codontrace 0.3.0b3__tar.gz → 0.3.0b4.dev0__tar.gz

codontrace-0.3.0b4.dev0/.github/workflows/ci.yml

@@ -0,0 +1,136 @@
+name: CI
+
+"on":
+  push:
+  pull_request:
+
+permissions:
+  contents: read
+
+env:
+  FORCE_JAVASCRIPT_ACTIONS_TO_NODE24: "true"
+  PIP_DISABLE_PIP_VERSION_CHECK: "1"
+
+jobs:
+  ubuntu-full-tests:
+    name: Full pytest (${{ matrix.os }}, Python ${{ matrix.python-version }})
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 60
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-latest]
+        python-version: ["3.11", "3.12", "3.13", "3.14"]
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+
+      - name: Set up Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+
+      - name: Install test dependencies
+        run: |
+          python -m pip install --upgrade pip setuptools wheel
+          python -m pip install -e . pytest
+
+      - name: Compile package, tests, examples, and tools
+        run: python -m compileall -q src tests examples tools
+
+      - name: Enforce core boundary with AST import guard
+        run: python tools/check_core_boundary.py
+
+      - name: Reject probe leftovers in the checkout
+        run: |
+          python - <<'PY'
+          import fnmatch
+          from pathlib import Path
+
+          skip_parts = {".git", ".venv", "venv", "__pycache__"}
+
+          def is_leftover(path: Path) -> bool:
+              name = path.name
+              if name.startswith(".grok_write_probe"):
+                  return True
+              if name.startswith(".size_test_") and path.suffix == ".txt":
+                  return True
+              return any(
+                  fnmatch.fnmatch(name, pattern)
+                  for pattern in ("*_FULL_RESTORE*", "*_backup*", "*.orig")
+              )
+
+          junk = sorted(
+              p
+              for p in Path(".").rglob("*")
+              if p.is_file()
+              and skip_parts.isdisjoint(p.parts)
+              and is_leftover(p)
+          )
+          if junk:
+              raise SystemExit("probe leftovers must not ship: " + ", ".join(map(str, junk)))
+          print("no probe leftovers")
+          PY
+
+      - name: Run full test suite
+        run: python -m pytest tests --disable-plugin-autoload
+
+  core-smoke-cross-os:
+    name: Core smoke (${{ matrix.os }}, Python ${{ matrix.python-version }})
+    runs-on: ${{ matrix.os }}
+    timeout-minutes: 30
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [windows-latest, macos-latest]
+        python-version: ["3.11", "3.12", "3.13", "3.14"]
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+
+      - name: Set up Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: pip
+
+      - name: Install test dependencies
+        run: |
+          python -m pip install --upgrade pip setuptools wheel
+          python -m pip install -e . pytest
+
+      - name: Compile package, tests, examples, and tools
+        run: python -m compileall -q src tests examples tools
+
+      - name: Enforce core boundary with AST import guard
+        run: python tools/check_core_boundary.py
+
+      - name: Run release-critical tests
+        run: python -m pytest tests/test_export_import.py tests/test_api_signature.py tests/test_genesis_causal_mechanism_hardening.py tests/test_genesis_causal_mechanism_runtime_wiring_integrity.py tests/test_genesis_integration_public_api_unification.py tests/test_genesis_integration_runtime_wiring.py tests/test_genesis_eval_bugfixes.py tests/release tests/genesis_gates --disable-plugin-autoload
+
+  lint-type:
+    name: Lint and typecheck (Ubuntu, Python 3.12)
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    continue-on-error: true
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+
+      - name: Set up Python
+        uses: actions/setup-python@v6
+        with:
+          python-version: "3.12"
+          cache: pip
+
+      - name: Install lint and typecheck dependencies
+        run: |
+          python -m pip install --upgrade pip setuptools wheel
+          python -m pip install -e ".[dev]"
+
+      - name: Ruff
+        run: python -m ruff check src tests
+
+      - name: Mypy
+        run: python -m mypy --strict src

