b1e9cdd182
- A1: fire the hard filter (Apify is precise); embedding vibe-ranker (embed.py) blends with the floor-free white-box (rank.py) to sift ~90 -> top 18 - A2: Opus curator (curate.py) reads the 18 -> honest <=count + Gemini-voiced report cards; graceful fallback to the white-box + templated cards - LLM via opencode.ai/zen gateway (llm.py): Opus chat + direct-OpenAI embeddings - Run LLM work off the event loop (asyncio.to_thread) so the Redis response publishes - C1: dedicated Postgres (app/db/) persists the per-user feed; get_scout_feed replays it so matches survive navigation/refresh - match contract + report-card fields (schema.py); skills.py; tests/
50 lines
2.7 KiB
Python
50 lines
2.7 KiB
Python
"""Phase 1 — the floor-free + coverage-aware invariants (the v1-collapse guardrails).
|
||
|
||
This is the test that would have caught the old engine: every factor is float-or-`None`
|
||
(no constant floors), and a sparse profile must NOT collapse all jobs to one neutral score.
|
||
"""
|
||
from app.engine import rank as R
|
||
|
||
_JOBS = [
|
||
{"title": "Product Manager", "seniority_level": "senior", "location_city": "New Delhi",
|
||
"location_country": "India", "salary_lpa": 30,
|
||
"details": {"skills": ["Product Management", "SQL"], "industry": "Fintech",
|
||
"description": "product manager roadmap stakeholders"}},
|
||
{"title": "Sales Lead", "seniority_level": "lead", "location_city": "Mumbai",
|
||
"location_country": "India", "salary_lpa": None,
|
||
"details": {"skills": ["Sales", "CRM"], "industry": "SaaS", "description": "sales quota targets"}},
|
||
{"title": "Product Manager", "seniority_level": "mid", "location_mode": "remote",
|
||
"location_city": "Remote", "details": {"skills": ["Product Management"], "description": "product manager"}},
|
||
{"title": "Designer", "seniority_level": None, "location_city": None, "salary_lpa": None, "details": {}},
|
||
]
|
||
|
||
|
||
def test_every_factor_returns_float_or_none():
|
||
p_full = R.profile_from({"title": "Product Manager", "location": ["New Delhi · India"],
|
||
"industry": ["Fintech"], "experience": ["Senior"], "years": 9,
|
||
"role": ["Product"], "targetComp": "₹25–30L"})
|
||
p_sparse = R.profile_from({"title": "Product Manager"})
|
||
for p in (p_full, p_sparse):
|
||
for job in _JOBS:
|
||
for name, fn in R.FACTORS.items():
|
||
v = fn(p, job)
|
||
assert v is None or (isinstance(v, float) and 0.0 <= v <= 1.0), f"{name} → {v!r}"
|
||
|
||
|
||
def test_sparse_profile_does_not_collapse():
|
||
# The v1 bug: sparse profile → everyone ~60. Assert real score variance across varied jobs.
|
||
p = R.profile_from({"title": "Product Manager", "location": ["New Delhi · India"]})
|
||
us = [u for u in (R.utility(p, j)[0] for j in _JOBS) if u is not None]
|
||
assert len(us) >= 2 and (max(us) - min(us)) > 0.1, f"collapsed to a flat band: {us}"
|
||
|
||
|
||
def test_absent_factor_redistributes_weight_not_floors():
|
||
# A job missing salary/industry/skills must still score on what IS present (no 60 injected).
|
||
p = R.profile_from({"title": "Product Manager", "location": ["New Delhi · India"]})
|
||
bare = {"title": "Product Manager", "location_city": "New Delhi", "location_country": "India",
|
||
"details": {"description": "product manager"}}
|
||
U, factors = R.utility(p, bare)
|
||
assert U is not None
|
||
assert factors["salary"] is None and factors["skill"] is None # absent
|
||
assert factors["location"] == 1.0 # present, real
|