feat(python): add trusted PyPI publishing and fix package metadata

.github/RELEASING-PYTHON.md

@@ -12,8 +12,8 @@ non-deterministic output; generated payload must never be added to Git.
 ## Validate without publishing
 
 Run **Python Standard Models Release** manually with version `1.0.0`, then run
-**Python Native Release** with version `4.1.0`. `workflow_dispatch` validates
-artifacts but cannot publish. The native run must pass Linux x86-64, Linux
+**Python Native Release** with version `4.1.0`. The default manual mode validates
+artifacts without publishing. The native run must pass Linux x86-64, Linux
 ARM64, macOS universal2, and Windows x86-64.
 
 For the Linux paths, maintainers can use `act` with rootless Podman and the

.github/act/python-models-standard.json

@@ -1,5 +1,6 @@
 {
   "inputs": {
-    "version": "1.0.0"
+    "version": "1.0.0",
+    "publish_existing_to_pypi": false
   }
 }

.github/act/python-native.json

@@ -1,5 +1,6 @@
 {
   "inputs": {
-    "version": "4.1.0"
+    "version": "4.1.0",
+    "publish_existing_to_pypi": false
   }
 }

.github/workflows/pages.yml

@@ -327,7 +327,10 @@ jobs:
         run: |
           set -euo pipefail
           mkdocs build --strict --config-file build/mkdocs/mkdocs.yml
-          rsync -a --delete --exclude '.git' --exclude '.git/' --exclude 'builds/' --exclude 'python/' build/mkdocs-site/ .gh-pages/
+          rsync -a --delete \
+            --exclude '.git' --exclude '.git/' \
+            --exclude 'builds/' --exclude 'python/simple/' \
+            build/mkdocs-site/ .gh-pages/
           mkdir -p .gh-pages/builds
           cp build/mkdocs-site/builds/index.html .gh-pages/builds/index.html
           cat > .gh-pages/.nojekyll <<EOF
@@ -337,7 +340,7 @@ jobs:
         shell: bash
         run: |
           set -euo pipefail
-          rm -rf .gh-pages/python
+          rm -rf .gh-pages/python/simple
           rm -f .gh-pages/.git
           if find .gh-pages -type l -print -quit | grep -q .; then
             echo 'Publication candidate contains a symbolic link.' >&2
@@ -391,7 +394,7 @@ jobs:
         shell: bash
         run: |
           set -euo pipefail
-          test ! -e candidate/python
+          test ! -e candidate/python/simple
           test ! -e candidate/.git
           if find candidate -type l -print -quit | grep -q .; then
             echo 'Publication candidate contains a symbolic link.' >&2
@@ -399,7 +402,8 @@ jobs:
           fi
           rsync -a --delete \
             --exclude '.git' --exclude '.git/' \
-            --exclude 'python' --exclude 'python/' candidate/ pages/
+            --exclude 'python/simple' --exclude 'python/simple/' \
+            candidate/ pages/
           cd pages
           git config user.name "github-actions[bot]"
           git config user.email "41898282+github-actions[bot]@users.noreply.github.com"

.github/workflows/python-models-standard-release.yml

@@ -11,6 +11,11 @@ on:
         required: true
         default: '1.0.0'
         type: string
+      publish_existing_to_pypi:
+        description: Publish the existing GitHub Release to PyPI
+        required: true
+        default: false
+        type: boolean
 
 permissions:
   contents: read
@@ -22,6 +27,7 @@ concurrency:
 jobs:
   build:
     name: Build and verify standard models
+    if: github.event_name == 'push' || inputs.publish_existing_to_pypi != true
     runs-on: ubuntu-latest
     outputs:
       version: ${{ steps.release.outputs.version }}
@@ -201,6 +207,99 @@ jobs:
           gh release edit "${RELEASE_TAG}" \
             --repo "${GITHUB_REPOSITORY}" --draft=false
 
