Give an agent a CV, languages, avoid and must texts, and read a posting's languages with a level and its deadline (migration 28)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
WolfandClaude Opus 5.5 committed 2026-10-03 21:54:16 +02:00
1 parent ed6d641373
commit 4fc474c437
15 files changed
+716 -79

No files matched your search

+2 -2
View File
@@ -148,7 +148,7 @@ def test_migration_20_seeds_an_old_file_and_leaves_collisions(tmp_path: Path) ->
"VALUES (?, ?, '', '')", (key, name))
conn.execute("PRAGMA user_version = 19")
reopened = Store(path)
assert reopened.migrations_applied == [20, 21, 22, 23, 24, 25, 26, 27]
assert reopened.migrations_applied == [20, 21, 22, 23, 24, 25, 26, 27, 28]
assert reopened.duplicate_companies() == [("orsted", ["orsted", "ørsted"])]
assert len(reopened.companies()) == 3
@@ -186,7 +186,7 @@ def test_migration_21_tags_the_favourites_and_the_boards_that_read_all_roles(tmp
conn.execute("ALTER TABLE companies DROP COLUMN all_countries")
conn.execute("PRAGMA user_version = 20")
reopened = Store(path)
assert reopened.migrations_applied == [21, 22, 23, 24, 25, 26, 27]
assert reopened.migrations_applied == [21, 22, 23, 24, 25, 26, 27, 28]
# The tag of 21 is All countries and All roles since 24.
tagged = {"mistral-ai": 1, "acme": 0, "black-forest-labs": 1, "liked": 0}
assert {c.key: c.all_roles for c in reopened.companies()} == tagged
+238
View File
@@ -0,0 +1,238 @@
"""The CV match's fields: an agent's CV, languages, avoid and must texts,
migration 28, the deadline from host and model, the languages extract reads
with a level, and the language flag. Nothing here reaches a model."""
from __future__ import annotations
import json
import sqlite3
from datetime import UTC, datetime
from pathlib import Path
from typing import Any
import pytest
from pydantic import ValidationError
from app import prep
from app.config import Config
from app.jobs import extract
from app.models import CvLanguage, RoleLanguage, cv_score, iso_day, language_flag
from app.storage import Store
from desk import modeltask
from desk.models import Item
from desk.tools import agent
from prepped import upsert
@pytest.fixture
def store(tmp_path: Path) -> Store:
return Store(tmp_path / "roledesk.db")
def posting(source_id: str, body: str = "We build things.", **raw: object) -> Item:
item = Item(source_id=source_id, location=f"https://boards.example/{source_id}",
title="Head of Data", body=body,
published_at=datetime(2026, 9, 20, tzinfo=UTC), raw=dict(raw))
item.source = "search"
return item
# ------------------------------------------------------------------ agents
def test_an_agent_keeps_its_cv_fields_when_a_save_leaves_them_out(store: Store) -> None:
made = store.save_agent("Data", "data leadership", cv_text=" VP Data. ",
cv_languages=[CvLanguage(language="DA", level="native"),
CvLanguage(language="en", level="fluent"),
CvLanguage(language="da", level="basic")],
avoid_text="Pure reporting.", must_text="Permanent contract.")
assert made.cv_text == "VP Data."
assert [(lang.language, lang.level) for lang in made.cv_languages] == \
[("da", "native"), ("en", "fluent")]
assert (made.avoid_text, made.must_text) == ("Pure reporting.", "Permanent contract.")
kept = store.save_agent("Data", "data leadership", agent_id=made.id)
assert (kept.cv_text, kept.avoid_text, kept.must_text) == \
("VP Data.", "Pure reporting.", "Permanent contract.")
