legal hold
+    retention_class: str
+    legal_hold: bool = False
+
+    # capture-completeness scope
+    gap_manifest: GapManifest = Field(default_factory=GapManifest)
+
+    # 12 anchor (post-seal) + 14 certification (post-seal) -> EXCLUDED from the hash
+    anchor_receipt: AnchorReceipt | None = None
+    certification_pointer: str | None = None
+
+    @model_validator(mode="after")
+    def _check_event_fields(self) -> Record:
+        if self.event_type is EventType.ACTION_TRIGGERED:
+            if not self.action_type:
+                raise ValueError("ACTION_TRIGGERED requires action_type")
+        elif self.action_type is not None:
+            raise ValueError("action_type must be unset for CONTEXT_RECEIVED")
+        return self
+
+    def hashable_dict(self) -> dict[str, Any]:
+        """The JSON-able record content that the hash/chain cover (post-seal fields removed)."""
+        data: dict[str, Any] = self.model_dump(mode="json")
+        for field in _UNHASHED_FIELDS:
+            data.pop(field, None)
+        return data
+
+    def canonical_bytes(self) -> bytes:
+        """Deterministic RFC 8785 (JCS) serialization of the hashable content."""
+        return canonicalize(self.hashable_dict())
+
+    def record_hash(self) -> bytes:
+        """The raw multihash over the canonical bytes."""
+        return multihash(self.canonical_bytes())
+
+    def record_hash_address(self) -> str:
+        """The content address (multibase string) of this record's hash."""
+        return content_address(self.canonical_bytes())
+
+
+def link(prev: Record, **fields: Any) -> Record:
+    """Build the next record, chaining ``prev_record_hash`` to ``prev``."""
+    return Record(prev_record_hash=prev.record_hash_address(), **fields)
+
+
+def verify_chain(records: list[Record]) -> bool:
+    """Return True iff each record's ``prev_record_hash`` matches the prior record."""
+    prev_addr: str | None = None
+    for rec in records:
+        if rec.prev_record_hash != prev_addr:
+            return False
+        prev_addr = rec.record_hash_address()
+    return True

tallwright_core/write.py

@@ -0,0 +1,36 @@
+# SPDX-License-Identifier: Apache-2.0
+"""Replacing a document's contents in one step, or not at all.
+
+The mirror of :mod:`tallwright_core.document`: that one refuses to READ a file that says two
+things, this one refuses to LEAVE a file that is half of one.
+
+``Path.write_text`` opens with ``"w"``: it truncates first and writes second. A crash or a full
+disk between those leaves a file that is neither the old contents nor the new " and the files this
+matters for are the ones that hold everything. `fleet.json` carries every tenant an operator has,
+every API-key hash and the witness identity they all anchor to; the operator config carries the
+checkpoint key's public half and the origin. Losing one of those to a torn write costs a customer
+list, not a cache.
+
+**This does not claim durability.** Without an ``fsync`` a crash can lose the write entirely. What
+it cannot do any more is lose it halfway, and a lost write leaves the OLD file " which is a thing
+somebody can still read " where a torn one leaves nothing.
+
+The scratch lives in the SAME directory, because ``os.replace`` is only atomic within one
+filesystem. Where the target directory is sealed append-only, the scratch has to live outside it
+instead and the rename goes in " that case is not this one and
+``Archive._stage_anchor_file`` handles it separately.
+"""
+
+from __future__ import annotations
+
+import os
+from pathlib import Path
+
+__all__ = ["write_document"]
+
+
+def write_document(path: Path, text: str) -> None:
+    """Write ``text`` to ``path`` so a reader sees either the old contents or all of the new."""
+    scratch = path.with_name(f".{path.name}.part")
+    scratch.write_text(text, encoding="utf-8")
+    os.replace(scratch, path)