{codontrace-0.3.0b3 → codontrace-0.3.0b4.dev0}/CHANGELOG.md

@@ -2,6 +2,240 @@
 
 ## Unreleased
 
+### Wave 2 — simulator-agnostic ClaimGate auditor
+
+CodonTrace Genesis standalone evidence auditor. **Does not** claim
+intelligence, collective intelligence, Tokyo Type 1 passed, or Avida
+replacement. Identity stays `0.3.0b4.dev0`. Phase A–E pins stay
+stable. ClaimGate is not loosened.
+
+- New package `codontrace.claimgate`: `claimgate_bundle_v1` +
+  `audit_bundle` → public 0–5 (`CLAIMS.md` §5 + §8). CLI
+  `python -m codontrace.claimgate audit bundle.json`.
+- `StrongClaimLadderResult.public_level` maps the nine internal rungs
+  through `docs/CLAIM_LADDER_MAP.md`. Digest payload unchanged.
+- Adapters: complete CodonTrace Genesis HARD_EXPERIMENT_01 v2;
+  Avida `.dat` and MABE2 DataFile CSV **skeletons** only.
+- HARD_EXPERIMENT_01 v2 still grades `runtime_observation` / public
+  level 1, labeled `assay_invalid` (bitwise-identical fitness across
+  arms; manipulation not realized), not a scientific null. MODES /
+  Tokyo Type 1 remain measurement neighbors, not a pass. ASME V&V 40
+  is context-of-use, not a numeric score.
+- Docs: `docs/CLAIMGATE_STANDALONE.md`, collapsed
+  `docs/CLAIM_LADDER.md`, `docs/design/*` literature notes.
+
+### HARD_EXPERIMENT_01 Wave 1b — survival calibration + assay gate
+
+CodonTrace Genesis harden of the same Wave 1 question. **Does not**
+claim intelligence, collective intelligence, Tokyo Type 1 passed, or
+Avida replacement. Identity stays `0.3.0b4.dev0`. Phase A–E pins stay
+stable. ClaimGate is not loosened.
+
+- Root-cause: research v1 (`30 / 40 / 16`) was an invalid null — every
+  arm extinct by tick 10, zero `EMIT_NEXUS`, zero adoptions. Energy
+  budget vs `COPY_SELF`, two food cells vs pop 16, no emit codon.
+- Overlay-only calibration (food inflow, ATP, basal cost, death
+  patience, initial genomes) so the existing capsule channel can fire.
+  Estimand / arms / seeds / analysis unchanged.
+- Assay gate: `assay_failed` when treatment mean adoptions ≈ 0 or
+  extinction ≈ 1; ceiling stays `runtime_observation`.
+- Research v2 (`docs/hard_experiment_01/results_v2.json`): treatment
+  adoptions 169, extinction 0, assay passed; primary dz = 0,
+  CI `[0,0]`, Holm p = 1.0; ceiling `runtime_observation`. Interpretable
+  null after the channel fired.
+
+### HARD_EXPERIMENT_01 Wave 1 — preregistered causal design
+
+CodonTrace Genesis confirmatory measurement on the Wave 0 helpers.
+**Does not** claim intelligence, collective intelligence, Tokyo Type 1
+passed, or Avida replacement. Identity stays `0.3.0b4.dev0`. Phase A–E
+pins stay stable. ClaimGate is not loosened.
+
+- Preregistration `docs/HARD_EXPERIMENT_01_PREREG.md` is committed
+  *before* campaign numbers. Campaigns record `prereg_digest`.
+- Four arms: `source_bias_on`, `source_bias_off`, `capsules_off`,
+  `capsules_shuffled` (`CapsuleShuffleMode.CONTENT`). Explicit DAG
+  (Okasha & Otsuka 2020) states which edge each arm cuts.
+- Dose-response `min_source_fitness ∈ {0,1,2,4}` with `FITNESS_WEIGHTED`
+  and a seed-fixed Spearman permutation trend (Goldsby 2012 pattern).
+- Paired dz + BCa 95% CI (10000) + sign-flip p; Holm across three
+  primary contrasts; `PairedComparisonResult.claim_downgraded` when CI
+  includes 0; missing outcomes dropped, never zero-filled.
+- Smoke (12 / 6 / 4) stays in CI at `runtime_observation`. Research
+  (30 / 40 / 16) may request existing `intervention_supported` only if
+  the preregistered decision rule holds and ClaimGate allows it.
+- Research v1 (`docs/hard_experiment_01/results_v1.json`) is a null
+  finding: dz = 0, CI `[0, 0]`, Holm p = 1.0, ceiling
+  `runtime_observation`. Replay matched. Capsule adoptions were 0.
+
+### Advisor hygiene + hard experiment 01
+
+CodonTrace Genesis hygiene and one measurement paper. **Does not** claim
+intelligence, collective intelligence, Tokyo Type 1 passed, or Avida
+replacement. Development identity `0.3.0b4.dev0`. Phase A–E default digest
+pins stay stable.
+
+- Deleted probe leftovers from the published tree:
+  `.grok_write_probe2.txt`, `.grok_write_probe3.txt`, `.size_test_10k.txt`.
+  Added `.gitignore` and `MANIFEST.in` excludes so they are never packaged.
+- Default PR CI now runs full `pytest tests` on Ubuntu × Python 3.11–3.14.
+  Cross-OS jobs keep the release-critical smoke subset (timeout split).