assert len(kept.cv_languages) == 2
cleared = store.save_agent("Data", "data leadership", agent_id=made.id, cv_text="",
cv_languages=[], avoid_text="", must_text="")
assert (cleared.cv_text, cleared.cv_languages, cleared.avoid_text, cleared.must_text) \
== ("", [], "", "")
def test_a_cv_language_is_a_code_and_a_level_of_the_ladder() -> None:
with pytest.raises(ValidationError):
CvLanguage(language="Danish", level="native")
with pytest.raises(ValidationError):
CvLanguage(language="da", level="good")
def test_new_agents_have_judge_limits_by_default(store: Store) -> None:
made = store.save_agent("Data")
assert (made.limits.judge_top, made.limits.judge_min) == (150, 0.55)
# --------------------------------------------------------------- migration
def test_v28_adds_the_cv_columns_and_the_host_deadline_and_runs_twice(tmp_path: Path) -> None:
path = tmp_path / "v27.db"
store = Store(path)
store.save_agent("Data", "data leadership")
store.upsert_role(posting("a", applicationDeadline="2026-10-18T00:00:00"))
store.upsert_role(posting("b"))
with sqlite3.connect(path) as conn:
conn.execute("DROP INDEX idx_roles_deadline")
conn.execute("DROP INDEX idx_role_judgements_agent")
conn.execute("DROP TABLE role_judgements")
for column in ("languages_required", "deadline_host", "deadline_model", "deadline"):
conn.execute(f"ALTER TABLE roles DROP COLUMN {column}")
for column in ("cv_text", "cv_languages", "avoid_text", "must_text"):
conn.execute(f"ALTER TABLE agents DROP COLUMN {column}")
conn.execute("PRAGMA user_version = 27")
migrated = Store(path)
assert migrated.migrations_applied == [28]
rows = {r["source_id"]: r for r in migrated._rows(
"SELECT source_id, languages_required, deadline_host, deadline FROM roles")}
assert (rows["a"]["deadline_host"], rows["a"]["deadline"]) == ("2026-10-18", None)
assert rows["a"]["languages_required"] is None and rows["b"]["deadline_host"] == ""
[data] = migrated.agents()
assert (data.cv_text, data.cv_languages, data.avoid_text, data.must_text) == \
("", [], "", "")
prep.run(migrated, remap=True, dupes=False)
assert migrated.role(posting("a").uid).deadline == "2026-10-18"
with sqlite3.connect(path) as conn:
conn.execute("PRAGMA user_version = 27")
assert Store(path).migrations_applied == [28]
# ---------------------------------------------------------------- deadline
def test_iso_day_reads_a_date_or_a_timestamp() -> None:
assert iso_day("2026-10-18T00:00:00") == "2026-10-18"
assert iso_day("2026-10-18") == "2026-10-18"
assert iso_day("18.10.2026") == iso_day("2026-13-01") == iso_day(None) == ""
def test_the_deadline_is_the_hosts_first_then_the_models(store: Store) -> None:
host = posting("host", applicationDeadline="2026-10-18T00:00:00")
model = posting("model")
upsert(store, host)
upsert(store, model)
store.set_extracted(host.uid, {"deadline": "2026-11-01"})
store.set_extracted(model.uid, {"deadline": "2026-11-01"})
prep.run_roles(store, [host.uid, model.uid])
assert store.role(host.uid).deadline == "2026-10-18"
assert store.role(model.uid).deadline == "2026-11-01"
# A pull whose host no longer states one leaves the model's.
upsert(store, posting("host"))
assert store.role(host.uid).deadline == "2026-11-01"
# ----------------------------------------------------------------- extract
class Model:
def __init__(self, reply: Any) -> None:
self.reply = reply
self.asked: list[list[dict[str, Any]]] = []
def __call__(self, settings: Any, **kwargs: Any) -> agent.Reply:
roles = kwargs["question"]["roles"]
self.asked.append(roles)
return agent.Reply(data={"roles": self.reply(roles)}, backend="api")
def answer(uid: str, **fields: Any) -> dict[str, Any]:
return {"uid": uid, "place": "", "country": "", "remote": "", "remote_where": [],
"seniority": "director", "salary_min": None, "salary_max": None,
"currency": "", "function": "", "permit": "", "languages_required": [],
"deadline": "", **fields}
