axiomize 1.6.0__tar.gz

axiomize-1.6.0/.gitattributes

@@ -0,0 +1,18 @@
+# Store all text as LF in the repo so the same blob checks out identically
+# everywhere; git still hands the working copy the OS-preferred ending via
+# core.autocrlf where configured.
+* text=auto eol=lf
+
+# Windows launchers must keep CRLF — cmd.exe mis-parses LF-only batch files
+# in some cases, and these are the documented entry points.
+*.bat   text eol=crlf
+*.cmd   text eol=crlf
+*.ps1   text eol=crlf
+
+# Binary assets — never normalize, never diff as text.
+*.png   binary
+*.jpg   binary
+*.jpeg  binary
+*.gif   binary
+*.ico   binary
+*.pdf   binary

axiomize-1.6.0/.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,15 @@
+---
+name: Bug report
+about: Something in the skill or tools is wrong
+title: "[BUG] "
+labels: bug
+---
+
+**What happened?**
+<!-- Include the exact phase/output where the skill misbehaved -->
+
+**Reproduction:** which idea/prompt triggered it?
+
+**Environment:** runtime (Claude Code / opencode / other), model, OS
+
+**Expected behavior per SKILL.md:** <!-- quote the rule that was violated, if any -->

axiomize-1.6.0/.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,10 @@
+# What does this PR change?
+<!-- one paragraph -->
+
+# Checklist
+
+- [ ] New/changed markdown passes `python skills/axiomize/tools/check_skill.py`
+- [ ] Perspective files follow the contract in CONTRIBUTING.md (all four sections)
+- [ ] Examples follow the 8-phase structure and include at least one rejected-lens rationale
+- [ ] Tool changes: sanity checks still exit non-zero on failure; default runtime < 60s
+- [ ] No emojis in content; units stated wherever quantities appear

axiomize-1.6.0/.github/dependabot.yml

@@ -0,0 +1,14 @@
+version: 2
+updates:
+  # Only github-actions. There is no pip manifest in this repo — the two
+  # runtime packages the validators need (numpy, scipy) are installed inline
+  # by the CI step, so there is nothing for the pip ecosystem to watch.
+  - package-ecosystem: "github-actions"
+    directory: "/"
+    schedule:
+      interval: "weekly"
+    labels:
+      - "dependencies"
+      - "github-actions"
+    commit-message:
+      prefix: "ci"

axiomize-1.6.0/.github/workflows/ci.yml

@@ -0,0 +1,70 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+# One run per ref. `pull_request` fires again on every push to a PR branch, and
+# the benchmark-regression step at the bottom replays every stored report
+# listed in benchmarks/ideas.json, so a superseded run is not cheap.
+concurrency:
+  group: ci-${{ github.ref }}
+  cancel-in-progress: true
+
+# Nothing here writes: the jobs only lint the skill and replay validators.
+permissions:
+  contents: read
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.9", "3.11", "3.13"]
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-python@v7
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Install dependencies
+        # Tracked manifest so a clean environment installs exactly what the
+        # tools import - keeps CI and local runs from drifting apart.
+        run: pip install -r requirements-test.txt
+      - name: Run test suite
+        # pytest is listed in requirements-test.txt so a clean environment
+        # has it. Fails the build on any regression in validate/fit/csv_check.
+        run: pytest tests/ -v
+      - name: Skill lint (frontmatter + links + compile)
+        run: python skills/axiomize/tools/check_skill.py
+      - name: Deterministic SIR checks
+        run: python skills/axiomize/tools/validate.py --model sir --beta 0.3 --gamma 0.1
+      - name: Stochastic (Gillespie) checks
+        run: python skills/axiomize/tools/validate.py --model gillespie --N 10000 --I0 1 --runs 300
+      - name: Stochastic subcritical/critical regression (fade-out theory)
+        # R0=0.5, I0=1: theory 2/3 (first-jump exact); R0=1, I0=3: theory 1/4.
+        # The old code asserted fade-out probability 1.0 for every R0 <= 1.
+        run: |
+          python skills/axiomize/tools/validate.py --model gillespie --beta 0.05 --gamma 0.1 --I0 1 --N 10000 --runs 400
+          python skills/axiomize/tools/validate.py --model gillespie --beta 0.1 --gamma 0.1 --I0 3 --N 10000 --runs 400
+      - name: Queue (Erlang-C) checks
+        run: python skills/axiomize/tools/validate.py --model queue --lam 60 --mu 20 --target-wait 3
+      - name: Calibration selftests
+        run: |
+          python skills/axiomize/tools/fit.py --model sir --selftest
+          python skills/axiomize/tools/fit.py --model logistic --selftest
+      - name: Parallel sweep engine
+        run: |
+          python skills/axiomize/tools/parallel_sweep.py --job sweep
+          python skills/axiomize/tools/parallel_sweep.py --job mc
+      - name: Benchmark regression (stored blind-test reports)
+        # Case ids come straight from ideas.json, so every case runs and the
+        # list can never silently fall behind the cases that exist.
+        run: |
+          ids="$(python -c "import json; print(' '.join(c['id'] for c in json.load(open('benchmarks/ideas.json', encoding='utf-8'))['cases']))")"
+          echo "grading cases: $ids"
+          for id in $ids; do
+            python skills/axiomize/tools/benchmark_runner.py --case "$id" --report "benchmarks/reports/$id.md" || exit 1
+          done