+- Started `engine.py` decomposition: `codontrace.engine_digest` (replay
+  hashes) and `codontrace.engine_results` (tick/snapshot/run identity).
+  Historical imports unchanged. Contract:
+  `docs/ENGINE_REPLAY_CONTRACT.md`.
+- Hard experiment 01 (`codontrace.genesis.hard_experiment_01`): 12-seed
+  paired source-bias on vs source-bias off vs capsules off; effect size;
+  independent replay digests; ClaimGate ceiling `runtime_observation`.
+  Example `examples/genesis_hard_experiment_01.py`. Write-up
+  `docs/HARD_EXPERIMENT_01.md`.
+- Phase H–L “not intelligence” pointers folded into
+  `docs/PHASE_INDEX.md`. Canonical honesty doc remains
+  `docs/WHY_NOT_INTELLIGENCE_YET.md`.
+- Full-suite honesty: register Phase H–L / RAG / hard-experiment digest
+  dataclasses in the replay policy sweep (measurement/reference, not
+  claim-granting). Remove the core `print()` from
+  `print_collective_intelligence_candidate_checklist` (examples may
+  print the rendered checklist). These tests were never in the old CI
+  subset.
+- Working-tree identity is `0.3.0b4.dev0` (`pyproject.toml`,
+  `codontrace.__version__`, `CITATION.cff`, release-label aliases). The
+  public PyPI tip remains `0.3.0b3`.
+- Contributor naming lives in `CONTRIBUTING.md` / `STYLE.md`. README opens
+  with a plain 30-second intro instead of an internal naming-order note.
+- Hard experiment 01 maps each arm to an explicit
+  `CapsuleTransferConfig` intervention (`hard_experiment_01_interventions`).
+
+### Phase L — Avida ORGANISM_MESSAGING / DEME_GROUP fidelity
+
+CodonTrace Genesis measurement-first follow-on after Phase K (`2614ec4`).
+**Does not** claim intelligence, collective intelligence, Tokyo Type 1
+passed, or Avida replacement. ClaimGate unchanged in the strict
+direction. No version bump. Phase A–E default digest pins stay stable.
+
+- ORGANISM_MESSAGING analog
+  (`evaluate_organism_messaging_group` /
+  `run_organism_messaging_fidelity_experiment`): per-organism FIFO
+  retrieve, faced-neighbor send, `rotate_cw`, `block_propagation`, Phase E
+  `DemeState` audit. Analog ISA, not Avida C++.
+- `DEME_GROUP` analog: optional `maybe_replicate_demes` with extra target
+  deme / germline copy recorded. Not avida.cfg hardware.
+- Goldsby 2012-aligned specialist measurement
+  (`run_goldsby_aligned_specialist_campaign`): behavioral vs genotypic
+  specialists, clonal-group DoL, colony-quota analog, ancestral vs evolved
+  isolation, delay dose-response. `is_goldsby_2012_pnas_experiment=False`.
+- Optional Phase K coordination-ablation evidence feed
+  (`feed_phase_k_coordination_ablation_evidence`): never auto-sets
+  ClaimGate flags; smoke never earns.
+- RAG corpus: `avida_cfg_organism_messaging`, `avida_cfg_deme_group`;
+  Goldsby messaging / 2012 digests updated.
+- Docs: `docs/PHASE_L_AVIDA_FIDELITY.md`.
+- Example: `examples/genesis_phase_l_avida_fidelity.py`.
+
+### Phase K — literature-grade CI depth
+
+CodonTrace Genesis measurement-first follow-on after Phase J (`899addc`).
+**Does not** claim intelligence, collective intelligence, Tokyo Type 1
+passed, or Avida replacement. ClaimGate unchanged in the strict
+direction. No version bump. Phase A–E default digest pins stay stable.
+
+- Evolved coordination instructions
+  (`run_evolved_coordination_instruction_experiment`): discrete
+  send/retrieve/broadcast/work genomes on the Phase E buffer; messaging
+  ablation can drop coordination payoff. Analog ISA, not Avida C++.
+- Goldsby-scale CPU-delay specialist harness
+  (`run_goldsby_cpu_delay_specialist_campaign`): 0/25/50-cycle analog
+  treatments; research default **50** replicates; isolation vs group
+  dual-task hooks. `is_goldsby_2012_pnas_experiment=False`.
+- Multi-generation Price analysis
+  (`run_multi_generation_price_analysis`): transmission term from
+  realized parent–offspring Δz; `price_equation_complete=False`.
+- Michod/Conlin conflict-suppression hooks
+  (`build_conflict_suppression_observation`): within-group variance,
+  cheater invasion, revertant isolation. Endogenous / transition flags
+  stay False.
+- RAG corpus: Price 1972 transmission digest; Goldsby / Michod / Conlin
+  / Okasha / Price 1970 updated.
+- Docs: `docs/PHASE_K_CI_DEPTH.md`.
+- Example: `examples/genesis_phase_k_ci_depth.py`.
+
+### Phase J — replay-verified CI candidate packs
+
