sartor 0.0.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,9 @@
1
+ __pycache__/
2
+ *.py[cod]
3
+ .venv/
4
+ dist/
5
+ build/
6
+ *.egg-info/
7
+ .DS_Store
8
+ ._*
9
+ exports/
sartor-0.0.1/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 Michael Borck
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
sartor-0.0.1/PKG-INFO ADDED
@@ -0,0 +1,54 @@
1
+ Metadata-Version: 2.5
2
+ Name: sartor
3
+ Version: 0.0.1
4
+ Summary: Privacy-first résumé tailoring: your evidence in, a tailored résumé out — every claim traceable, nothing invented. Bring your own model (OpenAI-compatible, Ollama, Anthropic).
5
+ Project-URL: Homepage, https://github.com/michael-borck/sartor
6
+ Project-URL: Repository, https://github.com/michael-borck/sartor
7
+ Project-URL: Issues, https://github.com/michael-borck/sartor/issues
8
+ Author-email: Michael Borck <michael.borck@curtin.edu.au>
9
+ License-Expression: MIT
10
+ License-File: LICENSE
11
+ Keywords: ats,career,cli,cv,llm,privacy,resume,tailoring
12
+ Classifier: Development Status :: 1 - Planning
13
+ Classifier: Environment :: Console
14
+ Classifier: Intended Audience :: End Users/Desktop
15
+ Classifier: License :: OSI Approved :: MIT License
16
+ Classifier: Operating System :: OS Independent
17
+ Classifier: Programming Language :: Python :: 3
18
+ Classifier: Programming Language :: Python :: 3.10
19
+ Classifier: Programming Language :: Python :: 3.11
20
+ Classifier: Programming Language :: Python :: 3.12
21
+ Classifier: Programming Language :: Python :: 3.13
22
+ Classifier: Topic :: Text Processing
23
+ Requires-Python: >=3.10
24
+ Description-Content-Type: text/markdown
25
+
26
+ # Sartor
27
+
28
+ > *sartor (Latin): a tailor.*
29
+
30
+ Privacy-first résumé tailoring. Give Sartor your real experience and one
31
+ job advertisement; it selects, reorders and rewrites only what your
32
+ evidence supports, then renders a finished résumé in Markdown, HTML, PDF
33
+ and DOCX — plus a pitch script for [Narra](https://github.com/michael-borck/narra)
34
+ and a deck for [SlideStream](https://github.com/michael-borck/slide-stream).
35
+
36
+ Every claim in the output traces back to a source you supplied.
37
+ Unsupported claims are flagged, never quietly invented. No account, no
38
+ telemetry; bring your own model (any OpenAI-compatible endpoint —
39
+ including local Ollama — or Anthropic).
40
+
41
+ **Status:** pre-code. Product definition in [docs/BRIEF.md](docs/BRIEF.md),
42
+ design in [docs/SPEC.md](docs/SPEC.md).
43
+
44
+ Planned interface:
45
+
46
+ ```bash
47
+ pip install sartor
48
+ sartor init my-profile
49
+ sartor ingest resume.pdf publications.md
50
+ sartor tailor --job acme-ad.txt --theme ats-plain --out exports/acme/
51
+ sartor serve # local web UI
52
+ ```
53
+
54
+ MIT © Michael Borck.
sartor-0.0.1/README.md ADDED
@@ -0,0 +1,29 @@
1
+ # Sartor
2
+
3
+ > *sartor (Latin): a tailor.*
4
+
5
+ Privacy-first résumé tailoring. Give Sartor your real experience and one
6
+ job advertisement; it selects, reorders and rewrites only what your
7
+ evidence supports, then renders a finished résumé in Markdown, HTML, PDF
8
+ and DOCX — plus a pitch script for [Narra](https://github.com/michael-borck/narra)
9
+ and a deck for [SlideStream](https://github.com/michael-borck/slide-stream).
10
+
11
+ Every claim in the output traces back to a source you supplied.
12
+ Unsupported claims are flagged, never quietly invented. No account, no
13
+ telemetry; bring your own model (any OpenAI-compatible endpoint —
14
+ including local Ollama — or Anthropic).
