ldy/JobSourceAgent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
41037fb2f559a1d0733695d33e180e10f40e8307
JobSourceAgent/tests/test_cascade.py
ldy 41037fb2f5 careers-llm-extract
2026-06-17 22:56:04 -04:00

662 lines
27 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

"""Tests for jobsource/careers/cascade.py — all network-free via monkeypatching."""
from __future__ import annotations
import pytest
from jobsource.careers import CareersResult, find_careers_page
from jobsource.careers.cascade import _detect_ats_in_page, _finalize, _safe_get_html
from jobsource.careers.ats import ATSBoard, ATSFetch, ATSResult
# ---------------------------------------------------------------------------
# Fake helpers
# ---------------------------------------------------------------------------
class FakeResponse:
def __init__(self, status_code: int, text: str = "", url: str = "https://example.com"):
self.status_code = status_code
self.text = text
self.url = url
class FakeClient:
pass
# ---------------------------------------------------------------------------
# _detect_ats_in_page
# ---------------------------------------------------------------------------
class TestDetectATSInPage:
def test_returns_board_from_page_html(self, monkeypatch):
html = '<a href="https://boards.greenhouse.io/vercel">Jobs</a>'
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: FakeResponse(200, html),
)
board = _detect_ats_in_page("https://vercel.com/careers", FakeClient())
assert board is not None
assert board.ats_name == "greenhouse"
assert board.slug == "vercel"
def test_returns_none_on_404(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: FakeResponse(404, ""),
)
assert _detect_ats_in_page("https://vercel.com/careers", FakeClient()) is None
def test_returns_none_on_exception(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: (_ for _ in ()).throw(RuntimeError("timeout")),
)
assert _detect_ats_in_page("https://vercel.com/careers", FakeClient()) is None
def test_returns_none_when_no_ats_in_page(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: FakeResponse(200, "<html>no ats here</html>"),
)
assert _detect_ats_in_page("https://acme.com/careers", FakeClient()) is None
# ---------------------------------------------------------------------------
# _safe_get_html
# ---------------------------------------------------------------------------
class TestSafeGetHtml:
def test_returns_text_on_200(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: FakeResponse(200, "page html"),
)
result = _safe_get_html("https://acme.com", FakeClient())
assert result == "page html"
def test_returns_none_on_404(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: FakeResponse(404, ""),
)
result = _safe_get_html("https://acme.com", FakeClient())
assert result is None
def test_returns_none_on_exception(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade.request_with_retries",
lambda *a, **kw: (_ for _ in ()).throw(RuntimeError("network error")),
)
result = _safe_get_html("https://acme.com", FakeClient())
assert result is None
# ---------------------------------------------------------------------------
# find_careers_page — tier ordering and early return
# ---------------------------------------------------------------------------
class TestCascadeTierOrdering:
def _patch_tiers(self, monkeypatch, *, ats=None, slug_guess=None,
url_pattern=None, homepage=None, sitemap=None, html="<html/>"):
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda website, client: html,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: ats,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
lambda *a, **kw: slug_guess,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: url_pattern,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
lambda *a, **kw: homepage,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: sitemap,
)
# Also stub detect_ats_in_url so _finalize doesn't try to do network calls
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None,
)
def test_ats_hit_returns_095_confidence(self, monkeypatch):
ats_result = ATSResult(
ats_name="greenhouse",
careers_url="https://boards.greenhouse.io/acme",
position_url="https://careers.acme.com/positions/1",
job_count=10,
)
self._patch_tiers(monkeypatch, ats=ats_result)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.careers_url == "https://boards.greenhouse.io/acme"
assert result.confidence == 0.95
assert result.method == "ats:greenhouse"
assert result.ats_name == "greenhouse"
assert result.position_url == "https://careers.acme.com/positions/1"
def test_url_pattern_hit_when_ats_misses(self, monkeypatch):
self._patch_tiers(monkeypatch, url_pattern="https://acme.com/careers")
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.careers_url == "https://acme.com/careers"
assert result.confidence == 0.80
assert result.method == "url_pattern"
assert result.ats_name is None
def test_homepage_scan_hit_when_ats_and_url_pattern_miss(self, monkeypatch):
self._patch_tiers(monkeypatch, homepage="https://acme.com/careers")
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.careers_url == "https://acme.com/careers"
assert result.confidence == 0.60
assert result.method == "homepage_scan"
def test_sitemap_hit_when_all_else_misses(self, monkeypatch):
self._patch_tiers(monkeypatch, sitemap="https://acme.com/careers")
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.careers_url == "https://acme.com/careers"
assert result.confidence == 0.50
assert result.method == "sitemap"
def test_all_miss_returns_none_method(self, monkeypatch):
self._patch_tiers(monkeypatch)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.careers_url is None
assert result.confidence == 0.0
assert result.method == "none"
def test_ats_hit_skips_later_tiers(self, monkeypatch):
"""When ATS resolves, slug_guess/url_pattern/homepage/sitemap should not be called."""
