505 lines
24 KiB
Python
505 lines
24 KiB
Python
"""Subbaustein-Qualität: Varianten-Konsens, Seed-Garantie, Nachfass, Outline-Review."""
|
||
|
||
import json
|
||
import re
|
||
|
||
import numpy as np
|
||
import pytest
|
||
|
||
import blocks as blx
|
||
from pipeline import GenContext
|
||
|
||
TOPIC = "t"
|
||
_MD_PATH = re.compile(r"(/\S+\.md)")
|
||
|
||
|
||
# ── _variant_clusters (pure) ─────────────────────────────────────────────────────────
|
||
|
||
def _sims(pairs, n):
|
||
m = np.eye(n)
|
||
for i, j, v in pairs:
|
||
m[i][j] = m[j][i] = v
|
||
return m
|
||
|
||
|
||
def test_variant_clusters_folds_paraphrases():
|
||
titles = ["Harte Umbrüche brauchen Marker", "Harte Umbrüche erfordern explizite Marker!",
|
||
"Tabs werden expandiert"]
|
||
cl = blx._variant_clusters(titles, [1, 1, 1], _sims([(0, 1, 0.95)], 3))
|
||
by_rep = {c["rep"]: c for c in cl}
|
||
assert by_rep[1]["mentions"] == 2 and sorted(by_rep[1]["members"]) == [0, 1] # longest wins
|
||
assert by_rep[2]["mentions"] == 1
|
||
|
||
|
||
def test_variant_clusters_negation_guard():
|
||
titles = ["Fenced können Absätze unterbrechen", "Fenced können Absätze nicht unterbrechen"]
|
||
cl = blx._variant_clusters(titles, [1, 1], _sims([(0, 1, 0.95)], 2))
|
||
assert len(cl) == 2 # antonyms never merge, no matter the cosine
|
||
|
||
|
||
# ── _subblocks_block integration (fake race + fake embeddings) ──────────────────────
|
||
|
||
def _fake_sims(texts):
|
||
"""Markertoken matrix: same first word → 0.95, else 0."""
|
||
n = len(texts)
|
||
m = np.eye(n)
|
||
key = lambda t: t.split()[0].casefold()
|
||
for i in range(n):
|
||
for j in range(n):
|
||
if i != j and key(texts[i]) == key(texts[j]):
|
||
m[i][j] = 0.95
|
||
return m
|
||
|
||
|
||
def _mk_race(finder_by_agent):
|
||
"""Key-routed _race fake. Finder round 1 → scripted per-agent subs; later finder and
|
||
catch-up rounds → nothing; clarify judges echo the consensus lines from their prompt."""
|
||
prompts = []
|
||
|
||
async def fake_race(topic, label, slots, quorum, timeout, provider, on_update=None,
|
||
cancelled=None, *, grace=None, min_runtime=None, max_runtime=None, late=None):
|
||
outs = []
|
||
for slot in slots:
|
||
key, prompt = slot["key"], slot["prompt"]
|
||
prompts.append((key, prompt))
|
||
fake_race.slots_seen.append(slot)
|
||
if "-subblock-final-" in key:
|
||
# no-tool judges reply as TEXT; the payload sink writes the j-file itself
|
||
kons = re.search(r"Konsens \(≥2 finders\):\n(.*?)\nUnsicher", prompt, re.S)
|
||
subs = [l[2:] for l in (kons.group(1).splitlines() if kons else [])
|
||
if l.startswith("- ") and l != "- (keiner)"]
|
||
if subs:
|
||
text = "<!-- block: Alpha -->\n" + "\n".join(f"- {s}" for s in subs)
|
||
outs.append(slot["payload"]((0, text, "")))
|
||
continue
|
||
if "-r1-" in key:
|
||
agent = int(key.rsplit("-", 1)[1])
|
||
subs = finder_by_agent.get(agent) or []
|
||
if subs: # Finder antworten als TEXT (Marker-Format), kein out_path mehr
|
||
text = "<!-- block: Alpha -->\n" + "\n".join(f"- {s}" for s in subs)
|
||
outs.append(slot["payload"]((0, text, "")))
|
||
outs = [o for o in outs if o]
|
||
return outs or None
|
||
fake_race.slots_seen = []
|
||
return fake_race, prompts
|
||
|
||
|
||
@pytest.fixture
|
||
def sub_env(testdb, tmp_path, monkeypatch):
|
||
monkeypatch.setattr(blx, "EMBEDDING_AKTIV", True)
|
||
monkeypatch.setattr(blx.embedding, "available", lambda: True)
|
||
monkeypatch.setattr(blx.embedding, "embed_sims", _fake_sims)
|
||
ctx = GenContext(topic=TOPIC, provider="claude", is_cancelled=lambda: False)
|
||
files = {"arbeit": tmp_path}
|
||
return testdb, ctx, files
|
||
|
||
|
||
async def _run(ctx, files, monkeypatch, finder_by_agent, seeds=None):
|
||
fake, prompts = _mk_race(finder_by_agent)
|
||
monkeypatch.setattr(blx, "_race", fake)
|
||
raw = await blx._subblocks_block(ctx, lambda *a, **k: None, files, {1: "Alpha — Grundkonzept"},
|
||
"", wipe=False, ns="x-", seeds=seeds)
|
||
return raw, prompts
|
||
|
||
|
||
async def test_variant_consensus_end_to_end(sub_env, monkeypatch):
|
||
"""3 Einzelfunde in 3 Formulierungen → EIN consensus-Repräsentant; Varianten gehen
|
||
nicht als „Unsicher" ins Panel."""