def test_extract_reads_levels_kinds_and_the_deadline(
store: Store, tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
item = posting("1")
upsert(store, item)
store.replace_scores(store.save_agent("Data", "data").id, [(item.uid, 0.9, "")])
model = Model(lambda roles: [answer(r["uid"], deadline="2026-10-30", languages_required=[
{"language": "DA", "level": "fluent", "kind": "required"},
{"language": "en", "level": "", "kind": "required"},
{"language": "sv", "level": "excellent", "kind": "plus"},
{"language": "da", "level": "basic", "kind": "plus"},
{"language": "danish", "level": "", "kind": "required"}]) for r in roles])
monkeypatch.setattr(agent, "backend", lambda settings: "api")
monkeypatch.setattr(agent, "ask_json", model)
config = Config(database=str(tmp_path / "roledesk.db"), extract={"enabled": True})
modeltask.run(extract.TASK, config, store)
role = store.role(item.uid)
assert role.languages == ["da", "en"]
assert [lang.model_dump() for lang in role.languages_required] == [
{"language": "da", "level": "fluent", "kind": "required"},
{"language": "en", "level": "", "kind": "required"},
{"language": "sv", "level": "", "kind": "plus"}]
assert role.deadline == "2026-10-30"
assert model.asked[0][0]["posted"] == role.posted_at[:10]
assert store.extract_pending() == 0
def test_a_role_extracted_before_is_read_once_more_for_the_new_fields(store: Store) -> None:
item = posting("1")
upsert(store, item)
store.replace_scores(store.save_agent("Data", "data").id, [(item.uid, 0.9, "")])
store.set_extracted(item.uid, {"seniority": "director"})
assert store.extract_pending() == 0
with store.transaction() as conn:
conn.execute("UPDATE roles SET languages_required = NULL")
assert store.extract_pending() == 1
def test_the_excerpt_keeps_the_opening_and_the_closing_lines() -> None:
short = "word " * 100
assert extract.excerpt(short) == short.strip()
long = "A" * 3000 + " " + "Z" * 3000
cut = extract.excerpt(long)
assert cut.startswith("A" * extract.HEAD) and cut.endswith("Z" * extract.TAIL)
assert len(cut) == extract.HEAD + extract.TAIL + 3
# ------------------------------------------------------------------- flags
def langs(*pairs: tuple[str, str]) -> list[CvLanguage]:
return [CvLanguage(language=code, level=level) for code, level in pairs]
def asks(*triples: tuple[str, str, str]) -> list[RoleLanguage]:
return [RoleLanguage(language=code, level=level, kind=kind) for code, level, kind in triples]
def test_the_language_flag() -> None:
mine = langs(("da", "native"), ("en", "fluent"), ("de", "professional"))
assert language_flag(asks(("sv", "fluent", "required")), mine) == \
("missing", "requires Swedish (fluent); yours: none")
assert language_flag(asks(("de", "fluent", "required")), mine) == \
("below", "requires German (fluent); yours: professional")
assert language_flag(asks(("de", "", "required"), ("da", "", "plus")), mine) == \
("plus", "Danish is a plus; yours: native")
assert language_flag(asks(("en", "fluent", "required"), ("fr", "", "plus")), mine) == \
("", "")
assert language_flag(asks(("sv", "fluent", "required")), []) == ("", "")
def test_the_cv_score_weighs_skills_and_career_above_experience() -> None:
assert cv_score(100, 100, 100) == 100.0
assert cv_score(80, 60, 40) == round((2400 + 1500 + 1200) / 85, 1)
def test_a_stored_language_list_survives_a_bad_entry(store: Store) -> None:
item = posting("1")
upsert(store, item)
with store.transaction() as conn:
conn.execute("UPDATE roles SET languages_required = ? WHERE uid = ?",
(json.dumps([{"language": "da", "level": "fluent", "kind": "required"},
{"language": "en", "level": "superb"}]), item.uid))
assert [lang.language for lang in store.role(item.uid).languages_required] == ["da"]
+1 -1
View File
@@ -261,7 +261,7 @@ def test_a_v17_file_gains_dup_of_with_its_first_rebuild(tmp_path: Path) -> None:
migrated = Store(path)
