careers-llm-extract

2026-06-17 22:56:04 -04:00
parent 113a4ced36
commit 41037fb2f5
7 changed files with 1350 additions and 28 deletions
--- a/tests/test_classify_llm.py
+++ b/tests/test_classify_llm.py
@@ -0,0 +1,256 @@
+"""Tests for careers/classify_llm.py — all network-free via monkeypatching."""
+from __future__ import annotations
+
+import pytest
+
+from jobsource.careers.classify_llm import (
+    Anchor,
+    CareerLinkResult,
+    JobLinkResult,
+    _LinkChoice,
+    classify_careers_link,
+    classify_job_link,
+    extract_anchors,
+)
+
+
+# ---------------------------------------------------------------------------
+# Fake helpers
+# ---------------------------------------------------------------------------
+
+
+class _FakeResult:
+    """Minimal stand-in for a Pydantic AI run result."""
+
+    def __init__(self, output: _LinkChoice) -> None:
+        self.output = output
+
+
+class _FakeAgent:
+    """Stand-in for a Pydantic AI Agent; returns a fixed _LinkChoice."""
+
+    def __init__(self, choice: _LinkChoice) -> None:
+        self._choice = choice
+
+    def run_sync(self, prompt: str) -> _FakeResult:
+        return _FakeResult(self._choice)
+
+
+class _RaisingAgent:
+    """Stand-in that raises on run_sync, simulating an LLM failure."""
+
+    def run_sync(self, prompt: str) -> _FakeResult:
+        raise RuntimeError("LLM unavailable")
+
+
+def _real_settings():
+    """Fake settings object with non-placeholder values."""
+    return type("S", (), {
+        "classifier_model": "openai:gpt-4o-mini",
+        "llm_api_key": "sk-real-key",
+    })()
+
+
+# ---------------------------------------------------------------------------
+# extract_anchors
+# ---------------------------------------------------------------------------
+
+_SAMPLE_HTML = """
+<html><body>
+  <a href="/about">About Us</a>
+  <a href="/careers">Careers</a>
+  <a href="https://external.com/jobs">External Jobs</a>
+  <a href="mailto:info@acme.com">Email</a>
+  <a href="#section">Anchor</a>
+  <a href="javascript:void(0)">JS link</a>
+  <a href="/careers">Duplicate careers link</a>
+</body></html>
+"""
+
+
+class TestExtractAnchors:
+    def test_resolves_relative_to_absolute(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        urls = [a.url for a in anchors]
+        assert "https://acme.com/about" in urls
+        assert "https://acme.com/careers" in urls
+
+    def test_preserves_already_absolute_urls(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        urls = [a.url for a in anchors]
+        assert "https://external.com/jobs" in urls
+
+    def test_skips_mailto(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        assert all("mailto:" not in a.url for a in anchors)
+
+    def test_skips_hash_fragments(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        # A pure # anchor resolves to base_url itself (https://acme.com) —
+        # it starts with http so it passes the scheme check but we skip hrefs
+        # that start with "#". Verify the JavaScript link is also gone.
+        assert all("javascript:" not in a.url for a in anchors)
+
+    def test_deduplicates_by_url(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        urls = [a.url for a in anchors]
+        assert urls.count("https://acme.com/careers") == 1
+
+    def test_captures_anchor_text(self):
+        anchors = extract_anchors(_SAMPLE_HTML, "https://acme.com")
+        career_anchor = next(a for a in anchors if a.url == "https://acme.com/careers")
+        assert career_anchor.text == "Careers"
+
+    def test_empty_html_returns_empty_list(self):
+        assert extract_anchors("", "https://acme.com") == []
+
+    def test_html_with_no_links_returns_empty_list(self):
+        assert extract_anchors("<html><body><p>No links here.</p></body></html>", "https://acme.com") == []
+
+    def test_text_truncated_at_max_length(self):
+        long_text = "x" * 200
+        html = f'<html><body><a href="/careers">{long_text}</a></body></html>'
+        anchors = extract_anchors(html, "https://acme.com")
+        assert len(anchors) == 1
+        assert len(anchors[0].text) <= 80
+
+
+# ---------------------------------------------------------------------------
+# Graceful degradation: placeholder settings → always None
+# ---------------------------------------------------------------------------
+
+
+class TestPlaceholderGate:
+    """When classifier_model or llm_api_key is a placeholder, classifiers no-op."""