+CodonTrace Genesis measurement-first follow-on after Phase I (PR #14 @
+`6a0fe1a`). **Does not** claim intelligence, collective intelligence, Tokyo
+Type 1 passed, or Avida replacement. ClaimGate unchanged in the strict
+direction. No version bump. Phase A–E default digest pins stay stable.
+
+- Replay-verified CI candidate pack
+  (`build_replay_verified_ci_candidate_pack`): runs Phase I research-scale
+  earnable campaigns, independently re-executes them, and earns
+  `replay_verification` only when campaign digests match. Same-object
+  “replay” is rejected. Smoke never earns flags.
+- When every candidate flag is honestly earned at research scale,
+  ClaimGate **allows** `collective_intelligence_candidate`. Bare
+  `collective_intelligence` / proved CI stay forbidden.
+- Price-equation / Okasha covariance scaffold
+  (`run_price_equation_covariance_scaffold`): last-generation MLS1
+  between/within partition on the two-task analog; transmission term not
+  estimated; `major_transition_in_individuality=False`.
+- RAG corpus: Price 1970 Nature; Okasha 2006 digest updated.
+- Docs: `docs/PHASE_J_REPLAY_CI.md`.
+- Example: `examples/genesis_phase_j_replay_ci.py`.
+
+### Phase I — collective-intelligence evidence harnesses
+
+CodonTrace Genesis measurement-first follow-on after Phase H (PR #13).
+**Does not** claim intelligence, collective intelligence, Tokyo Type 1
+passed, or Avida replacement. ClaimGate unchanged in the strict direction.
+No version bump. Phase A–E default digest pins stay stable.
+
+- Heldout unfamiliar-partner generalization
+  (`run_heldout_unfamiliar_partner_experiment`): familiar vs heldout lineage,
+  distinct partner digests, leakage status. Smoke does **not** earn
+  ClaimGate `heldout_protocol`.
+- Evolved (not assigned) division of labor
+  (`run_evolved_division_of_labor_experiment`): heritable two-task preference,
+  Gorelick-style NMI, all-A ablation. Does **not** auto-set `ablation_result`.
+- MLS evolutionary *outcome* vs organism-only
+  (`run_mls_evolutionary_outcome_experiment`): majority strategy contrast
+  (Okasha MLS2 analog), not a rank table.
+- Export-of-fitness scaffold (`build_export_of_fitness_observation`): isolation
+  vs group competence; isolation collapse marked as payoff construction;
+  `major_transition_in_individuality=False`.
+- `earn_collective_intelligence_candidate_flags`: research-scale evidence
+  objects may earn candidate flags; smoke never earns;
+  `replay_verification` is not earnable from Phase I.
+- RAG corpus: Gorelick 2004; Okasha 2006; Michod 2007 PNAS.
+- Docs: `docs/PHASE_I_CI_EVIDENCE.md`; WHY_NOT / PHASE_H cross-links.
+- Example: `examples/genesis_phase_i_ci_evidence.py`.
+
+### Phase H — literature RAG + collective-intelligence pathway scaffolding
+
+CodonTrace Genesis measurement-first kickoff toward CI then broader AI/OEE.
+**Does not** claim intelligence, collective intelligence, Tokyo Type 1 passed,
+or Avida replacement. ClaimGate unchanged in the strict direction. No version
+bump. Phase A–E default digest pins stay stable.
+
+- Literature RAG substrate (`codontrace.genesis.rag`): JSONL+markdown corpus
+  of citable digests (Nature MI 2021; Goldsby PNAS 2012; Goldsby/Ofria
+  messaging/germline; Channon 2024; Chromaria; Stanley open-endedness;
+  Szathmáry/Michod; JaxLife; arXiv 2607.09560; arXiv 2604.00810; bioRxiv
+  2023.03.15.532780) with has/lacks/next-experiment maps. Retriever:
+  `ingest_document`, `search_corpus(query, k)`, `cite_sources`. Deterministic
+  TF-IDF + feature-hash; replay digest. Example:
+  `examples/genesis_rag_query.py`.
+- Communication ablation harness (`run_communication_ablation_experiment`):
+  messaging on vs off, multi-seed, Cohen's *d*. Status
+  `measured_runtime_observation`. Does **not** set ClaimGate `ablation_result`.
+- Group-vs-individual effect-size campaign (research default `seed_count=12`,
+  smoke 2) with explicit `mean_delta` / `cohens_d` fields.
+- Goldsby 2012 task-switching cost hook on DoL metrics
+  (`TaskSwitchingCostConfig`, `apply_task_switching_cost_to_dol`).
+- `collective_intelligence_candidate` checklist printer lists missing flags
+  and never auto-sets them.
+- Docs: `docs/PHASE_H_CI_AI_PATH.md`, `docs/rag/`, honesty updates to
+  `docs/WHY_NOT_INTELLIGENCE_YET.md`.
+
 ## 0.3.0b3 — Phase F scientific gaps, Phase G materials, ClaimGate honesty (2026-09-11)
 
