profgen 0.0.1rc1__tar.gz

profgen-0.0.1rc1/.coveragerc

@@ -0,0 +1,28 @@
+# .coveragerc to control coverage.py
+[run]
+branch = True
+source = profgen
+# omit = bad_file.py
+
+[paths]
+source =
+    src/
+    */site-packages/
+
+[report]
+# Regexes for lines to exclude from consideration
+exclude_lines =
+    # Have to re-enable the standard pragma
+    pragma: no cover
+
+    # Don't complain about missing debug-only code:
+    def __repr__
+    if self\.debug
+
+    # Don't complain if tests don't hit defensive assertion code:
+    raise AssertionError
+    raise NotImplementedError
+
+    # Don't complain if non-runnable code isn't run:
+    if 0:
+    if __name__ == .__main__.:

profgen-0.0.1rc1/.github/workflows/ci.yml

@@ -0,0 +1,145 @@
+# GitHub Actions configuration **EXAMPLE**,
+# MODIFY IT ACCORDING TO YOUR NEEDS!
+# Reference: https://docs.github.com/en/actions
+
+name: tests
+
+on:
+  push:
+    # Avoid using all the resources/limits available by checking only
+    # relevant branches and tags. Other branches can be checked via PRs.
+    branches: [main]
+    tags: ['v[0-9]*', '[0-9]+.[0-9]+*']  # Match tags that resemble a version
+  pull_request:  # Run in every PR
+  workflow_dispatch:  # Allow manually triggering the workflow
+  schedule:
+    # Run roughly every 15 days at 00:00 UTC
+    # (useful to check if updates on dependencies break the package)
+    - cron: '0 0 1,16 * *'
+
+permissions:
+  contents: read
+
+concurrency:
+  group: >-
+    ${{ github.workflow }}-${{ github.ref_type }}-
+    ${{ github.event.pull_request.number || github.sha }}
+  cancel-in-progress: true
+
+jobs:
+  prepare:
+    runs-on: ubuntu-latest
+    outputs:
+      wheel-distribution: ${{ steps.wheel-distribution.outputs.path }}
+    steps:
+      - uses: actions/checkout@v4
+        with: {fetch-depth: 0}  # deep clone for setuptools-scm
+      - uses: actions/setup-python@v5
+        id: setup-python
+        with: {python-version: "3.13"}
+      - name: Run static analysis and format checkers
+        run: pipx run pre-commit run --all-files --show-diff-on-failure
+      - name: Type-check with mypy --strict
+        run: >-
+          pipx run --python '${{ steps.setup-python.outputs.python-path }}'
+          tox -e typecheck
+      - name: Build package distribution files
+        run: >-
+          pipx run --python '${{ steps.setup-python.outputs.python-path }}'
+          tox -e clean,build
+      - name: Record the path of wheel distribution
+        id: wheel-distribution
+        run: echo "path=$(ls dist/*.whl)" >> $GITHUB_OUTPUT
+      - name: Store the distribution files for use in other stages
+        # `tests` and `publish` will use the same pre-built distributions,
+        # so we make sure to release the exact same package that was tested
+        uses: actions/upload-artifact@v4
+        with:
+          name: python-distribution-files
+          path: dist/
+          retention-days: 1
+
+  test:
+    needs: prepare
+    strategy:
+      matrix:
+        python:
+        - "3.13"  # minimum supported (see requires-python in pyproject.toml)
+        platform:
+        - ubuntu-latest
+        - macos-latest
+        - windows-latest
+    runs-on: ${{ matrix.platform }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        id: setup-python
+        with:
+          python-version: ${{ matrix.python }}
+      - name: Retrieve pre-built distribution files
+        uses: actions/download-artifact@v4
+        with: {name: python-distribution-files, path: dist/}
+      - name: Run tests
+        run: >-
+          pipx run --python '${{ steps.setup-python.outputs.python-path }}'
+          tox --installpkg '${{ needs.prepare.outputs.wheel-distribution }}'
+          -- -rFEx --durations 10 --color yes  # pytest args
+      - name: Generate coverage report
+        run: pipx run coverage lcov -o coverage.lcov
+      - name: Upload partial coverage report
+        continue-on-error: true   # Coveralls not configured for this repo; don't fail CI on it
+        uses: coverallsapp/github-action@master
+        with:
+          path-to-lcov: coverage.lcov
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          flag-name: ${{ matrix.platform }} - py${{ matrix.python }}
+          parallel: true
+
+  finalize:
+    needs: test
+    runs-on: ubuntu-latest
+    steps:
+      - name: Finalize coverage report
+        continue-on-error: true   # Coveralls not configured for this repo; don't fail CI on it
+        uses: coverallsapp/github-action@master
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          parallel-finished: true
+
+  # Final version tags (e.g. v0.0.1) -> real PyPI. Pre-release tags (…rc…) are
+  # excluded here and routed to TestPyPI by the publish-testpypi job below.
+  publish:
+    needs: finalize
+    if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/') && !contains(github.ref, 'rc') }}
+    runs-on: ubuntu-latest
+    # Trusted Publishing (OIDC) — no API token/secret. The PyPI publisher is
+    # registered against this repo's `ci.yml` workflow and the `pypi` environment.
+    environment: pypi
+    permissions:
+      id-token: write   # required for the OIDC token exchange with PyPI
+      contents: read
+    steps:
+      - name: Retrieve pre-built distribution files
+        uses: actions/download-artifact@v4
+        with: {name: python-distribution-files, path: dist/}
+      - name: Publish to PyPI (Trusted Publishing)
+        uses: pypa/gh-action-pypi-publish@release/v1
+
+  # Pre-release tags (e.g. v0.0.1rc1) -> TestPyPI dry-run. Same OIDC mechanism,
+  # registered against this repo's `ci.yml` workflow and the `testpypi` environment.
+  publish-testpypi:
+    needs: finalize
+    if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/') && contains(github.ref, 'rc') }}
+    runs-on: ubuntu-latest
+    environment: testpypi
+    permissions:
+      id-token: write
+      contents: read
+    steps:
+      - name: Retrieve pre-built distribution files
+        uses: actions/download-artifact@v4
+        with: {name: python-distribution-files, path: dist/}
+      - name: Publish to TestPyPI (Trusted Publishing, dry-run)
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          repository-url: https://test.pypi.org/legacy/