axiomize-1.6.0/.github/workflows/pages.yml

@@ -0,0 +1,75 @@
+name: Pages
+
+on:
+  push:
+    branches: [main]
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pages: write
+  id-token: write
+
+concurrency:
+  group: pages
+  cancel-in-progress: true
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - uses: actions/setup-python@v7
+        with:
+          python-version: "3.12"
+      - run: pip install mkdocs-material
+      - name: Stage skill docs
+        # SKILL.md and the perspective heads are written to be browsed inside
+        # skills/axiomize/, so their relative links do not survive being
+        # concatenated into a single docs/ page. Everything that is NOT staged
+        # into docs/ (i.e. anything but rigor.md and archetypes.md) is rewritten
+        # to an absolute GitHub URL. `mkdocs build --strict` below is what keeps
+        # this list honest: a new link shape that no rule covers fails the build
+        # instead of shipping a 404 to the site.
+        run: |
+          cp skills/axiomize/rigor.md docs/rigor.md
+          cp skills/axiomize/archetypes.md docs/archetypes.md
+          SKILL_URL='https://github.com/Furox-Art/axiomize/tree/main/skills/axiomize'
+          {
+            echo "# The workflow"
+            cat skills/axiomize/SKILL.md
+            printf '\n\n# The lenses\n\n'
+            for f in skills/axiomize/perspectives/*.md; do
+              echo "## $(basename "$f" .md)"
+              echo
+              head -n 12 "$f"
+              echo
+            done
+          } | sed -E "
+              s#\]\(perspectives/#](${SKILL_URL}/perspectives/#g
+              s#\]\(templates/#](${SKILL_URL}/templates/#g
+              s#\]\(first-principles\.md#](${SKILL_URL}/first-principles.md#g
+              s#\]\((agent-based|causal-inference|control|decision-theory|demographic|deterministic|game-theory|information-theory|network|optimization|reliability|spatial|spc|stochastic|thermodynamic)\.md#](${SKILL_URL}/perspectives/\1.md#g
+            " > docs/skill.md
+          {
+            echo "# Worked examples"
+            for f in examples/*.md; do
+              echo "## $(basename "$f" .md)"
+              echo
+              head -n 8 "$f"
+              echo
+            done
+          } > docs/examples.md
+      - run: mkdocs build --strict
+      - uses: actions/upload-pages-artifact@v5
+        with:
+          path: site
+  deploy:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: github-pages
+      url: ${{ steps.deployment.outputs.page_url }}
+    steps:
+      - id: deployment
+        uses: actions/deploy-pages@v5

axiomize-1.6.0/.gitignore

@@ -0,0 +1,16 @@
+__pycache__/
+*.pyc
+.ipynb_checkpoints/
+.DS_Store
+*.aux
+*.log
+*.out
+texput.log
+
+# mkdocs output, plus the pages that the Pages workflow stages into docs/ from
+# skills/axiomize before building. Generated on every build, never authored.
+site/
+docs/rigor.md
+docs/archetypes.md
+docs/skill.md
+docs/examples.md