ats_result = ATSResult(
ats_name="lever", careers_url="https://jobs.lever.co/acme",
)
later_called: list[str] = []
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda *a, **kw: "<html/>",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: ats_result,
)
def make_tracker(name):
def fn(*a, **kw):
later_called.append(name)
return None
return fn
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
make_tracker("slug_guess"),
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
make_tracker("url_pattern"),
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
make_tracker("homepage_scan"),
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
make_tracker("sitemap"),
)
find_careers_page("https://www.acme.com", client=FakeClient())
assert later_called == []
def test_failing_tier_falls_through(self, monkeypatch):
"""A tier that raises should not abort the cascade."""
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda *a, **kw: "<html/>",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: (_ for _ in ()).throw(RuntimeError("ats exploded")),
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: "https://acme.com/careers",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: None,
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "url_pattern"
assert result.careers_url == "https://acme.com/careers"
# ---------------------------------------------------------------------------
# ATS-URL upgrade in _finalize
# ---------------------------------------------------------------------------
class TestFinalizeATSUpgrade:
def test_lever_url_upgrades_to_ats_lever(self, monkeypatch):
"""When a heuristic finds a Lever URL, _finalize upgrades to ats:lever."""
from jobsource.careers.ats import ATSBoard
lever_board = ATSBoard(
ats_name="lever",
slug="acme",
careers_url="https://jobs.lever.co/acme",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: lever_board,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats._FETCH_DISPATCH",
{"lever": lambda board, client: ATSFetch(first_url="https://jobs.lever.co/acme/abc", job_count=5)},
)
result = _finalize(
"https://jobs.lever.co/acme", "homepage_scan", 0.60,
"https://www.acme.com", FakeClient(),
)
assert result.method == "ats:lever"
assert result.confidence == 0.95
assert result.position_url == "https://jobs.lever.co/acme/abc"
assert result.ats_name == "lever"
def test_non_ats_url_no_upgrade(self, monkeypatch):
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None,
)
result = _finalize(
"https://acme.com/careers", "url_pattern", 0.80,
"https://www.acme.com", FakeClient(),
)
assert result.method == "url_pattern"
assert result.confidence == 0.80
assert result.position_url is None
def test_ats_upgrade_from_page_html(self, monkeypatch):
"""URL pattern finds /careers; fetching that page reveals Greenhouse embed → upgrade."""
gh_board = ATSBoard(
ats_name="greenhouse",
slug="vercel",
careers_url="https://boards.greenhouse.io/vercel",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None, # URL string itself is not an ATS URL
)
monkeypatch.setattr(
"jobsource.careers.cascade._detect_ats_in_page",
lambda url, client: gh_board, # page HTML reveals Greenhouse
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats._FETCH_DISPATCH",
{"greenhouse": lambda board, client: ATSFetch(first_url="https://job-boards.greenhouse.io/vercel/jobs/123", job_count=73)},
)
result = _finalize(
"https://vercel.com/careers", "url_pattern", 0.80,
"https://vercel.com", FakeClient(),
)
assert result.method == "ats:greenhouse"
assert result.confidence == 0.95
assert result.careers_url == "https://boards.greenhouse.io/vercel"
assert result.position_url == "https://job-boards.greenhouse.io/vercel/jobs/123"
assert result.ats_name == "greenhouse"
def test_ats_upgrade_fetch_failure_falls_back_to_original(self, monkeypatch):
"""If ATS fetch during upgrade fails, return the original heuristic result."""