15
+
16
+ **Status:** pre-code. Product definition in [docs/BRIEF.md](docs/BRIEF.md),
17
+ design in [docs/SPEC.md](docs/SPEC.md).
18
+
19
+ Planned interface:
20
+
21
+ ```bash
22
+ pip install sartor
23
+ sartor init my-profile
24
+ sartor ingest resume.pdf publications.md
25
+ sartor tailor --job acme-ad.txt --theme ats-plain --out exports/acme/
26
+ sartor serve # local web UI
27
+ ```
28
+
29
+ MIT © Michael Borck.
@@ -0,0 +1,85 @@
1
+ # Sartor
2
+
3
+ Sartor (Latin: tailor) is a privacy-first tool that tailors your real
4
+ experience to one job advertisement and renders a finished résumé — every
5
+ claim traceable to evidence you supplied, nothing invented.
6
+
7
+ ## Replaces
8
+
9
+ Hand-editing a Word résumé per application, and pasting a résumé plus a job
10
+ ad into a cloud chatbot that quietly embellishes it. It complements
11
+ Career Compass (exploration: *which* job?) by doing production
12
+ (*this* job, *this* document).
13
+
14
+ ## Main jobs
15
+
16
+ 1. Build a master profile from source material: an existing résumé
17
+ (PDF, DOCX, HTML, MD, TXT), plus extra evidence — publications,
18
+ projects, performance notes, framework documents.
19
+ 2. Tailor to a target: paste or load a job description; Sartor selects,
20
+ reorders and rewrites only supported material, reports keyword/skill
21
+ coverage and gaps, and flags anything it cannot evidence.
22
+ 3. Render and hand off: styled MD, HTML, PDF and DOCX from one canonical
23
+ model, plus a pitch script for Narra and a Markdown deck for
24
+ SlideStream.
25
+
26
+ ## Data
27
+
28
+ - MasterProfile: canonical JSON Resume (basics, work, education, skills,
29
+ projects, publications), human-editable, the single source of truth.
30
+ - Source: one ingested file, with extracted text and stable span ids.
31
+ - EvidenceLink: output bullet → source span(s) that support it.
32
+ - JobTarget: parsed advertisement (must-haves, nice-to-haves, keywords,
33
+ framework skills).
34
+ - TailoredResume: a MasterProfile subset + rewrites + coverage report.
35
+ - Theme: named style (ats-plain, modern, academic) applied at render.
36
+ - ExportBundle: rendered files + `manifest.json` (provenance, model,
37
+ inputs, timestamps).
38
+
39
+ ## Shape
40
+
41
+ Python package first: `pip install sartor`, a `sartor` CLI, and
42
+ `sartor serve` (localhost web UI). Desktop is a Tauri v2 shell running
43
+ `sartor serve` as a sidecar — the proven Narra/SlideStream pattern, not a
44
+ new Electron app. Cross-platform (macOS, Windows, Linux). Local data in
45
+ a plain project folder the user owns. No accounts, no telemetry.
46
+
47
+ LLM providers — exactly two:
48
+
49
+ - OpenAI-compatible (`/v1/chat/completions`), key optional. Ollama is
50
+ just this provider at `http://localhost:11434/v1`.
51
+ - Anthropic (native Messages API).
52
+
53
+ The only network egress is to the configured LLM, and (phase 2, opt-in)
54
+ to a search provider — search queries are derived terms only, never
55
+ résumé text (the Career Compass rule).
56
+
57
+ ## System hooks
58
+
59
+ CLI. File-based handoffs out: `pitch.md` → Narra (podcast),
60
+ `deck.md` → SlideStream (PPTX, narrated video). JSON Resume in/out, so
61
+ Career Compass can consume a profile later. No direct app-to-app calls
62
+ in v1.
63
+
64
+ ## Reference
65
+
66
+ Career Compass for privacy posture and provider UX; Narra and
67
+ SlideStream for packaging (core + CLI + serve + Tauri); Winnow and
68
+ CiteSight for claim-level provenance and verification.