 Public beta bump from `0.3.0b2`. Includes Phase F+#9 scientific-gap measurement, Phase G+#10 named-materials substrate, and post-F/G suite ClaimGate honesty hygiene (#11 @ `0a5ce17`). Product name **CodonTrace Genesis**. Package/import remains `codontrace`. License `AGPL-3.0-or-later`. This is not an Avida replacement and does not prove intelligence, collective intelligence, AGI, OEE, Tokyo Type 1, realistic chemistry, or wet-lab equivalence.

{codontrace-0.3.0b3 → codontrace-0.3.0b4.dev0}/CITATION.cff

@@ -1,7 +1,7 @@
 cff-version: 1.2.0
 message: "If you use CodonTrace Genesis, please cite the versioned software release."
 title: "CodonTrace Genesis"
-version: "0.3.0b3"
+version: "0.3.0b4.dev0"
 authors:
   - family-names: "Jamei"
     given-names: "Parvaz"

{codontrace-0.3.0b3 → codontrace-0.3.0b4.dev0}/CONTRIBUTING.md

@@ -1,5 +1,15 @@
 # Contributing
 
+## Product naming
+
+- Public product name: **CodonTrace Genesis**
+- PyPI / import package: `codontrace`
+- Do not drop the Genesis qualifier in docs, PRs, issues, or papers
+- `codontrace.genesis` is a module path, not a second product name
+- Do not put internal naming-order lists in the README
+
+See [`STYLE.md`](STYLE.md).
+
 ## Development setup
 
 ```bash

{codontrace-0.3.0b3 → codontrace-0.3.0b4.dev0}/MANIFEST.in

@@ -7,6 +7,10 @@ include RELEASE_EVIDENCE.md
 include CITATION.cff
 include pyproject.toml
 recursive-include docs *.md
+recursive-include docs/hard_experiment_01 *.json
+recursive-include src/codontrace/genesis/rag/corpus *.jsonl
+recursive-include src/codontrace/genesis/rag/corpus *.md
+include src/codontrace/genesis/ilw/INTEGRATION_DAG.json
 recursive-include examples *.py
 recursive-include tests *.py
 recursive-include tools *.py
@@ -24,3 +28,5 @@ global-exclude .mypy_cache
 global-exclude .ruff_cache
 global-exclude *.html
 global-exclude *.patch
+global-exclude .grok_write_probe*.txt
+global-exclude .size_test_*.txt