tallwright_core-0.1.0.dist-info/METADATA

@@ -0,0 +1,42 @@
+Metadata-Version: 2.5
+Name: tallwright-core
+Version: 0.1.0
+Summary: Tallwright - shared record schema, content-addressing, and RFC 6962 Merkle primitives
+Author: Tallwright
+License-Expression: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Keywords: ai-agents,audit,compliance,forensics,merkle,tamper-evident
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Legal Industry
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Requires-Dist: pydantic<3,>=2
+Requires-Dist: rfc8785<1,>=0.1
+Description-Content-Type: text/markdown
+
+# tallwright-core
+
+The dependency-light foundation every other Tallwright package builds on. No
+networked or external-service code lives here.
+
+- **`hashing.py`** — SHA-256 self-describing multihash + multibase content-addressing
+- **`merkle.py`** — RFC 6962 Merkle tree: root, inclusion + consistency proofs, and
+  independent (RFC 9162) verifiers
+- **`schema.py`** — the 14-field record envelope,
+  RFC 8785 (JCS) canonicalization, record hash, and hash-chain (`link` / `verify_chain`).
+  Post-seal fields (`anchor_receipt`, `certification_pointer`) are excluded from the hash.
+
+Status: **implemented + tested** (the first approved production code). Anchoring, signing,
+storage, and capture live in later packages.
+
+## License
+
+Copyright 2026 David Jöbstl. Licensed under the Apache License, Version 2.0: the package carries the license
+text and the NOTICE. The exhibit
+format's specification and its test vectors are under the same license; the rest of Tallwright is
+proprietary.

tallwright_core-0.1.0.dist-info/RECORD

@@ -0,0 +1,13 @@
+tallwright_core/__init__.py,sha256=veR1oRr5cgnXJZZX4IV7ZyhLQeG4BzG3m-nNMi64J8Y,2002
+tallwright_core/canonical.py,sha256=Wp_A24lWyaPnU-de9rhmS6dG1p_HE64emAD5FOLDE8E,4374
+tallwright_core/document.py,sha256=6gUItl8iaQr-OXGr3WS7gcdcnAz6jKov89-KMzOVUhg,5197
+tallwright_core/hashing.py,sha256=1F6SX31sCNCaEV23UpN6Dv3Z8uYdihdSX9qZ-peWDcU,9296
+tallwright_core/merkle.py,sha256=xSUCJ2kMkRCF2G2irWxspFalh-NpsjPE9x2kWeIeses,6328
+tallwright_core/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+tallwright_core/schema.py,sha256=WT5f7oMDpZMjpNOdZY34BJaWk_NdND2tKjnbeNlNn2Q,6788
+tallwright_core/write.py,sha256=DVD0JaWyvTez8NFBPt2LqFM4MpUMjXJmPdNxbkZvbz4,1710
+tallwright_core-0.1.0.dist-info/METADATA,sha256=f8_2l70RJv-ntWZVwaiZzNBArZZHZ_Tjl_78jZHB7FM,1749
+tallwright_core-0.1.0.dist-info/WHEEL,sha256=zOwg4jB6zX2kU910N-cMawjivD6tO8NEWvE12je1bVk,87
+tallwright_core-0.1.0.dist-info/licenses/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+tallwright_core-0.1.0.dist-info/licenses/NOTICE,sha256=V9DbIbqPvS6ooaxdVebPZgfjXOgZcZcz7hLYGUeRMc8,262
+tallwright_core-0.1.0.dist-info/RECORD,,

tallwright_core-0.1.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.32.0
+Root-Is-Purelib: true
+Tag: py3-none-any

tallwright_core-0.1.0.dist-info/licenses/LICENSE

@@ -0,0 +1,202 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

tallwright_core-0.1.0.dist-info/licenses/NOTICE

@@ -0,0 +1,6 @@
+Tallwright public verifier
+Copyright 2026 David Jöbstl
+
+Licensed under the Apache License, Version 2.0: tallwright-core, tallwright-verifier, the
+specification of the exhibit format and its test vectors. The license text is in the LICENSE
+file beside this one.