axiomize-1.6.0/CHANGELOG.md

@@ -0,0 +1,123 @@
+# Changelog
+
+All notable changes to Axiomize are documented here. Format follows [Keep a Changelog](https://keepachangelog.com/); versioning is SemVer.
+
+## [1.5.0] - 2026-08-24
+
+### Added
+- First-principles protocol (`skills/axiomize/first-principles.md`) for ideas with no matching archetype: derive the model from conservation/accounting identities instead of forcing a known template
+- Two novel-domain benchmark reports exercising that path — async-alignment and telephone-fidelity — neither drawn from the archetype catalog
+- Novel-territory appendix in the report template, recording which quantities were derived rather than borrowed
+- PyPI packaging (`pyproject.toml`, src-layout shim, `axiomize-*` console commands, tracked `requirements-test.txt` manifest); the wheel is assembled straight from `skills/axiomize/` so no file is duplicated
+
+### Fixed
+- Gillespie fade-out theory now uses the gambler's-ruin jump-chain result valid for every R0 (previously asserted probability 1.0 for all R0 ≤ 1); critical point verified as 1/(I0+1)
+- `validate.py --days` is honored by the SIR report and sweep (previously ignored)
+- `csv_check.py` fails hard on unknown column names and treats unsorted time as a data-quality failure (previously silently fell back to columns 0/1 and sorted before grading)
+- Benchmark archetype gate requires the significant tokens of an archetype alternative (the old split-on-"/" reduced "M/M/c queueing" to the letter M and always passed); rejection check accepts the template's "Rejected lenses" heading
+- `report_to_latex.py` neutralizes dangerous TeX macros (`\input`, `\write18`, ...) in math spans and catches pdflatex timeouts (the previous pre-pass preserved live `\input` commands)
+- Playground locates the tools directory when launched via `gradio app.py` and validates the SIR population N before fitting
+
+### Changed
+- SKILL.md routes to the first-principles protocol when archetype matching fails
+
+## [1.4.1] - 2026-08-24
+
+### Fixed
+- LaTeX converter hardening: all 11 worked examples and 8 benchmark reports now compile with zero errors
+- `texput.log`, a LaTeX build artifact, had been committed in 1.4.0; removed and `*.aux` / `*.log` / `*.out` added to `.gitignore`
+
+## [1.4.0] - 2026-08-24
+
+### Added
+- LaTeX/PDF export for reports (`skills/axiomize/tools/report_to_latex.py`): booktabs tables, verbatim code blocks, unicode transliteration
+- Sample rendered report checked in as `docs/report-sample.tex` / `.pdf`
+
+## [1.3.2] - 2026-08-24
+
+### Added
+- Animated demo embedded in the README
+- Benchmark regression wired into CI: the eight stored blind-test reports are replayed on every run, so a scoring regression fails the build
+
+## [1.3.1] - 2026-08-24
+
+### Added
+- Eight blind-test benchmark reports committed under `benchmarks/reports/` (8/8 PASS, mean score 9.35) plus `docs/benchmark-results.md` summarising them
+- Epidemiology and operations domain packs filled out
+
+### Fixed
+- QA-wave defects: fade-out theory formula, Erlang-C overflow on large offered load, `csv_check` reporting a false PASS, and `fit` bounds
+
+### Removed
+- Registry submission kit (`docs/submissions.md`) — premature
+
+## [1.3.0] - 2026-08-24
+
+### Added
+- Three new lenses: decision theory (deep uncertainty, EVPI), demographic/actuarial, spatial statistics — fifteen total
+- Benchmark runner: `benchmark_runner.py` grades produced reports against ideas.json cases automatically
+- `fit.py --json` for machine-readable calibration output
+- Local playground: Gradio UI wrapping csv_check + calibration (`playground/app.py`)
+- Example gallery page and registry submission kit in docs
+- Domain pack: project management
+
+### Fixed
+- benchmark_runner --case-list no longer requires --report
+
+## [1.2.0] - 2026-08-24
+
+### Added
+- Three new lenses: reliability engineering, statistical process control, thermodynamic analogies (twelve total)
+- Three new worked examples: fleet maintenance (reliability), marketing attribution (causal), sensor placement (information theory) — eleven examples total
+- Benchmark suite: `benchmarks/ideas.json` with 8 standard test cases + scoring rubric
+- Domain packs: economics and ecology (alongside epidemiology and operations)
+- Beginner tutorial: `docs/tutorial.md`
+- New tool: `csv_check.py` — data quality pre-check before calibration
+- Parallel subagent wave executed for lens/example authoring; orchestrator integration pattern documented by example
+
+### Fixed
+- f-string backslash incompatibility breaking Python 3.9/3.11 in parallel_sweep.py
+
+## [1.1.0] - 2026-08-24
+
+### Added
+- Three new lenses: game theory, causal inference, information theory (nine total)
+- Archetype catalog expanded 16 → 30 entries
+- Five new worked examples: network rumor, greenhouse control, startup growth (Bass + calibration), insurance ruin risk, café pricing war — network and control lenses now have dedicated examples
+- Mermaid coupling diagrams in Phase 2 and the report template
+- `fit.py`: AIC/BIC diagnostics, residual autocorrelation flag, `--compare` mode ranking models on the same data
+- `tools/index_reports.py`: rebuilds `reports/INDEX.md`; sessions now cross-reference earlier reports
+- Glossary template supporting basic-tier readers
+- GitHub Pages site (`mkdocs-material`) via Pages workflow
+- Domain packs: epidemiology, operations
+- Publishing checklist for external registries
+- CI: Python 3.9 / 3.11 / 3.13 matrix with dependency floors
+
+## [1.0.0] - 2026-08-24
+
+First tagged release.
+
+### Added
+- 8-phase workflow: parse → decompose → parameters → assumptions → multi-perspective modeling → comparison → implementation → falsifiability
+- Six mathematical lenses: deterministic, stochastic, optimization, agent-based, network, control
+- Archetype catalog mapping idea patterns to canonical models (SIR, Bass, newsvendor, M/M/c, logistic, Lotka–Volterra...)
+- Three-tier rigor ladder (basic / standard / research) with plain-language guarantee and automatic escalation on threshold risk
+- Parallel Dispatch Protocol: lenses run as independent subagents from frozen self-contained briefs; assumption conflicts surface at merge
+- Standardized report template with confidence ledger and research appendix
+- Tools bundled inside the skill:
+  - `validate.py` — SIR / Gillespie CTMC / Erlang-C queue modes with sanity checks and sweeps
+  - `fit.py` — parameter calibration from CSV data with confidence intervals and self-tests
+  - `parallel_sweep.py` — real process-pool execution of grids and Monte Carlo chunks
+  - `check_skill.py` — metadata, link and compile linter (wired into CI)
+- Three worked examples: epidemic SIR, retail inventory, coffee-shop staffing
+- GitHub Actions CI running all validation modes
+
+[1.5.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.5.0
+[1.4.1]: https://github.com/Furox-Art/axiomize/releases/tag/v1.4.1
+[1.4.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.4.0
+[1.3.2]: https://github.com/Furox-Art/axiomize/releases/tag/v1.3.2
+[1.3.1]: https://github.com/Furox-Art/axiomize/releases/tag/v1.3.1
+[1.3.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.3.0
+[1.2.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.2.0
+[1.1.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.1.0
+[1.0.0]: https://github.com/Furox-Art/axiomize/releases/tag/v1.0.0