from jobsource.careers.ats import ATSBoard
gh_board = ATSBoard(
ats_name="greenhouse",
slug="acme",
careers_url="https://boards.greenhouse.io/acme",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: gh_board,
)
def boom(board, client):
raise RuntimeError("api down")
monkeypatch.setattr(
"jobsource.careers.cascade._ats._FETCH_DISPATCH",
{"greenhouse": boom},
)
result = _finalize(
"https://boards.greenhouse.io/acme", "sitemap", 0.50,
"https://www.acme.com", FakeClient(),
)
# Upgrade failed — returns the original heuristic result
assert result.method == "sitemap"
assert result.confidence == 0.50
# ---------------------------------------------------------------------------
# CareersResult model
# ---------------------------------------------------------------------------
class TestCareersResultModel:
def test_defaults(self):
r = CareersResult()
assert r.careers_url is None
assert r.confidence == 0.0
assert r.method == "none"
assert r.ats_name is None
assert r.position_url is None
def test_full(self):
r = CareersResult(
careers_url="https://boards.greenhouse.io/acme",
confidence=0.95,
method="ats:greenhouse",
ats_name="greenhouse",
position_url="https://careers.acme.com/positions/1",
)
assert r.careers_url == "https://boards.greenhouse.io/acme"
assert r.ats_name == "greenhouse"
# ---------------------------------------------------------------------------
# Tier 1b — slug-guess in cascade ordering
# ---------------------------------------------------------------------------
class TestSlugGuessTier:
"""Tests that slug-guess wires correctly into the cascade between HTML-ATS and url_pattern."""
def _patch_for_slug_guess(self, monkeypatch, *, slug_guess_result=None,
url_pattern=None):
"""Patch cascade so HTML-ATS always misses; control slug_guess and url_pattern."""
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda website, client: "<html/>",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
lambda *a, **kw: slug_guess_result,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: url_pattern,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None,
)
def test_slug_guess_hit_returns_090_confidence(self, monkeypatch):
"""A slug-guess hit should produce confidence=0.90 and method=ats:{name}:slug_guess."""
slug_result = ATSResult(
ats_name="greenhouse",
careers_url="https://boards.greenhouse.io/anthropic",
position_url="https://boards.greenhouse.io/anthropic/jobs/1",
job_count=370,
)
self._patch_for_slug_guess(monkeypatch, slug_guess_result=slug_result)
result = find_careers_page(
"https://www.anthropic.com",
company_name="Anthropic",
client=FakeClient(),
)
assert result.confidence == 0.90
assert result.method == "ats:greenhouse:slug_guess"
assert result.careers_url == "https://boards.greenhouse.io/anthropic"
assert result.ats_name == "greenhouse"
assert result.position_url == "https://boards.greenhouse.io/anthropic/jobs/1"
def test_slug_guess_hit_blocks_url_pattern(self, monkeypatch):
"""If slug-guess hits, url_pattern should not be called."""
url_pattern_called: list[bool] = []
slug_result = ATSResult(
ats_name="lever",
careers_url="https://jobs.lever.co/acme",
position_url="https://jobs.lever.co/acme/xyz",
job_count=5,
)
self._patch_for_slug_guess(
monkeypatch, slug_guess_result=slug_result, url_pattern="https://acme.com/careers"
)
# Override url_pattern with a tracker
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: url_pattern_called.append(True) or None,
)
find_careers_page("https://www.acme.com", client=FakeClient())
assert url_pattern_called == []
def test_slug_guess_miss_falls_through_to_url_pattern(self, monkeypatch):
"""When slug-guess misses, the cascade continues to url_pattern."""
self._patch_for_slug_guess(
monkeypatch,
slug_guess_result=None,
url_pattern="https://acme.com/careers",
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "url_pattern"
assert result.confidence == 0.80
def test_company_name_forwarded_to_slug_guess(self, monkeypatch):
"""company_name must be passed through to recover_via_slug_guess."""
received: list[tuple] = []
def fake_recover(website, company_name, client):
received.append((website, company_name))
return None
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda *a, **kw: "<html/>",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
fake_recover,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: None,
)
find_careers_page(
"https://www.anthropic.com",
company_name="Anthropic",
client=FakeClient(),
)
assert len(received) == 1
assert received[0] == ("https://www.anthropic.com", "Anthropic")
def test_slug_guess_tier_error_falls_through(self, monkeypatch):
"""A slug-guess tier that raises should not abort the cascade."""
