Skip to content

v2.0.0-beta.1

Latest

Choose a tag to compare

@github-actions github-actions released this 16 Jul 00:48

Arisbe v2.0.0-beta.1 — "Moses"

The last beta before the second-order crossing. This release tags the completed
first-order territory — Peirce's Alpha and Beta graphs under Frithjof Dau's
formalization, made operational as an environment for doing logic in pictures
at the point where the project pauses, deliberately, on the edge of logic about
the graphs themselves (quotation, (forces s φ); see
docs/SECOND_ORDER_FRONTIER.md). Moses looks over;
the crossing is the next line's work.

What Arisbe is

An environment for doing logic in pictures, not pictures of logic. The
fundamental entity is the Universe of Discourse — a diachronic process of
reasoning in which a single Existential Graph is one synchronic snapshot. The
central engineering problem the codebase solves is inerrant correspondence
between a graph's linear written form and its graphical drawn form: the contract
(docs/LINEAR_GRAPHICAL_CORRESPONDENCE.md)
is runtime-attested at every service boundary — when picture and proposition come
apart, the system refuses to serve the drawing.

What this release contains

  • The exact-correspondence engine — a cut is its drawn curve; containment,
    ligature crossing-sequences, label extents, and argument order are checked as
    exact facts about the literal picture. Layout is deterministic: the same
    graph draws bit-identically across processes.
  • Three web modesOrganon (read-only archive with modal and audit lenses),
    Ergasterion (freeform draw-then-read composition, challenge mode, rule
    workshop), Agon (the Endoporeutic Game: contest and interpretation registers).
    One rendering engine end to end; every served (graph, drawing) pair attested.
  • The interpretation register — a proposition peeled outside-in against a
    domain model M (three-valued, open-world), with Horn materialization, theory
    queries over pure T-boxes, and model rendering with an honest horizon.
  • The validity discipline — nothing contingent stands at depth 0: M resides
    as a supposition at level 1 of a standing world-scroll ~[ M ~[ ] ]; every
    change to M is an explicit rule-licensed step, every verdict a recorded,
    forever-recomputable peel, guarded by a standing corpus gate
    (tests/test_corpus_polarity_discipline.py). The register of fidelity to
    Peirce — devotions, departures, adversarial examination — is written down
    (docs/FIDELITY_AND_DEPARTURES.md).
  • Automated model development — the Agon as engine of change: proposer
    membranes (closed, raise-only, raise-and-resolve, wiki-dispute), three LLM
    roles under an incorruptible mechanical referee, disuse-decay, a bounded
    paced checkpointed live runner, and live sources (Wikidata, weather,
    prediction-scored sports). Twelve pre-registered live runs, each disposed
    into a findings log (runs/).
  • Modality without Gamma — ◇/□ read off the branching diachronic history
    (no modal mark), surfaced as Organon lenses.
  • Import and export — EGIF/CGIF/CLIF/FOPL round-trips (corpus-tested),
    OWL/RDF ontology import to attested UoDs, authentic-Peirce LaTeX/TikZ export,
    provenance-to-citation, and the verifier exposed as an MCP service for LLM
    agents (src/mcp_server.py).
  • The book — the documentation as a rendered Quarto book (42 chapters),
    served at /book, with role-aware on-ramps from newcomer to Peirce scholar.

Honest state (known issues)

  • Two regimes. The attested corpus enacts the validity discipline; the live
    automated loops
    still run the older sheet-level regime, honestly flagged in
    their documents, pending a separate ordered decision
    (M_RESIDENCE_AND_THE_VALIDITY_DISCIPLINE §8.1).
  • Run 12 (MLB, prediction-scored) is in flight during the all-star break;
    its disposal lands in runs/RUN_12_LOG.md after play resumes.
  • Standing polish threads (not release-blocking): layout at ontology scale
    (BFO/SUMO density), FOPL panel display nuance, non-visual projection (a11y).
  • Two imported ontologies (bfo_core, colore_field) carry structures whose
    EGIF linearization is lossy (cross-sibling vertex references from the OWL
    disjointness expansion); the corpus handles them structurally and records it
    in their annotations — a per-axiom variable-freshening pass in the importer
    is the candidate fix.

Quickstart

git clone https://github.com/mijahauan/Arisbe && cd Arisbe
uv sync --extra dev --extra web && npm install
uv run uvicorn --app-dir src web_api.main:app --port 8000
# open http://localhost:8000  (the book is at /book)

Verified by ~3350 tests including the correspondence invariant suite, the
mathematical core suite, and the corpus polarity gate; CI runs the full suite
on every push and on this tag.