69
+
70
+ ## Look
71
+
72
+ Documents, not app chrome, carry the design: three résumé themes in v1
73
+ (`ats-plain` default — single column, parseable; `modern`; `academic`).
74
+ Web UI follows the calm, plain-language style of the Buddy apps.
75
+
76
+ ## Not in version one
77
+
78
+ - PPTX, Reveal.js, podcast or video rendering (handoff files only).
79
+ - SFIA / O*NET / ESCO mapping and web-search grounding (phase 2; SFIA
80
+ data is licence-restricted, so it is user-supplied or looked up,
81
+ never bundled).
82
+ - Job-board scraping, auto-apply, LinkedIn import, accounts, sync,
83
+ cover letters (Career Compass already drafts those).
84
+ - Inventing metrics, dates or skills. Unsupported claims are flagged
85
+ or dropped — this is a feature, not a limitation.
@@ -0,0 +1,129 @@
1
+ # Sartor — Specification (draft)
2
+
3
+ Status: draft, pre-code. The one-page product definition is
4
+ [BRIEF.md](BRIEF.md); this file records the design reasoning, pipeline
5
+ and phasing so implementation stories can be cut from it.
6
+
7
+ ## One line
8
+
9
+ `sartor tailor profile/ --job ad.txt --theme ats-plain --out exports/acme/`
10
+ — a tailored résumé in MD, HTML, PDF and DOCX, a coverage report, and
11
+ handoff files for podcast and video, without your data leaving the
12
+ machine (unless you point it at a cloud LLM).
13
+
14
+ ## Why a new app, not a Career Compass feature
15
+
16
+ Career Compass answers *which career, what gaps, what next* and is
17
+ deliberately session-only: nothing persists, export-don't-save. Tailoring
18
+ is the opposite shape: a durable master profile reused across many
19
+ applications, and a document-rendering engine (themes, PDF/DOCX fidelity,
20
+ ATS checks) that Compass has no reason to grow. Splitting keeps Compass a
21
+ teaching tool for students and Sartor a production tool for applicants.
22
+ The seam between them is a file: JSON Resume.
23
+
24
+ ## Why Python, not Node
25
+
26
+ - Document I/O is the core workload: `pypdf`, `python-docx`,
27
+ `python-pptx`, HTML→PDF (`weasyprint`), Markdown and templating are
28
+ all mature in Python; Node equivalents exist but are thinner.
29
+ - The shipping pattern is already proven twice in this codebase family:
30
+ Python core + Typer CLI + optional FastAPI `serve` + Tauri v2 sidecar
31
+ desktop (Narra, SlideStream). Desktop therefore costs a copied shell,
32
+ not a second codebase. Electron was considered and rejected: it would
33
+ duplicate Compass's stack to gain nothing for document rendering.
34
+ - PyPI name `sartor` verified available 2026-11-22 (also npm and
35
+ `github.com/michael-borck/sartor`). Alternatives, also verified free
36
+ that day: `tailor-cv`, `bespoke-cv`, `vericv`, `refit-cv`.
37
+
38
+ ## Providers (exactly two)
39
+
40
+ | Provider | Endpoint | Key |
41
+ |---|---|---|
42
+ | `openai-compatible` | any `/v1/chat/completions` (OpenAI, OpenRouter, Groq, LM Studio, …) | optional |
43
+ | `anthropic` | Anthropic Messages API | required |
44
+
45
+ Ollama is not a third provider: it is `openai-compatible` with
46
+ `base_url: http://localhost:11434/v1` and the key ignored. Config lives
47
+ in `~/.sartor.yaml` or `./sartor.yaml`, keys by `${ENV_VAR}` reference
48
+ only, never stored in the browser (Narra rule). A `sartor doctor`
49
+ command tests the connection and lists models before any generation.
50
+
51
+ ## Inputs
52
+
53
+ - Existing résumé: PDF, DOCX, HTML, MD, TXT, or JSON Resume (lossless).
54
+ - Additional evidence: any number of the same formats — publication
55
+ lists, project write-ups, transcripts, reference letters.
56
+ - Job target: pasted text, file, or (phase 2) a URL fetched and cleaned.