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda *a, **kw: "<html/>",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_and_fetch",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.recover_via_slug_guess",
lambda *a, **kw: (_ for _ in ()).throw(RuntimeError("slug_guess exploded")),
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.probe_url_patterns",
lambda *a, **kw: "https://acme.com/careers",
)
monkeypatch.setattr(
"jobsource.careers.cascade._ats.detect_ats_in_url",
lambda url: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.scan_homepage_links",
lambda *a, **kw: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: None,
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "url_pattern"
assert result.careers_url == "https://acme.com/careers"
# ---------------------------------------------------------------------------
# Tier 5 — Cheap-LLM careers-link classification
# ---------------------------------------------------------------------------
class TestLLMClassifyTier:
"""Tier 5 fires after sitemap, no-ops when LLM returns None, and is skipped on earlier hits."""
def _patch_all_deterministic_miss(self, monkeypatch, *, homepage_html: str = "<html/>") -> None:
"""Patch tiers 1–4 to all miss; leaves Tier 5 for the caller to control."""
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda website, client: homepage_html,
)
monkeypatch.setattr("jobsource.careers.cascade._ats.detect_and_fetch", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._ats.recover_via_slug_guess", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._heuristics.probe_url_patterns", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._heuristics.scan_homepage_links", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._heuristics.parse_sitemap", lambda *a, **kw: None)
# Prevent _finalize from making network calls
monkeypatch.setattr("jobsource.careers.cascade._ats.detect_ats_in_url", lambda url: None)
def test_llm_classify_hit_returns_055_confidence(self, monkeypatch):
from jobsource.careers.classify_llm import Anchor, CareerLinkResult
self._patch_all_deterministic_miss(
monkeypatch,
homepage_html='<html><body><a href="/careers">Careers</a></body></html>',
)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.extract_anchors",
lambda html, base_url: [Anchor(url="https://acme.com/careers", text="Careers")],
)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.classify_careers_link",
lambda anchors: CareerLinkResult(careers_url="https://acme.com/careers", confidence=0.85),
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "llm_classify"
assert result.confidence == 0.55 # fixed tier-level confidence from cascade
assert result.careers_url == "https://acme.com/careers"
def test_llm_classify_none_falls_through_to_method_none(self, monkeypatch):
from jobsource.careers.classify_llm import Anchor
self._patch_all_deterministic_miss(monkeypatch)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.extract_anchors",
lambda html, base_url: [Anchor(url="https://acme.com/about", text="About")],
)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.classify_careers_link",
lambda anchors: None,
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "none"
assert result.careers_url is None
def test_tier5_skipped_when_homepage_html_is_none(self, monkeypatch):
"""When the homepage fetch returns None, Tier 5 is silently bypassed."""
llm_called: list[bool] = []
self._patch_all_deterministic_miss(monkeypatch)
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda website, client: None,
)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.classify_careers_link",
lambda anchors: llm_called.append(True) or None,
)
find_careers_page("https://www.acme.com", client=FakeClient())
assert llm_called == []
def test_tier5_not_called_when_sitemap_hits(self, monkeypatch):
"""When sitemap resolves, the cascade returns before reaching Tier 5."""
llm_called: list[bool] = []
monkeypatch.setattr(
"jobsource.careers.cascade._safe_get_html",
lambda website, client: "<html/>",
)
monkeypatch.setattr("jobsource.careers.cascade._ats.detect_and_fetch", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._ats.recover_via_slug_guess", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._heuristics.probe_url_patterns", lambda *a, **kw: None)
monkeypatch.setattr("jobsource.careers.cascade._heuristics.scan_homepage_links", lambda *a, **kw: None)
monkeypatch.setattr(
"jobsource.careers.cascade._heuristics.parse_sitemap",
lambda *a, **kw: "https://acme.com/careers",
)
monkeypatch.setattr("jobsource.careers.cascade._ats.detect_ats_in_url", lambda url: None)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.classify_careers_link",
lambda anchors: llm_called.append(True) or None,
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "sitemap"
assert llm_called == []
def test_tier5_error_falls_through_gracefully(self, monkeypatch):
"""A Tier 5 exception should not abort the cascade."""
self._patch_all_deterministic_miss(monkeypatch)
monkeypatch.setattr(
"jobsource.careers.cascade._classify_llm.extract_anchors",
lambda html, base_url: (_ for _ in ()).throw(RuntimeError("classify boom")),
)
result = find_careers_page("https://www.acme.com", client=FakeClient())
assert result.method == "none"
assert result.careers_url is None
Reference in New Issue View Git Blame Copy Permalink