The oracle had two import names, `tools.oracle` for the tools and
`paperspec` for other repositories, which could load the same code twice
in one process and needed a gate keyed on the module name. It now lives
in the standard src layout under its one name.
- git mv tools/oracle -> src/paperspec; the 22 rule modules are unchanged.
- The tools import `paperspec`. `tools/__init__.py` now holds the Python
3.14 and cryptography/jsonschema check and puts src/ first on sys.path,
so the tools run this checkout's code with or without pixi. The
package's own __init__ checks nothing and exposes __version__.
- pyproject.toml: version from git tags via setuptools-scm
(guess-next-dev, node-and-timestamp, src/paperspec/_version.py,
gitignored). The packages stay listed rather than found because
schemas/, sql/ and vectors/ stay at the root: the spec links to them and
the site serves the schemas at their $id URLs.
- pixi.toml: a `dev` feature and environment with setuptools,
setuptools-scm and python-build, and `pixi run -e dev wheel`.
- check.py scans src/paperspec for rule IDs; README, CONTRIBUTING, the
FAQ, the overview, the vectors and tools READMEs and the changelog point
at the new path. The FAQ no longer says the oracle is not a library.
pixi.lock is not re-solved in this commit: with a dynamic version, pixi
builds the package metadata to lock, which needs the conda environment
installed. Run `pixi lock` and commit the result before pushing; the site
workflow installs with --locked.
pixi run check passes; no computed value, vector, fixture, schema or SQL
file changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Other repositories (paperstack, then paperpack) import the reference
implementation instead of reimplementing the specification; scienting-core
will take the role later behind the same vectors.
- pyproject.toml: distribution and import name `paperspec`, version 0.1.0,
Python >= 3.12, no base dependencies; `tools/oracle` is the package, and
`schemas/`, `sql/` and `vectors/` ship as the data subpackages
`paperspec.schemas`, `paperspec.sql` and `paperspec.vectors`. Extras
`schema`, `crypto` and `tools`.
- tools/oracle/__init__.py: the Python 3.14 and cryptography/jsonschema gate
runs only when imported as `tools.oracle` by the repository's own tools.
- crypto.py imports `cryptography` on first use of HKDF and Ed25519, with an
ImportError naming `paperspec[crypto]`; schema.py imports `jsonschema` and
`referencing` in `Store.validator` and `Store.registry`. package.py, db.py
and receipts.py then import on a bare interpreter.
- pixi.toml installs the package editable (pixi.lock re-solved; no conda pin
changed), so the tools and the package resolve to the same code.
- tools/README.md names the library surface and the extras; CHANGELOG.md
records the packaging change under 0.1.0, since check.py requires the
changelog's first heading to equal the format version.
No computed value, vector, fixture, schema or SQL file changed.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- hashing.algorithm names the hash function in every document; a consumer refuses one it does not implement (doc.hash.algorithm)
- ML-DSA-65 (FIPS 204) is an implemented signature scheme: pure-Python reference in tools/oracle/mldsa.py checked against NIST ACVP vectors, a receipt signed by the new roster key dana, and signature vectors for key generation and a deterministic signature
- roster keys may carry a did:key, which the validator checks against the public key (doc.roster.did), and a witness purpose for parties that cosign log checkpoints
- model provenance: a correction's author and an annotation's generator name the model with provider, version and the hashes of its weights, prompt and parameters, and where it ran
- the transparency log gains the tlog-tiles static layout (fixtures/records/log/), a project settings.log with origin, witnesses and threshold, and an acceptance rule; the fixture checkpoint is cosigned by a witness
- store.sql: roster_keys accepts cose and ml-dsa-65 and gains a did column
- spec chapters 01, 08, 11, 18, 20 and 23, the rationale, glossary, fixture, vector and tool guides updated; stress harness covers ML-DSA
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Keys (spec/02 §2.4), the rule paperstack settled on:
- A work key is a Scholar-style citation key, a hyphen and the first four
Crockford base32 characters of SHA-256("scienting-key/1:" + identity),
where the identity is the first of doi, arxiv, isbn, pmid, pmcid and
openalex in same_as, else the folded title, year and family name. A
collision lengthens the later key by two characters; a key is minted
once and never recomputed. A document key adds .si, .v<seq> or .<type>
for versions other than the version of record.