+  publish-pypi:
+    name: Publish standard models to PyPI
+    if: >-
+      always() &&
+      ((github.event_name == 'push' && needs.publish.result == 'success') ||
+       (github.event_name == 'workflow_dispatch' && inputs.publish_existing_to_pypi))
+    needs: [build, publish]
+    runs-on: ubuntu-latest
+    environment: python-pypi
+    permissions:
+      attestations: read
+      contents: read
+      id-token: write
+
+    steps:
+      - name: Select PyPI release source
+        id: release
+        shell: bash
+        env:
+          BUILT_VERSION: ${{ needs.build.outputs.version }}
+          BUILT_TAG: ${{ needs.build.outputs.tag }}
+          REQUESTED_VERSION: ${{ inputs.version }}
+        run: |
+          set -euo pipefail
+          if [[ "${GITHUB_EVENT_NAME}" == 'push' ]]; then
+            version="${BUILT_VERSION}"
+            tag="${BUILT_TAG}"
+          else
+            version="${REQUESTED_VERSION}"
+            tag="python-models-standard@${version}"
+          fi
+          [[ "${version}" =~ ^(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)$ ]]
+          printf 'version=%s\ntag=%s\n' "${version}" "${tag}" >> "${GITHUB_OUTPUT}"
+
+      - name: Download current-run release candidate
+        if: github.event_name == 'push'
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
+        with:
+          name: python-models-standard-release-${{ steps.release.outputs.version }}
+          path: release
+
+      - name: Download existing GitHub Release
+        if: github.event_name == 'workflow_dispatch'
+        shell: bash
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+        run: |
+          set -euo pipefail
+          mkdir release
+          gh release download "${RELEASE_TAG}" \
+            --repo "${GITHUB_REPOSITORY}" --dir release \
+            --pattern '*.whl' --pattern '*.tar.gz' --pattern SHA256SUMS
+
+      - name: Verify and stage PyPI distributions
+        shell: bash
+        run: |
+          set -euo pipefail
+          cd release
+          sha256sum --check SHA256SUMS
+          awk '{print $2}' SHA256SUMS | LC_ALL=C sort > expected-files
+          [[ "$(wc -l < expected-files)" -eq 2 ]]
+          if grep -Ev '^[A-Za-z0-9_.+-]+(\.whl|\.tar\.gz)$' expected-files; then
+            echo 'SHA256SUMS contains an invalid distribution filename.' >&2
+            exit 1
+          fi
+          find . -maxdepth 1 -type f \( -name '*.whl' -o -name '*.tar.gz' \) \
+            -printf '%f\n' | LC_ALL=C sort > actual-files
+          diff -u expected-files actual-files
+          mkdir ../pypi
+          while IFS= read -r artifact; do
+            cp -- "${artifact}" ../pypi/
+          done < expected-files
+
+      - name: Verify GitHub build provenance
+        shell: bash
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+        run: |
+          set -euo pipefail
+          for artifact in pypi/*; do
+            gh attestation verify "${artifact}" \
+              --repo "${GITHUB_REPOSITORY}" \
+              --signer-workflow "${GITHUB_REPOSITORY}/.github/workflows/python-models-standard-release.yml" \
+              --source-ref "refs/tags/${RELEASE_TAG}"
+          done
+
+      - name: Publish distributions with PyPI Trusted Publishing
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: pypi/
+
   publish-index:
     name: Publish Python package index
     if: github.event_name == 'push'

.github/workflows/python-release.yml

@@ -11,6 +11,11 @@ on:
         required: true
         default: '4.1.0'
         type: string
+      publish_existing_to_pypi:
+        description: Publish the existing GitHub Release to PyPI
+        required: true
+        default: false
+        type: boolean
 
 permissions:
   contents: read
@@ -22,6 +27,7 @@ concurrency:
 jobs:
   prepare:
     name: Prepare versioned sources and sdist
+    if: github.event_name == 'push' || inputs.publish_existing_to_pypi != true
     runs-on: ubuntu-latest
     outputs:
       version: ${{ steps.release.outputs.version }}
@@ -419,6 +425,105 @@ jobs:
           gh release edit "${RELEASE_TAG}" \
             --repo "${GITHUB_REPOSITORY}" --draft=false
 
