Complete Python 3 rewrite with native prediction workflow and regression parity

.github/workflows/python.yml

@@ -0,0 +1,61 @@
+name: CopraRNA 3
+on:
+  push:
+  pull_request:
+  workflow_dispatch:
+
+env:
+  OPENBLAS_CORETYPE: Haswell
+
+jobs:
+  python:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        include:
+          - python: '3.11'
+            biopython: '1.84'
+          - python: '3.12'
+            biopython: '1.88'
+          - python: '3.13'
+            biopython: '1.88'
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python }}
+      - run: python -m pip install '.[test,build]' 'biopython==${{ matrix.biopython }}'
+      - run: python -m pytest -q
+      - run: python -m build
+      - name: Check installed wheel outside the source directory
+        shell: bash
+        run: |
+          python -m venv --system-site-packages /tmp/coprarna-wheel
+          /tmp/coprarna-wheel/bin/python -m pip install --force-reinstall --no-deps dist/*.whl
+          cd /tmp
+          /tmp/coprarna-wheel/bin/coprarna --version
+          /tmp/coprarna-wheel/bin/coprarna helpers
+          /tmp/coprarna-wheel/bin/coprarna helper archive-readme
+
+  scientific:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: mamba-org/setup-micromamba@v2
+        with:
+          environment-file: environment.yml
+          cache-environment: true
+      - name: Install Python package and browser
+        shell: bash -el {0}
+        run: |
+          python -m pip install --no-deps .
+          python -m pip install ".[web]"
+          python -m playwright install chromium --only-shell
+      - name: Native scientific workflow and browser integration
+        shell: bash -el {0}
+        env:
+          COPRARNA_PIPELINE_TESTS: '1'
+          COPRARNA_BROWSER_TESTS: '1'
+          COPRARNA_REFERENCE_TESTS: '1'
+        run: python -m pytest -q

.gitignore

@@ -1 +1,9 @@
 /.project
+
+__pycache__/
+*.py[cod]
+.pytest_cache/
+build/
+dist/
+*.egg-info/
+core_genome_archive/metadata/*.lock

AGENTS.md

@@ -0,0 +1,67 @@
+# CopraRNA 3 architecture and release policy
+
+[Issue #4](https://github.com/BackofenLab/CopraRNA/issues/4) prepares version 3.0.0
+from `issue-1-python3-bugfixes`. Its explicit removal of the 2.1 code base supersedes
+the earlier instruction to retain original source files. Historical sources remain
+in Git history; production and tests must never invoke them. Keep substantial
+review changes in separate commits.
+
+## Architecture
+
+- `pipeline.py`, `cli.py`: configuration, isolated batches, cleanup and genome CLI.
+- `organisms.py`: on-demand NCBI assembly resolution, versioned metadata, cache
+  validation and duplicate-assembly rejection; no static organism allow-list.
+- `homology.py`, `genome.py`, `ncbi.py`, `annotation.py`: shared compressed genomes,
+  streaming GenBank parsing, short job-local native species codes and clustering.
+- `intarna.py`, `parameters.py`: IntaRNA >=3 / IntaRNAsTar, native parameter files,
+  gzipped interactions and removal of near-full-length perfect helices.
+- `energy.py`, `combination.py`, `numerics.py`, `statistics.py`: validated GEV and
+  comparative statistics; preserve the validated NumPy/SciPy and BLAS kernel pins.
+- `conservation.py`, `alignments.py`: consensus matrices and Jalview exports.
+- `plots.py`, `phylogeny.py`: PDF/SVG scientific figures and tree rendering.
+- `david.py`, `enrichment.py`, `webserver.py`: optional service and web exports.
+- `io.py`: shared compressed/plain I/O and protected input/output namespaces.
+- `helpers.py`: active Python helper contracts; no original-script aliases or
+  defect-emulation mode. `resources.py` provides current report resources only.
+
+## Documentation
+
+Document each new/changed public function with its purpose, input/output contract,
+file side effects and failure behavior where relevant. For scientific algorithms,
+explain coordinate systems, matrix dimensions, statistical conventions and retained
+oracle differences. Comment non-obvious numerical choices, native interfaces and
+cache publication/lifetime rules. Record release-facing changes and rationale in
+`ChangeLog.md`; keep `README.md` consistent with CLI and packaging.
+
+## Validation
+
+Retain exact independently captured scientific fixtures for unchanged behavior.
+Their provenance and checksums live in `tests/fixtures`; do not regenerate expected
+numbers from the Python implementation. Compare R-data objects semantically.
+Test changed behavior against explicit contracts and real native integrations.
+No legacy source/interpreter environments or generated availability catalogs belong
+in the release. The current source version, package metadata and option-file
+version must agree.
+
+Native integration uses `COPRARNA_PIPELINE_TESTS=1`; reference-genome extraction
+uses `COPRARNA_REFERENCE_TESTS=1`; optional browser integration uses
+`COPRARNA_BROWSER_TESTS=1`. CI checks Python 3.11–3.13, builds and installed wheels.
+Refresh the 17 NCBI reference records with `python -m tools.refresh_core_genomes`;
+verify exact identities, checksums and annotation dates in the archive manifest.
+Live external-service availability is not certified by offline request tests.
+Check new dependency/UI releases before changing pins.
+
+## Invariants
+
+Preserve the user's original FASTA and native parameter file, including inputs
+named `input_sRNA.fa`. When adding outputs, update `io._JOB_OUTPUTS`; preserve
+cleanup destinations that resolve to original inputs. Every independent native
+entry point retains its version check and detected-version cache signature.
+
+Keep genomes compressed and reuse shared paths directly. Cache metadata pins exact
+assembly/sequence versions until explicitly refreshed. Publish only completed
+validated downloads/resolutions, under accession locks; failed refreshes preserve
+prior entries. A chromosome and plasmid from the same assembly must never count
+as independent comparative organisms. Scientific report labels are unversioned;
+job metadata and cache paths preserve exact versions. SVG/PDF remain primary
+figure formats; no PostScript/Ghostscript dependency is permitted.