{{ p.title }}{% if p.subtitle %} · {{ p.subtitle }}{% endif %}
+ {% if p.description %}{{ p.description }}
{% endif %} + {% if p.tech %}Technologies: {{ p.tech | join(", ") }}
{% endif %}Education
- {% for e in cv.education %} -Certifications
- {% for c in cv.certifications %} -Skills & Technical Proficiency
+ {% for grp in cv.skills %} +{{ grp.category }}: {{ grp["items"] | join(", ") }}
{% endfor %}Sam Okafor
' in html - for href in ("mailto:sam@example.com", "tel:+447700900456", "https://linkedin.com/in/sam-okafor", - "https://github.com/samokafor", "https://samokafor.dev"): - assert f'href="{href}"' in html - assert "BSc (Hons) Computer Science | 06/2025 | Northfield University" in html and "Grade: 2:1" in html - assert "Software Engineering Intern | Brightwave Analytics | 06/2024 – 09/2024" in html + assert _headings(html) == ["Professional Summary", "Education", "Experience", "Projects", + "Skills & Technical Proficiency"] + assert 'Sam Okafor
' in html + contact = re.search(r'(.*?)
', html, re.S).group(1) + assert re.sub(r"<[^>]+>|\s+", " ", contact).split() == "Bristol | +44 7700 900456 | sam@example.com | LinkedIn | GitHub | Website".split() + for href in ("mailto:sam@example.com", "https://linkedin.com/in/sam-okafor", "https://github.com/samokafor", + "https://samokafor.dev"): + assert f'href="{href}"' in contact + assert "tel:" not in contact # the phone is plain text + assert _entry(html, "BSc (Hons) Computer Science | Northfield University") == "09/2022 – 06/2025" + assert "Grade: 2:1
" in html + assert _entry(html, "Software Engineering Intern | Brightwave Analytics") == "06/2024 – 09/2024" assert "Cut dashboard load time from 6 s to 1.5 s" in html - assert "AWS Cloud Practitioner | Amazon Web Services | 03/2025" in html + education = html.split("Education
")[1].split("(.*?)
', doc_render.render_cv_html(_tailored_cv())).group(1) + summary = re.search(r"Professional Summary
\s*(.*?)
", doc_render.render_cv_html(_tailored_cv())).group(1) assert summary == "Computer Science graduate who builds full-stack web apps and data pipelines." -def test_portfolio_prose_becomes_two_clean_sentences_with_labelled_links(llm): +def test_projects_show_a_titled_paragraph_and_a_technologies_line(llm): llm.reply = "{}" html = doc_render.render_cv_html(_tailored_cv()) projects = html.split("Projects
")[1].split("")[0] - assert re.findall(r'Technologies: Python, PyTorch
' in projects + assert "Skills")[1].split("")[0] + skills = html.split("Skills & Technical Proficiency
")[1].split("")[0] missing = [s for grp in GRADUATE["skills"] for s in grp["skills"] if s not in skills] - assert missing == [] and skills.count('class="skill-row"') == doc_render.MAX_SKILL_GROUPS + assert missing == [] and skills.count("") == doc_render.MAX_SKILL_GROUPS
-def _regrouping(groups: dict[str, list[str]]) -> str:
- """The model's reply: the base CV with its skills regrouped as given."""
+def _choosing(groups: dict[str, list[str]]) -> str:
+ """The model's reply: the base CV with the skills it chose for the job."""
cv = agent._persona_to_cvdoc(GRADUATE)
return json.dumps(cv | {"skills": [{"category": c, "items": items} for c, items in groups.items()]})
-ALL_SKILLS = {
- "Languages": ["Python", "TypeScript", "Java", "SQL"],
- "Web": ["React", "Next.js", "Tailwind CSS", "FastAPI", "Node.js", "Spring Boot"],
- "Data & Machine Learning": ["PostgreSQL", "Redis", "pandas", "scikit-learn", "PyTorch"],
- "Cloud & Tools": ["AWS", "Docker", "Git", "GitHub Actions", "CI/CD"],
- "Testing": ["pytest", "Jest", "Playwright"],
+CHOSEN = { # 13 of the profile's 23 skills, the way the candidate's own CV groups them
+ "Languages": ["Python", "SQL"],
+ "Backend & Data": ["FastAPI", "PostgreSQL", "Redis"],
+ "ML & AI": ["scikit-learn", "PyTorch", "pandas"],
+ "Tools": ["Docker", "Git", "GitHub Actions", "AWS", "pytest"],
}
-def test_the_models_regrouping_is_used_when_it_keeps_every_skill(llm):
- llm.reply = _regrouping({**ALL_SKILLS, "Languages": ["python", "TypeScript", "Java", "SQL"]})
+def test_the_model_chooses_the_skills_for_the_job(llm):
+ llm.reply = _choosing({**CHOSEN, "Languages": ["python", "SQL"]})