profgen-0.0.1rc1/.gitignore

@@ -0,0 +1,73 @@
+# Real candidate CVs — contain PII (names, emails, phone numbers).
+# NEVER commit these. Anonymised, generated fixtures live elsewhere.
+tests/samples/
+
+# Local preview output of the synthetic CV corpus (regenerable).
+.fixture_preview/
+
+# Generated example profiles (regenerable via examples/build_example_profile.py).
+examples/output_profiles/
+
+# Temporary and binary files
+*~
+*.py[cod]
+*.so
+*.cfg
+!.isort.cfg
+!setup.cfg
+*.orig
+*.log
+*.pot
+__pycache__/*
+.cache/*
+.*.swp
+*/.ipynb_checkpoints/*
+.DS_Store
+
+# Project files
+.ropeproject
+.project
+.pydevproject
+.settings
+.idea
+.vscode
+tags
+
+# Package files
+*.egg
+*.eggs/
+.installed.cfg
+*.egg-info
+
+# Unittest and coverage
+htmlcov/*
+.coverage
+.coverage.*
+.tox
+junit*.xml
+coverage.xml
+.pytest_cache/
+
+# Build and docs folder/files
+build/*
+dist/*
+sdist/*
+docs/api/*
+docs/_rst/*
+docs/_build/*
+cover/*
+MANIFEST
+
+# Per-project virtualenvs
+.venv*/
+.conda*/
+.python-version
+# Generated by setuptools_scm
+src/profgen/_version.py
+
+# Local secrets / API keys — never commit
+.env
+.env.*
+
+# Private branded assets for `make profile` (confidential templates + style maps)
+local/

profgen-0.0.1rc1/.pre-commit-config.yaml

@@ -0,0 +1,29 @@
+ci:
+  autoupdate_schedule: quarterly
+
+default_language_version:
+  python: python3.13
+
+repos:
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.6.9        # pin to a tag; update as needed
+    hooks:
+      - id: ruff
+        args: [--fix, --exit-non-zero-on-fix]
+      - id: ruff-format
+
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.6.0        # pin to a tag; update as needed
+    hooks:
+      - id: check-added-large-files
+      - id: end-of-file-fixer
+      - id: trailing-whitespace
+      - id: check-merge-conflict
+
+
+  # If you want Black as well, uncomment this block and keep its rev pinned
+  # - repo: https://github.com/psf/black
+  #   rev: 24.10.0     # example pinned release
+  #   hooks:
+  #     - id: black
+  #       args: ["--line-length=100"]