+  publish-pypi:
+    name: Publish native distributions to PyPI
+    if: >-
+      always() &&
+      ((github.event_name == 'push' && needs.publish.result == 'success') ||
+       (github.event_name == 'workflow_dispatch' && inputs.publish_existing_to_pypi))
+    needs: [prepare, assemble, publish]
+    runs-on: ubuntu-latest
+    environment: python-pypi
+    permissions:
+      attestations: read
+      contents: read
+      id-token: write
+
+    steps:
+      - name: Select PyPI release source
+        id: release
+        shell: bash
+        env:
+          BUILT_VERSION: ${{ needs.prepare.outputs.version }}
+          BUILT_TAG: ${{ needs.prepare.outputs.tag }}
+          REQUESTED_VERSION: ${{ inputs.version }}
+        run: |
+          set -euo pipefail
+          if [[ "${GITHUB_EVENT_NAME}" == 'push' ]]; then
+            version="${BUILT_VERSION}"
+            tag="${BUILT_TAG}"
+          else
+            version="${REQUESTED_VERSION}"
+            tag="python@${version}"
+          fi
+          [[ "${version}" =~ ^(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)$ ]]
+          printf 'version=%s\ntag=%s\n' "${version}" "${tag}" >> "${GITHUB_OUTPUT}"
+
+      - name: Download current-run release candidate
+        if: github.event_name == 'push'
+        uses: actions/download-artifact@d3f86a106a0bac45b974a628896c90dbdf5c8093 # v4
+        with:
+          name: python-native-release-${{ steps.release.outputs.version }}
+          path: release
+
+      - name: Download existing GitHub Release
+        if: github.event_name == 'workflow_dispatch'
+        shell: bash
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+        run: |
+          set -euo pipefail
+          mkdir release
+          gh release download "${RELEASE_TAG}" \
+            --repo "${GITHUB_REPOSITORY}" --dir release \
+            --pattern '*.whl' --pattern '*.tar.gz' --pattern SHA256SUMS
+
+      - name: Verify required model dependency on PyPI
+        run: >-
+          curl --fail --silent --show-error
+          https://pypi.org/pypi/radixor-models-standard/1.0.0/json
+          > /dev/null
+
+      - name: Verify and stage PyPI distributions
+        shell: bash
+        run: |
+          set -euo pipefail
+          cd release
+          sha256sum --check SHA256SUMS
+          awk '{print $2}' SHA256SUMS | LC_ALL=C sort > expected-files
+          [[ "$(wc -l < expected-files)" -eq 5 ]]
+          if grep -Ev '^[A-Za-z0-9_.+-]+(\.whl|\.tar\.gz)$' expected-files; then
+            echo 'SHA256SUMS contains an invalid distribution filename.' >&2
+            exit 1
+          fi
+          find . -maxdepth 1 -type f \( -name '*.whl' -o -name '*.tar.gz' \) \
+            -printf '%f\n' | LC_ALL=C sort > actual-files
+          diff -u expected-files actual-files
+          mkdir ../pypi
+          while IFS= read -r artifact; do
+            cp -- "${artifact}" ../pypi/
+          done < expected-files
+
+      - name: Verify GitHub build provenance
+        shell: bash
+        env:
+          GH_TOKEN: ${{ github.token }}
+          RELEASE_TAG: ${{ steps.release.outputs.tag }}
+        run: |
+          set -euo pipefail
+          for artifact in pypi/*; do
+            gh attestation verify "${artifact}" \
+              --repo "${GITHUB_REPOSITORY}" \
+              --signer-workflow "${GITHUB_REPOSITORY}/.github/workflows/python-release.yml" \
+              --source-ref "refs/tags/${RELEASE_TAG}"
+          done
+
+      - name: Publish distributions with PyPI Trusted Publishing
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: pypi/
+
   publish-index:
     name: Publish Python package index
     if: github.event_name == 'push'

README.md

@@ -30,7 +30,7 @@ It also retains the operational advantages of a compiled artifact model: predict
 For Python, one installation provides the native runtime and the separate
 standard package of 20 precompiled models:
 
-From PyPI, once publication is enabled:
+From PyPI:
 
 ```bash
 python -m pip install --only-binary=:all: radixor

docs/python/fast-track.md

@@ -19,9 +19,8 @@ standard-model package with 20 precompiled language models.
       --index-url https://leogalambos.github.io/Radixor/python/simple/ radixor
     ```
 
-PyPI publication is pending, and the GitHub index becomes live with the first
-Python releases. Until then, follow the source-checkout procedure on
-[Installation and Builds](installation.md).
+Both indexes provide the same released distributions. Use PyPI as the primary
+source or the GitHub index as an independent alternative.
 
 Radixor supports CPython 3.9 and newer. A JVM, Java dependency, and source
 dictionary are not required.

docs/python/installation.md

@@ -7,17 +7,12 @@ development or unsupported platforms.
 
 ## Install from PyPI
 
-PyPI is the intended primary index once the Radixor projects are approved and
-published there:
+PyPI is the primary package index:
 
 ```bash
 python -m pip install --only-binary=:all: radixor
 ```
 
-PyPI publication is not live yet. Until the `radixor` and
-`radixor-models-standard` project pages exist, this command cannot install the
-project.
-
 ## Install compiled packages from GitHub
 
 Python releases are published as immutable GitHub Release assets. A small
@@ -29,10 +24,7 @@ python -m pip install --only-binary=:all: \
 ```
 
 The index links directly to checksummed wheel assets in GitHub Releases; Pages
-does not duplicate the package files. It is not live until the first Python
-model and native releases have been published. This was verified before the
-initial release: the URL returned HTTP 404 and the repository contained no
-Python Release assets.
+does not duplicate the package files.
 
 Do not configure the GitHub index as an `--extra-index-url`: `pip` does not
 prioritize one index over another. Use it as the sole `--index-url`, as shown
@@ -93,10 +85,8 @@ gh attestation verify radixor-<version>-<wheel-tags>.whl \
 
 Python packages do **not** reuse the OpenPGP key configured for Java/Maven
 Central publications. Java's `SIGNING_KEY` and `SIGNING_PASSWORD` produce Maven
-signatures; Python currently uses release checksums plus GitHub's
-identity-bound build-provenance attestation. A future PyPI publication should
-use PyPI Trusted Publishing and its supported attestations rather than copying
-the Java signing mechanism.
+signatures; Python uses release checksums, GitHub identity-bound build
+provenance, and PyPI Trusted Publishing with supported digital attestations.
 