57
+ - Source material: style guides ("address these selection criteria"),
58
+ framework extracts (e.g. user-supplied SFIA skill definitions).
59
+
60
+ ## Pipeline
61
+
62
+ 1. **Ingest** — extract text per file; assign stable span ids
63
+ (`src://resume.pdf#s12`). Everything stays in the project folder.
64
+ 2. **Distil** — LLM drafts a MasterProfile (JSON Resume) from the
65
+ sources. The user reviews and edits it; this human-owned file is the
66
+ truth for all later runs. Re-ingesting never silently overwrites it.
67
+ 3. **Parse target** — LLM extracts a JobTarget: must-have skills,
68
+ nice-to-haves, keywords, seniority, explicit criteria.
69
+ 4. **Tailor** — select and order profile entries against the target;
70
+ rewrite bullets for relevance and keyword match. Rewrites may
71
+ rephrase, never add facts. Each generated bullet must cite the
72
+ EvidenceLink(s) it rests on.
73
+ 5. **Verify** — deterministic checks first (every bullet has evidence;
74
+ dates, titles and numbers in the output appear in the sources; no
75
+ skill named that no source supports), then an optional LLM
76
+ entailment pass. Failures are reported per bullet: `supported`,
77
+ `weak`, `unsupported → dropped/flagged`. This is the Winnow/CiteSight
78
+ pattern applied to résumés, and it is the product's differentiator.
79
+ 6. **Report** — coverage: which target requirements are evidenced,
80
+ which are gaps (with "add evidence or accept the gap", never
81
+ "invent something"), ATS keyword table.
82
+ 7. **Render** — from the one TailoredResume model:
83
+ - `resume.md`, `resume.html` (themed), `resume.pdf`
84
+ (HTML + WeasyPrint, so HTML and PDF never diverge), `resume.docx`
85
+ (python-docx), `resume.json` (JSON Resume).
86
+ - Handoffs: `pitch.md` (first-person narrative script → Narra
87
+ podcast), `deck.md` (one claim per slide → SlideStream PPTX/MP4).
88
+ - `manifest.json`: inputs with hashes, provider/model, theme,
89
+ verification results — a run is auditable and replayable.
90
+
91
+ ## Frameworks and web search (phase 2)
92
+
93
+ - SFIA, O*NET, ESCO, AQF mapping of both profile and target, so skills
94
+ share a vocabulary ("SFIA PROG level 4"). Career Compass's F6 notes
95
+ apply. SFIA content is licence-restricted: user supplies it, or codes
96
+ are looked up live; nothing is bundled or redistributed.
97
+ - Search grounding (company/role context, current terminology) follows
98
+ the Study Buddy → Career Compass chain: DuckDuckGo default,
99
+ Brave/Serper/SearXNG optional, automatic fallback. Only derived
100
+ queries leave the device; résumé text never does.
101
+
102
+ ## Interfaces
103
+
104
+ - Library: `from sartor import core` (Narra-style), all logic here.
105
+ - CLI: `init`, `ingest`, `profile`, `tailor`, `render`, `verify`,
106
+ `doctor`, `serve`.
107
+ - Web UI (`serve`): upload → review profile → paste job → review
108
+ tailoring with evidence flags inline → download bundle.
109
+ - Desktop: Tauri v2 shell + Python sidecar via `uv` (copied from
110
+ Narra's `desktop/`), installers on `desktop-v*` tags.
111
+
112
+ ## Phasing
113
+
114
+ - **M1** — core + CLI: ingest, distil, tailor, verify, MD/HTML/PDF/DOCX
115
+ render, two providers, `doctor`. One theme (`ats-plain`).
116
+ - **M2** — `serve` web UI, remaining themes, handoff files, manifest.
117
+ - **M3** — framework mapping + search grounding (both opt-in).
118
+ - **M4** — Tauri desktop, signed installers, PyPI release polish.
119
+
120
+ ## Open questions (decide before M1 stories)
121
+
122
+ - PDF engine: WeasyPrint (HTML fidelity, heavier install) vs Typst
123
+ (fast, separate template language)? Lean WeasyPrint for M1.