profgen-0.0.1rc1/.readthedocs.yml

@@ -0,0 +1,24 @@
+# Read the Docs configuration file
+# https://docs.readthedocs.io/en/stable/config-file/v2.html
+
+version: 2
+
+build:
+  os: ubuntu-22.04
+  tools:
+    python: "3.13"
+
+sphinx:
+  configuration: docs/conf.py
+
+formats:
+  - pdf
+
+python:
+  install:
+    # Install the package with its `docs` extra (single source of truth in
+    # pyproject.toml). No docs/requirements.txt.
+    - method: pip
+      path: .
+      extras:
+        - docs

profgen-0.0.1rc1/AUTHORS.md

@@ -0,0 +1,3 @@
+# Contributors
+
+- Kevin Steptoe <kevin.steptoe@gmail.com>

profgen-0.0.1rc1/CHANGELOG.md

@@ -0,0 +1,170 @@
+# Changelog
+
+## Unreleased
+
+### De-brand & bring-your-own-template
+
+- Remove all brand-specific naming from the user-facing surface ahead of the
+  public PyPI release: the output is now described as a generic "standardised
+  Word profile". The pipeline, the no-invented-facts rule, employer anonymisation
+  (`Project N | <domain>`) and the grounding check are unchanged.
+- Render against **five logical roles** (`title`, `date_heading`, `body`,
+  `bullet`, `legal`) resolved to paragraph-style names through a *style map*. The
+  neutral `DEFAULT_STYLE_MAP` maps them to `Profile Title`, `Profile Date`,
+  `Normal`, `List Bullet` and `Profile Legal`.
+- **Bring your own template:** `convert --template my.docx --style-map my.toml`
+  renders against your branded document, with the TOML file mapping the five roles
+  to that document's own style names. `load_style_map` loads it; partial maps fall
+  back to the defaults. This lets a private or corporate template be applied
+  without it living in the package.
+- `convert`'s default output is now `<source-stem>_profile.docx`.
+- Add a `make profile CV=cv.pdf [OUT=..] [OFFLINE=1]` convenience target that
+  renders a branded profile using a gitignored `local/template.docx` and
+  `local/style-map.toml`; copy `examples/style-map.example.toml` to
+  `local/style-map.toml` and edit it. `local/` and `.env` are gitignored, keeping
+  confidential templates and API keys out of the repository.
+
+### Increment 0 — Tooling & dependencies foundation
+
+- Add runtime dependencies: `pydantic`, `python-docx`, `pdfplumber`, `anthropic`.
+- Add optional `pdf-fast` extra (`pymupdf`) for the swappable PDF backend.
+- Add `reportlab` and `mypy` to the `dev` extra; fold `docs` + `pdf-fast` in.
+- Wire `mypy --strict` (scoped to `src/`) into the tooling.
+- Replace the single-command CLI with a Click group exposing `convert` and
+  `make-template` (stubs at this stage); add the `cv-formatter` entry-point alias.
+- Remove the scaffold's placeholder `api.py` (superseded by the forthcoming
+  `pipeline.py`).
+- Add an offline guard (socket-blocking autouse fixture) and CLI smoke tests;
+  enforce `mypy --strict` in CI via a `typecheck` tox env.
+- Add a synthetic, fully-fictitious CV fixture corpus (`tests/fixtures/`) with a
+  `cv_corpus` test fixture; record the profile-format decisions in the spec.
+
+### Increment 1 — The Candidate data contract
+
+- Add the Pydantic v2 `Candidate` model and `EmploymentEntry` / `ProjectEntry` /
+  `EducationEntry` sub-models — the single contract shared by the whole pipeline.
+  Every field defaults to `"Not stated"` / `[]`; nothing is required.
+- Export `Candidate` (and sub-models, `NOT_STATED`, `is_not_stated`) from the
+  package root; enable the `pydantic.mypy` plugin.
+
+### Increment 2 — Extractors (stage 2, fully offline)
+
+- Add `profgen.extractors`: `ExtractedDocument` + `Extractor` protocol, plain-text
+  (UTF-8 with CP1252 fallback), `.docx` (python-docx, paragraphs + tables in
+  document order) and `.pdf` (pdfplumber default, pymupdf swappable via the
+  `pdf-fast` extra) backends, with extension-based `extract()` dispatch.
+- Extractors do no interpretation — verbatim text only; `normalized_text`
+  collapses whitespace for grounding-friendly matching.
+
+### Increment 3 — LLM structuring layer (stage 3)
+
+- Add `profgen.llm`: the `StructuringClient` protocol — the single interface that
+  turns an `ExtractedDocument` into a typed `Candidate` — with two implementations
+  behind it.
+- Add `ClaudeStructuringClient`, the production client. It calls Claude
+  (`claude-sonnet-4-6`) with forced, strict tool use against one tool whose schema
+  is the `Candidate` model, `temperature=0` and `max_tokens=8192`. Constructing it
+  touches no network or environment; the Anthropic SDK client is built lazily on
+  first use (or injected for testing). It is wired but, by design, never exercised
+  in CI.
+- Add `HeuristicStructuringClient`, a deterministic, network-free parser that reads
+  the synthetic corpus by section headers. It is the `--offline` and test path and
+  invents nothing — absent fields stay at the `"Not stated"` / `[]` defaults.
+- Add `StructuringError` for transport, API and malformed-response failures, and
+  the `SYSTEM_PROMPT` / `TOOL_NAME` prompt constants. The system prompt encodes the
+  no-invented-facts rule, the `"Not stated"` sentinel, the no-derived-fields rule,
+  British English, and the semiconductor skill bucketing.
+
+### Increment 4 — Grounding check & review report (stages 4 & 6)
+
+- Add `profgen.pipeline` with four pure, deterministic, LLM-independent functions,
+  all re-exported from the package root.
+- `check_grounding` — the anti-hallucination guard. For each groundable entity
+  (company, tool, certification, institution, project name, project domain) it
+  confirms the value appears in the verbatim stage-2 text via a