 ## Build through Gradle
 

docs/python/quick-start.md

@@ -26,9 +26,8 @@ source .venv/bin/activate       # Windows: .venv\Scripts\activate
       --index-url https://leogalambos.github.io/Radixor/python/simple/ radixor
     ```
 
-PyPI publication is pending. The GitHub option becomes live when the first
-Python GitHub Releases populate the Pages-backed package index. See
-[Installation and Builds](installation.md) for availability and source builds.
+Both indexes provide the same released distributions. See
+[Installation and Builds](installation.md) for provenance and source builds.
 
 The `radixor` wheel contains code. Its required
 `radixor-models-standard` dependency contains 20 precompiled models. The

python/README.md

@@ -17,12 +17,13 @@ same-input methodology, radixor won all **18 / 18** direct comparisons with
 PyStemmer 3.1.0 (Snowball's C `libstemmer`) in the published 2026-08-08 run.
 At batch size 100, the geometric-mean speedup was **1.67×**. The complete
 machine metadata and current results are in the [Python performance
-documentation](../docs/python/performance.md); benchmark implementation and
-fairness notes are in [`benchmarks/`](benchmarks/README.md).
+documentation](https://leogalambos.github.io/Radixor/python/performance/);
+benchmark implementation and fairness notes are in the
+[`benchmarks/` source directory](https://github.com/leogalambos/Radixor/blob/main/python/benchmarks/README.md).
 
 ## Installation
 
-From PyPI, once publication is enabled:
+From PyPI:
 
 ```bash
 python -m pip install --only-binary=:all: radixor
@@ -35,9 +36,9 @@ python -m pip install --only-binary=:all: \
   --index-url https://leogalambos.github.io/Radixor/python/simple/ radixor
 ```
 
-The GitHub command becomes usable after the first model and native releases
-populate that index. See the [installation guide](../docs/python/installation.md)
-for current availability and source-checkout builds.
+The GitHub index points to the same immutable model and native release assets.
+See the [installation guide](https://leogalambos.github.io/Radixor/python/installation/)
+for provenance and source-checkout builds.
 
 Wheels are provided for Linux, macOS, and Windows (Python 3.9+). The install
 also resolves the mandatory pure `radixor-models-standard` dependency
@@ -249,5 +250,6 @@ individual model versions recorded in the manifest.
 
 ## License
 
-The native/API package is BSD-3-Clause — see [LICENSE](LICENSE). Model data
-is separately licensed under CC BY-SA 3.0 in its packaged notices.
+The native/API package is BSD-3-Clause — see the
+[license text](https://github.com/leogalambos/Radixor/blob/main/python/LICENSE).
+Model data is separately licensed under CC BY-SA 3.0 in its packaged notices.

python/models-standard/README.md

@@ -17,8 +17,9 @@ tag version, and deterministically compiles all model resources there.
 Only gzip-compressed Radixor v7 (`.rxc`) tries are shipped. They are release
 artifacts, not checked-in repository files. Canonical textual dictionaries
 remain in the Radixor source repository and are not included in this wheel or
-source distribution. Model data is licensed under CC BY-SA 3.0; see
-`LICENSE-MODEL-DATA.txt` and the generated per-model notices.
+source distribution. Model data is licensed under CC BY-SA 3.0; see the
+[model-data license](https://github.com/leogalambos/Radixor/blob/main/python/models-standard/LICENSE-MODEL-DATA.txt)
+and the generated per-model notices.
 
 The checked-out directory is intentionally only a packaging skeleton and is
 not directly buildable as the complete data distribution. From the repository

python/models-standard/pyproject.toml

@@ -16,6 +16,12 @@ classifiers = [
     "Topic :: Text Processing :: Linguistic",
 ]
 
+[project.urls]
+Homepage = "https://github.com/leogalambos/Radixor"
+Documentation = "https://leogalambos.github.io/Radixor/python/"
+Repository = "https://github.com/leogalambos/Radixor"
+Issues = "https://github.com/leogalambos/Radixor/issues"
+
 [tool.setuptools]
 include-package-data = true
 

python/pyproject.toml

@@ -57,6 +57,12 @@ classifiers = [
     "Topic :: Text Processing :: Linguistic",
 ]
 
+[project.urls]
+Homepage = "https://github.com/leogalambos/Radixor"
+Documentation = "https://leogalambos.github.io/Radixor/python/"
+Repository = "https://github.com/leogalambos/Radixor"
+Issues = "https://github.com/leogalambos/Radixor/issues"
+
 [tool.maturin]
 python-source = "."
 module-name = "radixor._radixor"