- Every fixture key is now minted (hale2026harmonic-damb,
okafor2025solvent-acth, writer2027fixed-ynr7.v3, tiny2026tiny-j4qm), so
every node and record ID was re-minted and the docs rewritten to match.
The validator checks the document key against the work key and version
(doc.key.work) and the hash part against the identifier (doc.key.suffix).
vectors/keys.jsonl pins the cases paperstack's tests pin.
Review:
- Spot-check seeds are ECVRF-EDWARDS25519-SHA512-TAI proofs (RFC 9381),
implemented in tools/oracle/vrf.py in plain Python and reproduced
against the RFC's Example 16; the sample ranks by the VRF output. The
earlier Ed25519 seed could be ground by a dishonest checker.
- RFC 9162 inclusion and consistency proofs (crypto.inclusion_path,
verify_inclusion, consistency_proof, verify_consistency) with
scienting-proof/1, vectors/proofs.jsonl and two fixture records.
- document.json carries `history`, the root over the corrections, with
the empty-tree hash for none (spec/11 §11.5, rule doc.history).
- A provisional transparency log (spec/23): log entries, signed-note
checkpoints with key hashes as transparency logs share them, the rule
that a logged revocation beats a self-asserted time, and vectors.
- Reserved signature schemes cose and ml-dsa-65 (spec/08 §8.4).
Identifiers and text:
- urn:scienting: becomes tag:scient.ing,2026: (RFC 4151), since the URN
namespace was never registered.
- Unicode 16.0 is pinned for normalization and categories (spec/01 §1.4);
the checker refuses another version.
Checking:
- check.py: keys, proofs, VRF and checkpoint vectors; the proof and
checkpoint records; a check that every ID and hash the prose quotes
exists in the fixtures or vectors; three more broken packages
(history-wrong, key-suffix, key-work).
- stress.py: key, VRF, proof and checkpoint properties.
- Docs: a keys section in the concepts and FAQ, rationale entries for
keys, VRF, proofs, tag URIs and the Unicode pin, a "Looking ahead"
section, media-type registration and OCI distribution notes.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- Add tools/validate.py: validates packages, and with --correction runs
proposed corrections through the gate against a package without
writing anything (spec/11 §11.4). The gate now refuses a proposal that
fails the correction schema, rule correction.schema.
- tools/load.py validates every package before loading, and with
--bundle --built-from builds a scienting-bundle/1 with the FTS5 search
index and bundle_meta (spec/13 §13.5). Search rows are defined per
unit (sentence, cell, equation outside a sentence, figure-text) with
the parent kind as context.
- Roster keys carry a role, and a key with the role owner may revoke any
receipt (spec/08 §8.7, spec/20); roster schema, store table and the
fixture roster updated. The checker signs a trial revocation to prove
the rule both ways.
- A qa: receipt raises the nodes it sampled except those in its errors
(spec/08 §8.2); the example paper's fidelity counts become 27 F3,
13 F2, 4 F1 and 1 F0.
- The destination chooses the keyed scheme, not the text policy:
hmac-team between the team's machines, hkdf-pages outside the team
(spec/05 §5.8, spec/12 §12.5).
- A TextQuoteSelector must repeat the anchor's plain quote and is absent
for hashed or keyed anchors (spec/18 §18.3).
- Add fixtures/records/suspect.json, a scienting-suspect/1 record.
- Oracle: hashes are kept incrementally and checked after every
correction against a full recomputation; load and write refuse unsafe
manifest paths; Ed25519, HKDF and schema validation come from the
cryptography and jsonschema packages; canonical JSON uses the built-in
encoder with an exact RFC 8785 fallback.
- The tools need Python 3.14 with cryptography and jsonschema; the py310
CI environment is dropped and the workflow runs the default only.
- Resolve every [CHECK] note in the docs, READMEs and context; add
CONTRIBUTING.md; every tool answers --help and refuses unknown options.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
alexm