124
+ - MasterProfile editing UX in CLI-only M1: hand-edited JSON/YAML is
125
+ acceptable; a review UI waits for `serve`.
126
+ - How strict is the verifier by default — drop unsupported bullets, or
127
+ keep-and-flag? Lean: flag in UI, drop only with `--strict`.
128
+ - Locale naming: "résumé" (US) vs "CV" (AU/UK/academic) — theme-level
129
+ setting, not a fork.
@@ -0,0 +1,44 @@
1
+ [build-system]
2
+ requires = ["hatchling"]
3
+ build-backend = "hatchling.build"
4
+
5
+ [project]
6
+ name = "sartor"
7
+ version = "0.0.1"
8
+ description = "Privacy-first résumé tailoring: your evidence in, a tailored résumé out — every claim traceable, nothing invented. Bring your own model (OpenAI-compatible, Ollama, Anthropic)."
9
+ readme = "README.md"
10
+ license = "MIT"
11
+ license-files = ["LICENSE"]
12
+ authors = [
13
+ { name = "Michael Borck", email = "michael.borck@curtin.edu.au" },
14
+ ]
15
+ keywords = ["resume", "cv", "tailoring", "ats", "career", "llm", "privacy", "cli"]
16
+ classifiers = [
17
+ "Development Status :: 1 - Planning",
18
+ "Environment :: Console",
19
+ "Intended Audience :: End Users/Desktop",
20
+ "License :: OSI Approved :: MIT License",
21
+ "Operating System :: OS Independent",
22
+ "Programming Language :: Python :: 3",
23
+ "Programming Language :: Python :: 3.10",
24
+ "Programming Language :: Python :: 3.11",
25
+ "Programming Language :: Python :: 3.12",
26
+ "Programming Language :: Python :: 3.13",
27
+ "Topic :: Text Processing",
28
+ ]
29
+ requires-python = ">=3.10"
30
+ dependencies = []
31
+
32
+ [project.scripts]
33
+ sartor = "sartor.cli:main"
34
+
35
+ [project.urls]
36
+ Homepage = "https://github.com/michael-borck/sartor"
37
+ Repository = "https://github.com/michael-borck/sartor"
38
+ Issues = "https://github.com/michael-borck/sartor/issues"
39
+
40
+ [tool.hatch.build.targets.wheel]
41
+ packages = ["src/sartor"]
42
+
43
+ [tool.hatch.build.targets.sdist]
44
+ exclude = ["dist", "**/._*"]
@@ -0,0 +1,10 @@
1
+ """Sartor — privacy-first résumé tailoring.
2
+
3
+ Pre-release placeholder (0.0.1): reserves the PyPI name while M1
4
+ (ingest → distil → tailor → verify → render) is built. See
5
+ docs/BRIEF.md and docs/SPEC.md in the repository.
6
+ """
7
+
8
+ __version__ = "0.0.1"
9
+
10
+ __all__ = ["__version__"]
@@ -0,0 +1,31 @@
1
+ """Command-line entry point. Placeholder until the M1 pipeline lands."""
2
+
3
+ from __future__ import annotations
4
+
5
+ import argparse
6
+
7
+ from sartor import __version__
8
+
9
+
10
+ def main(argv: list[str] | None = None) -> int:
11
+ parser = argparse.ArgumentParser(
12
+ prog="sartor",
13
+ description=(
14
+ "Privacy-first résumé tailoring: your evidence in, a tailored "
15
+ "résumé out — every claim traceable, nothing invented."
16
+ ),
17
+ )
18
+ parser.add_argument(
19
+ "--version", action="version", version=f"sartor {__version__}"
20
+ )
21
+ parser.parse_args(argv)
22
+ print(
23
+ f"sartor {__version__} — pre-release placeholder.\n"
24
+ "M1 (ingest, tailor, verify, render) is in development: "
25
+ "https://github.com/michael-borck/sartor"
26
+ )
27
+ return 0
28
+
29
+
30
+ if __name__ == "__main__":
31
+ raise SystemExit(main())