assert migrated.migrations_applied == [18, 19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
assert dup_of(migrated) == {first.uid: None, second.uid: first.uid}
with sqlite3.connect(path) as conn:
assert conn.execute("SELECT 1 FROM sqlite_master WHERE name = 'idx_roles_dup_of'"
+3 -2
View File
@@ -32,7 +32,7 @@ def answer(uid: str, **fields: Any) -> dict[str, Any]:
empty: dict[str, Any] = {
"place": "", "country": "", "remote": "", "seniority": "", "salary_min": None,
"salary_max": None, "currency": "", "function": "", "permit": "",
"languages_required": []}
"languages_required": [], "deadline": ""}
return {"uid": uid, **empty, **fields}
@@ -139,7 +139,8 @@ def test_closed_and_answered_roles_are_not_offered(
conn.execute("UPDATE roles SET closed_at = '2026-09-22T00:00:00+00:00' "
"WHERE source_id = 'closed'")
conn.execute("UPDATE roles SET extracted_at = '2026-09-22T00:00:00+00:00', "
"seniority_model = 'senior', remote_where_model = '[]' "
"seniority_model = 'senior', remote_where_model = '[]', "
"languages_required = '[]' "
"WHERE source_id = 'done'")
model = stub(monkeypatch, lambda roles: [answer(r["uid"]) for r in roles])
+1 -1
View File
@@ -434,7 +434,7 @@ def test_migration_23_gives_every_agent_no_terms(tmp_path: Path) -> None:
conn.execute("ALTER TABLE agents DROP COLUMN fetch_terms")
conn.execute("PRAGMA user_version = 22")
migrated = Store(path)
assert migrated.migrations_applied == [23, 24, 25, 26, 27]
assert migrated.migrations_applied == [23, 24, 25, 26, 27, 28]
assert migrated.agents()[0].fetch_terms == []
+1 -1
View File
@@ -120,7 +120,7 @@ def test_migration_24_splits_the_tag_and_moves_terms_off(tmp_path: Path) -> None
conn.execute("UPDATE companies SET all_roles=1 WHERE key='mistral-ai'")
conn.execute("PRAGMA user_version = 23")
migrated = Store(path)
assert migrated.migrations_applied == [24, 25, 26, 27]
assert migrated.migrations_applied == [24, 25, 26, 27, 28]
flags = {c.key: (c.all_countries, c.all_roles) for c in migrated.companies()}
assert flags == {"mistral-ai": (1, 1), "acme": (0, 1), "plain": (0, 0)}
params = {s.target: s.params for s in migrated.subscriptions(source="ats")}
+1 -1
View File
@@ -316,7 +316,7 @@ def test_migration_22_fills_the_host_words_from_raw(tmp_path: Path) -> None:
migrated = Store(path)
assert migrated.migrations_applied == [22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [22, 23, 24, 25, 26, 27, 28]
one = row(migrated, "1")
assert (one["seniority_host"], one["employment_host"], one["remote_host"],
one["country_host"]) == ("senior", "full_time", "true", "DK")
+2 -2
View File
@@ -530,7 +530,7 @@ def test_migration_27_keeps_an_older_readers_remote_as_the_hosts_word(tmp_path:
migrated = Store(path)
assert migrated.migrations_applied == [27]
assert migrated.migrations_applied == [27, 28]
assert migrated.gazetteer().anywhere is not None
row = migrated._row("SELECT remote_host, remote_model, remote_where_model FROM roles "
"WHERE uid = ?", (board,))
@@ -540,7 +540,7 @@ def test_migration_27_keeps_an_older_readers_remote_as_the_hosts_word(tmp_path:
assert area(migrated, search.uid) == ("", "")
with sqlite3.connect(path) as conn:
conn.execute("PRAGMA user_version = 26")
assert Store(path).migrations_applied == [27]
assert Store(path).migrations_applied == [27, 28]
def test_a_role_waits_for_extract_until_it_was_asked_where(store: Store) -> None:
+10 -10
View File
@@ -70,7 +70,7 @@ def test_schema_creates_on_an_empty_file(tmp_path: Path) -> None:
"application_events", "agents", "scores", "state"} <= tables
assert "idx_roles_cluster" in indexes
assert store.migrations_applied == []
assert LATEST == 27
assert LATEST == 28
# A second open finds it current and applies nothing.
assert Store(path).migrations_applied == []
@@ -173,7 +173,7 @@ def test_a_v1_file_migrates_to_the_shape_a_new_file_has(tmp_path: Path) -> None:
assert Store(old).migrations_applied == list(range(2, LATEST + 1))