+  whitespace-collapsed, case-folded substring match, and returns a British-English
+  note for every entity that does not. Honestly-absent (`"Not stated"` / `[]`)
+  values are never flagged; company is verified for grounding even though it is
+  never rendered (SPEC §14.1).
+- `collect_missing_information` — lists every scalar left at the `"Not stated"`
+  sentinel and every empty list, in model field order.
+- `annotate_candidate` — the composition seam: runs both of the above and returns
+  a copy of the `Candidate` with the pipeline-populated `source_confidence_notes`
+  and `missing_information` fields set, without mutating the input.
+- `render_review_report` — renders the deterministic `*.review.md` body (missing
+  information / assumptions made / items to verify before customer submission) from
+  an already-annotated candidate, in British English. (End-to-end `run_pipeline`
+  orchestration remains a later increment.)
+
+### Increment 5 — Word rendering & starter template (stage 5)
+
+- Add `profgen.template.word_renderer`: the deterministic, fully-offline Word
+  renderer. `render_profile(candidate, *, template_path=None, style_map=None)`
+  builds the `python-docx` `Document`; `write_profile(candidate, output_path, *,
+  template_path=None, style_map=None)` renders and saves it.
+- Use the **style-donor** approach (SPEC §8): keep the donor's header, footer and
+  fonts, and write variable-length content programmatically against five logical
+  roles (`title`, `date_heading`, `body`, `bullet`, `legal`, exposed as the
+  `ROLE_TITLE/ROLE_DATE/ROLE_BODY/ROLE_BULLET/ROLE_LEGAL` constants) — no
+  Jinja-style loops in the `.docx`. A *style map* resolves each role to a concrete
+  paragraph-style name; the neutral `DEFAULT_STYLE_MAP` maps them to
+  `Profile Title`, `Profile Date`, `Normal`, `List Bullet` and `Profile Legal`. A
+  house style is applied by supplying a donor plus a style map that points the
+  roles at that document's own style names; an absent style falls back to the
+  default.
+- Add `make_template(path)`, which writes a starter `.docx` carrying the default
+  named styles, and wire the `make-template` CLI command to it (prints
+  `Wrote starter template to <path>`).
+- Enforce **employer anonymisation** (§14.1): company names are never rendered;
+  experience is shown as `Project N | <domain>`. Enforce **no derived fields**
+  (§14.3): the skills table's "Years Experience" column renders `"Not stated"`
+  for every row. Honestly render `"Not stated"` where the source was silent.
+- All rendered text is British English. (`convert` orchestration remains a stub
+  until Increment 6.)
+
+### Increment 6 — End-to-end `convert` orchestration (offline path)
+
+- Add `profgen.run_pipeline(source, output, *, offline=False, template_path=None,
+  style_map=None, client=None) -> PipelineResult` — the full orchestration
+  (SPEC §3): extract →
+  structure → annotate/ground → render `.docx` → write the sibling `*.review.md`.
+  An injected `client` wins; otherwise `--offline` selects the deterministic
+  `HeuristicStructuringClient` and the default path selects `ClaudeStructuringClient`
+  (`ANTHROPIC_API_KEY` required, reached only on first `structure`). The offline /
+  injected path is fully deterministic and performs no network access.
+- Add the frozen `PipelineResult` dataclass (`candidate`, `profile_path`,
+  `review_path`, `needs_verification`); both re-exported from the package root.
+  `needs_verification` is `True` when any entity was ungrounded or any field was
+  left unstated.
+- Wire the `convert` CLI command to `run_pipeline`. It writes the profile to
+  `--output` (default `<source-stem>_profile.docx` in the CWD) plus the sibling
+  `*.review.md` (`out.docx` → `out.review.md`), and prints a manual-verification
+  warning to stderr when `needs_verification` is set. Missing `ANTHROPIC_API_KEY`
+  or an online failure is reported as a friendly message advising `--offline`.
+
+### Increment 7 — Examples, real-path smoke, README & acceptance hardening
+
+- Add `examples/build_example_profile.py` — the runnable offline example (SPEC §12
+  criterion 4). It drives `run_pipeline` on the bundled
+  `examples/input_cvs/sample_cv.txt` with `offline=True`, needs **no API key and
+  makes no network call**, and writes `sample_profile.docx` plus its sibling
+  `sample_profile.review.md` into the gitignored `examples/output_profiles/`. The
+  run is deterministic.
+- Add `examples/smoke_real_path.py` — the only place that exercises the real Claude
+  path (`offline=False`), double-guarded behind `PROFGEN_REAL_SMOKE=1` **and**
+  `ANTHROPIC_API_KEY`. When either is absent it prints a one-line message and exits
+  `0` without touching the network, so it is inert in the offline suite and **never
+  runs in CI** by design.
+- Rewrite the placeholder PyScaffold `README.md` into a real description: install,
+  the `convert` / `make-template` usage (incl. `--offline` and the `cv-formatter`
+  alias), and the no-invented-facts guarantee.
+- Add the Sphinx User Guide and finalise the docs sweep; confirm the API reference
+  (autodoc) still builds cleanly.
+- Final acceptance sweep against SPEC §12: all eight criteria hold; `ruff`,
+  `mypy --strict` (scoped to `src/`) and offline `pytest` are green. The SPEC §13
+  open seams (OCR for scanned PDFs; a fixed-layout `docxtpl` placeholder renderer)
+  are deliberately left unbuilt.
+
+<!-- No tagged release yet. The project is pre-release; setuptools_scm derives
+     the version from git (currently 0.0.x.devN). The first real release will be
+     tagged once the pipeline is functional. -->