+
+    def test_classify_careers_link_returns_none_with_placeholder(self):
+        anchors = [Anchor(url="https://acme.com/careers", text="Careers")]
+        assert classify_careers_link(anchors) is None
+
+    def test_classify_job_link_returns_none_with_placeholder(self):
+        anchors = [Anchor(url="https://acme.com/jobs/123", text="Engineer")]
+        assert classify_job_link(anchors) is None
+
+    def test_classify_careers_link_returns_none_on_empty_list(self, monkeypatch):
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm.get_settings",
+            _real_settings,
+        )
+        assert classify_careers_link([]) is None
+
+    def test_classify_job_link_returns_none_on_empty_list(self, monkeypatch):
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm.get_settings",
+            _real_settings,
+        )
+        assert classify_job_link([]) is None
+
+
+# ---------------------------------------------------------------------------
+# Success paths via _build_link_agent monkeypatch
+# ---------------------------------------------------------------------------
+
+
+class TestClassifySuccess:
+    def _patch_agent(self, monkeypatch, choice: _LinkChoice) -> None:
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm.get_settings",
+            _real_settings,
+        )
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            lambda instructions: _FakeAgent(choice),
+        )
+
+    def test_careers_pick_returns_mapped_url(self, monkeypatch):
+        anchors = [
+            Anchor(url="https://acme.com/about", text="About"),
+            Anchor(url="https://acme.com/careers", text="Careers"),
+        ]
+        self._patch_agent(monkeypatch, _LinkChoice(index=1, confidence=0.95))
+        result = classify_careers_link(anchors)
+        assert result is not None
+        assert isinstance(result, CareerLinkResult)
+        assert result.careers_url == "https://acme.com/careers"
+        assert result.confidence == 0.95
+
+    def test_job_pick_returns_mapped_url(self, monkeypatch):
+        anchors = [
+            Anchor(url="https://acme.com/jobs/", text="All Jobs"),
+            Anchor(url="https://acme.com/jobs/123", text="Software Engineer"),
+        ]
+        self._patch_agent(monkeypatch, _LinkChoice(index=1, confidence=0.85))
+        result = classify_job_link(anchors)
+        assert result is not None
+        assert isinstance(result, JobLinkResult)
+        assert result.job_url == "https://acme.com/jobs/123"
+        assert result.confidence == 0.85
+
+    def test_careers_first_index_zero(self, monkeypatch):
+        anchors = [Anchor(url="https://acme.com/jobs", text="Jobs")]
+        self._patch_agent(monkeypatch, _LinkChoice(index=0, confidence=0.70))
+        result = classify_careers_link(anchors)
+        assert result is not None
+        assert result.careers_url == "https://acme.com/jobs"
+
+
+# ---------------------------------------------------------------------------
+# Abstain and failure paths
+# ---------------------------------------------------------------------------
+
+
+class TestClassifyFailurePaths:
+    def _patch_real_settings(self, monkeypatch) -> None:
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm.get_settings",
+            _real_settings,
+        )
+
+    def test_abstain_index_minus_one_returns_none(self, monkeypatch):
+        self._patch_real_settings(monkeypatch)
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            lambda instructions: _FakeAgent(_LinkChoice(index=-1, confidence=0.0)),
+        )
+        anchors = [Anchor(url="https://acme.com/about", text="About")]
+        assert classify_careers_link(anchors) is None
+
+    def test_out_of_range_index_returns_none(self, monkeypatch):
+        self._patch_real_settings(monkeypatch)
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            lambda instructions: _FakeAgent(_LinkChoice(index=99, confidence=0.9)),
+        )
+        anchors = [Anchor(url="https://acme.com/careers", text="Careers")]
+        assert classify_careers_link(anchors) is None
+
+    def test_run_sync_exception_returns_none(self, monkeypatch):
+        self._patch_real_settings(monkeypatch)
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            lambda instructions: _RaisingAgent(),
+        )
+        anchors = [Anchor(url="https://acme.com/careers", text="Careers")]
+        assert classify_careers_link(anchors) is None
+
+    def test_build_agent_exception_returns_none(self, monkeypatch):
+        self._patch_real_settings(monkeypatch)
+
+        def boom(instructions: str):
+            raise RuntimeError("import error")
+
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            boom,
+        )
+        anchors = [Anchor(url="https://acme.com/careers", text="Careers")]
+        assert classify_careers_link(anchors) is None
+
+    def test_job_classify_run_sync_exception_returns_none(self, monkeypatch):
+        self._patch_real_settings(monkeypatch)
+        monkeypatch.setattr(
+            "jobsource.careers.classify_llm._build_link_agent",
+            lambda instructions: _RaisingAgent(),
+        )
+        anchors = [Anchor(url="https://acme.com/jobs/123", text="Engineer")]
+        assert classify_job_link(anchors) is None