# ALTER appends, so the order differs; every read is by column name.
for table in ("roles", "agents", "agent_prompts", "scores", "vectors", "companies",
"role_tags", "agent_companies", "role_votes"):
"role_tags", "agent_companies", "role_votes", "role_judgements"):
assert sorted(_columns(old, table)) == sorted(_columns(fresh, table))
@@ -187,7 +187,7 @@ def test_a_v14_file_gains_the_watchlist_columns_and_leaves_companies_unrated(
migrated = Store(path)
assert migrated.migrations_applied == [15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
acme = migrated.company("acme")
assert acme is not None
assert (acme.ats, acme.careers_url) == ("teamtailor", "https://a.example")
@@ -392,7 +392,7 @@ def test_v14_makes_each_intent_the_agents_one_prompt_with_its_vector(tmp_path: P
migrated = Store(path)
assert migrated.migrations_applied == [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
assert migrated.agent(backend.id).prompts == [ # type: ignore[union-attr]
Prompt(name="Backend", text="python backend")]
assert migrated.agent(blank.id).prompts == [] # type: ignore[union-attr]
@@ -416,7 +416,7 @@ def test_v16_gives_a_company_no_reason_and_keeps_its_row(tmp_path: Path) -> None
migrated = Store(path)
assert migrated.migrations_applied == [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
acme = migrated.company("acme")
assert acme is not None
assert (acme.careers_url, acme.notes, acme.hand_reason) == (
@@ -445,7 +445,7 @@ def test_v17_folds_every_place_and_levels_the_titles_extract_has_not_read(
migrated = Store(path)
assert migrated.migrations_applied == [17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
with sqlite3.connect(path) as conn:
rows = {uid: (fold, level) for uid, fold, level in conn.execute(
"SELECT uid, place_fold, seniority FROM roles")}
@@ -688,7 +688,7 @@ def test_v19_gives_every_role_its_places_and_splits_its_level(tmp_path: Path) ->
migrated = Store(path)
assert migrated.migrations_applied == [19, 20, 21, 22, 23, 24, 25, 26, 27]
assert migrated.migrations_applied == [19, 20, 21, 22, 23, 24, 25, 26, 27, 28]
assert [p.name for p in migrated.role_places(joined.uid)] == ["Copenhagen", "Malmö"]
assert [p.name for p in migrated.role_places(detailed.uid)] == ["Aarhus"]
assert migrated.role_places(stated.uid) == []
@@ -730,7 +730,7 @@ def test_v25_clears_the_model_levels_the_ladder_splits_and_sets_the_level(
migrated = Store(path)
assert migrated.migrations_applied == [25, 26, 27]
assert migrated.migrations_applied == [25, 26, 27, 28]
with sqlite3.connect(path) as conn:
found = {sid: rest for sid, *rest in conn.execute(
"SELECT source_id, seniority_model, seniority_stated, seniority_title, seniority "
@@ -748,12 +748,12 @@ def test_v26_adds_posting_url_empty_and_runs_twice_without_harm(tmp_path: Path)
migrated = Store(path)
assert migrated.migrations_applied == [26, 27]
assert migrated.migrations_applied == [26, 27, 28]
role = migrated.role(posting("a").uid)
assert role is not None and role.posting_url == ""
with sqlite3.connect(path) as conn:
conn.execute("PRAGMA user_version = 25") # the column is there already
assert Store(path).migrations_applied == [26, 27]
assert Store(path).migrations_applied == [26, 27, 28]
def test_a_pull_keeps_the_posting_url(store: Store) -> None:
+6 -3
View File
@@ -1182,8 +1182,10 @@ def test_an_agents_limits_are_saved_listed_and_checked(client: TestClient) -> No
0..1 is refused."""
default = client.post("/api/agents", json={"name": "Wide"}).json()
assert default["limits"] == {"extract_top": 500, "extract_min": 0.55,
"summarise_top": 30, "summarise_min": 0.65}
narrow = {"extract_top": 50, "extract_min": 0.6, "summarise_top": 5, "summarise_min": 0.7}
"summarise_top": 30, "summarise_min": 0.65,
"judge_top": 150, "judge_min": 0.55}
narrow = {"extract_top": 50, "extract_min": 0.6, "summarise_top": 5, "summarise_min": 0.7,
"judge_top": 20, "judge_min": 0.6}
saved = client.post("/api/agents", json={"name": "Narrow", "limits": narrow})
assert saved.status_code == 200 and saved.json()["limits"] == narrow
assert [a["limits"]["extract_top"] for a in client.get("/api/agents").json()] == [50, 500]
@@ -1227,7 +1229,8 @@ def test_list_roles_serves_the_row_projection_and_not_the_big_columns(
"uid", "company_key", "company", "title", "place", "country", "remote", "function",
"seniority", "salary_min", "salary_max", "currency", "salary_text", "salary_state",
"original_url", "posting_url", "posted_at", "score", "gated_reason", "prompt", "vote",
"voted_at", "read_at", "tags", "excerpt", "copies", "remote_where",
"voted_at", "read_at", "tags", "excerpt", "copies", "remote_where", "deadline",
"lang_flag", "lang_note", "cv_score",
}
assert row["uid"] == uid
assert row["company"] == "Acme" and row["place"] == "Berlin, Germany"