profgen-0.0.1rc1/CONTRIBUTING.md

@@ -0,0 +1,31 @@
+# Contributing
+
+Contributions are welcome! Please follow these guidelines:
+
+## Development Setup
+
+1. Clone the repository
+2. Install development dependencies: `make dev`
+3. Install pre-commit hooks: `pre-commit install`
+
+## Code Style
+
+This project uses [Ruff](https://docs.astral.sh/ruff/) for linting and formatting.
+Run `make lint` to check and `make format` to auto-format.
+
+## Testing
+
+Run tests with `make test` or `pytest` directly.
+
+## Pull Requests
+
+1. Fork the repository
+2. Create a feature branch
+3. Make your changes
+4. Run tests and linting
+5. Submit a pull request
+
+## Issue Reporting
+
+Please use the issue tracker to report bugs or request features.
+Include as much detail as possible.

profgen-0.0.1rc1/LICENSE.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2026 Kevin Steptoe
+
+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.

profgen-0.0.1rc1/MANIFEST.in

@@ -0,0 +1,9 @@
+# Keep confidential, repo-private files out of the published sdist.
+# setuptools_scm includes every git-tracked file by default, so the private
+# design docs and agent definitions are explicitly excluded here.
+exclude cv_formatter_SPEC.md
+exclude cv_formatter_IMPLEMENTATION_PLAN.md
+exclude cv_formatter_ROADMAP.md
+exclude CLAUDE.md
+prune .claude
+prune tests/samples