cv = _tailored_cv()
- assert cv["skills"] == [{"category": c, "items": items} for c, items in ALL_SKILLS.items()] # profile spelling
- assert "at most 6 categories" in llm.prompts[-1]
+ assert cv["skills"] == [{"category": c, "items": items} for c, items in CHOSEN.items()] # profile spelling
+ assert "choose the 20-30 skills" in llm.prompts[-1] and "Leave out spoken languages" in llm.prompts[-1]
@pytest.mark.parametrize("change", [
- pytest.param(lambda g: g["Testing"].remove("Jest"), id="drops a skill"),
- pytest.param(lambda g: g["Testing"].append("Selenium"), id="adds one"),
- pytest.param(lambda g: g["Testing"].append("Python"), id="repeats one"),
- pytest.param(lambda g: g.update({"Misc": [g["Web"].pop()], "More": [g["Web"].pop()]}), id="7 groups"),
- pytest.param(lambda g: g.update({"": g.pop("Testing")}), id="unnamed group"),
+ pytest.param(lambda g: g["Tools"].append("Kubernetes"), id="invents one"),
+ pytest.param(lambda g: g["Tools"].append("Python"), id="repeats one"),
+ pytest.param(lambda g: g["Tools"].append("GitHub Actions CI/CD pipelines"), id="renames one"),
])
-def test_a_regrouping_that_changes_the_skills_is_ignored(llm, change):
- groups = copy.deepcopy(ALL_SKILLS)
+def test_skills_the_model_invents_renames_or_repeats_are_dropped(llm, change):
+ groups = copy.deepcopy(CHOSEN)
change(groups)
- llm.reply = _regrouping(groups)
- assert _tailored_cv()["skills"] == agent._persona_to_cvdoc(GRADUATE)["skills"]
+ llm.reply = _choosing(groups)
+ assert _tailored_cv()["skills"] == [{"category": c, "items": items} for c, items in CHOSEN.items()]
+
+
+def test_an_unnamed_row_is_called_other():
+ base = [{"category": "All", "items": [f"Skill {i}" for i in range(12)]}]
+ chosen = agent._chosen_skills([{"category": "", "items": [f"Skill {i}" for i in range(12)]}], base)
+ assert chosen == [{"category": "Other", "items": [f"Skill {i}" for i in range(12)]}]
+
+
+@pytest.mark.parametrize("n, used", [(9, False), (10, True), (40, True), (41, False)])
+def test_too_few_real_skills_or_too_many_falls_back_to_the_full_list(n, used):
+ base = [{"category": "All", "items": [f"Skill {i}" for i in range(50)]}]
+ chosen = agent._chosen_skills([{"category": "All", "items": [f"Skill {i}" for i in range(n)]}], base)
+ assert (chosen is not None) is used
+
+
+@pytest.mark.parametrize("core, location", [
+ ({"city": "Bristol", "country": "United Kingdom"}, "Bristol, United Kingdom"),
+ ({"city": "London, United Kingdom", "country": "United Kingdom"}, "London, United Kingdom"),
+ ({"city": "Bristol"}, "Bristol"),
+ ({"country": "United Kingdom"}, "United Kingdom"),
+])
+def test_the_cv_shows_city_and_country(core, location):
+ assert agent._cv_location(core) == location
@pytest.mark.parametrize("categories, rows", [
- (["A", "B", "C", "D", "E", "F", "G", "H"], ["A", "B", "C", "D", "E", "Other"]),
- (["A", "Other", "B", "C", "D", "E", "F"], ["A", "B", "C", "D", "E", "Other"]), # absorbs, moves last
+ (["A", "B", "C", "D", "E", "F", "G", "H"], ["A", "B", "C", "D", "Other"]),
+ (["A", "Other", "B", "C", "D", "E", "F"], ["A", "B", "C", "D", "Other"]), # absorbs, moves last
(["A", "B", "C"], ["A", "B", "C"]),
])
def test_extra_skill_groups_fold_into_other(categories, rows):
@@ -284,7 +328,7 @@ def test_length_limits_hold_whatever_the_model_returns():
data = doc_render.build_cv_template_data(cv)
assert [r["title"] for r in data["experience"]] == ["Engineer 3", "Engineer 2", "Engineer 1", "Engineer 0"]
assert max(len(r["bullets"]) for r in data["experience"]) == 4
- assert len(data["projects"]) == 3 and len(data["education"]) == 4 and len(data["certifications"]) == 6
+ assert len(data["projects"]) == 4 and len(data["education"]) == 4 and len(data["certifications"]) == 6
assert len(data["skills"][0]["items"]) == 15 # skills are never cut
assert len(data["summary"]) <= 420 and data["summary"].endswith(".") # ends on a whole sentence
assert len(data["projects"][0]["description"]) <= 520 and data["projects"][0]["description"].endswith(".")
@@ -306,7 +350,8 @@ class Pages:
steps: list[str] = []
def count(html):
- Pages.steps.append(re.search(r"