axiomize-1.6.0/CONTRIBUTING.md

@@ -0,0 +1,60 @@
+# Contributing to Axiomize
+
+Thanks for helping make idea-to-model rigor a standard agent capability.
+
+## Adding a perspective (the most valuable PRs)
+
+Each file in `skills/axiomize/perspectives/` follows a fixed contract. A new lens PR must include **all** sections:
+
+```
+# Perspective: <Name> (<one-line essence>)
+
+## When Applicable
+- Explicit triggers tied to the Phase 2 classifications (flow / interaction / decision / uncertainty)
+- What questions this lens answers that others cannot
+
+## Model Forms
+- The 2-4 canonical formalisms, with checklists for building them correctly
+- Concrete functional forms (not "model it appropriately")
+
+## Standard Analysis Output
+- Numbered list of artifacts every analysis must produce
+
+## Strengths / Blind Spots
+- (+) what this view uniquely sees
+- (-) what this view cannot see (honesty is the product)
+```
+
+Rules for perspective content:
+
+1. Every equation symbol must be defined inline.
+2. Units are mandatory where dimensional.
+3. Include an "analysis ladder" (cheap → expensive methods) if more than one fidelity level exists.
+4. No filler prose — a domain expert should be able to build a first model from your file alone.
+
+Candidate lenses not yet covered: queueing networks beyond M/M/c, survival analysis with competing risks beyond reliability scope.
+
+## Adding worked examples (`examples/`)
+
+Follow the 8-phase structure exactly as in `examples/epidemic-sir.md`. Requirements:
+
+- Real-ish parameter ranges with source classes (`lit.` / `data` / `est.`)
+- At least two perspectives actually built, plus at least one **explicitly rejected with a one-line reason**
+- A falsifiability section naming observations that would kill the model
+- If you add runnable tooling, extend `skills/axiomize/tools/validate.py` (see below)
+
+## Extending `skills/axiomize/tools/validate.py`
+
+Every new model mode must print sanity checks and exit non-zero when they fail. Accepted checks: conservation laws, bounds/monotonicity, agreement with a closed-form theory result (within stated tolerance), or distributional consistency across Monte Carlo runs. CI runs all modes — keep default parameters under ~60s total runtime.
+
+## Style
+
+- Markdown for docs; Python 3.9+ stdlib + numpy/scipy only.
+- No comments in code unless explaining a non-obvious formula's origin.
+- English for repo content.
+
+## Submitting
+
+1. Fork & branch (`feat/<topic>`).
+2. Run all validate modes locally before the PR.
+3. Describe what lens/example adds to coverage that no existing file provides.

axiomize-1.6.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Furkan
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.