|
||
db, ctx, files = sub_env
|
||
raw, prompts = await _run(ctx, files, monkeypatch, {
|
||
1: ["Umbruch braucht Marker"],
|
||
2: ["Umbruch erfordert explizite Marker!"],
|
||
3: ["Umbruch verlangt zwei Leerzeichen als Marker"],
|
||
})
|
||
assert raw == {"Alpha": ["Umbruch verlangt zwei Leerzeichen als Marker"]} # longest = rep
|
||
rows = await db.list_subblocks(TOPIC, "alpha")
|
||
status = sorted(r["status"] for r in rows)
|
||
assert status == ["consensus", "variant", "variant"]
|
||
clarify_prompts = [p for k, p in prompts if "-subblock-final-" in k]
|
||
assert clarify_prompts and "Umbruch braucht Marker" not in clarify_prompts[0]
|
||
|
||
|
||
async def test_seed_promotes_single_find(sub_env, monkeypatch):
|
||
"""Seed deckt einen verworfenen Einzelfund lexikalisch → Promotion zu consensus."""
|
||
db, ctx, files = sub_env
|
||
raw, _ = await _run(ctx, files, monkeypatch, {
|
||
1: ["Alpha Grundlagen", "Zeilenumbruch Regeln im Detail"],
|
||
2: ["Alpha Grundlagen"],
|
||
}, seeds=["Zeilenumbruch Regeln"])
|
||
assert "Zeilenumbruch Regeln im Detail" in raw["Alpha"]
|
||
row = next(r for r in await db.list_subblocks(TOPIC, "alpha")
|
||
if r["sub_norm"] == "zeilenumbruch regeln im detail")
|
||
assert row["status"] == "consensus"
|
||
|
||
|
||
async def test_seed_inserted_when_nothing_found(sub_env, monkeypatch):
|
||
"""Seed ohne jeden Fund wird als eigener consensus-Sub eingefügt (Facts-Gate prüft später)."""
|
||
db, ctx, files = sub_env
|
||
raw, _ = await _run(ctx, files, monkeypatch, {
|
||
1: ["Alpha Grundlagen"], 2: ["Alpha Grundlagen"],
|
||
}, seeds=["Fußnoten Syntax"])
|
||
assert "Fußnoten Syntax" in raw["Alpha"]
|
||
row = next(r for r in await db.list_subblocks(TOPIC, "alpha")
|
||
if r["sub_title"] == "Fußnoten Syntax")
|
||
assert row["status"] == "consensus"
|
||
|
||
|
||
async def test_seed_covered_no_duplicate(sub_env, monkeypatch):
|
||
"""Seed lexikalisch von einem consensus-Sub abgedeckt → nichts eingefügt."""
|
||
db, ctx, files = sub_env
|
||
raw, _ = await _run(ctx, files, monkeypatch, {
|
||
1: ["Tabs werden zu Leerzeichen expandiert"], 2: ["Tabs werden zu Leerzeichen expandiert"],
|
||
}, seeds=["Tabs"])
|
||
assert raw == {"Alpha": ["Tabs werden zu Leerzeichen expandiert"]}
|
||
|
||
|
||
async def test_wipe_false_is_idempotent(sub_env, monkeypatch):
|
||
"""Zweiter Karten-Lauf kumuliert keine Mentions (per-Block-Wipe)."""
|
||
db, ctx, files = sub_env
|
||
await _run(ctx, files, monkeypatch, {1: ["Alpha Grundlagen"], 2: ["Alpha Grundlagen"]})
|
||
first = {r["sub_norm"]: r["mentions"] for r in await db.list_subblocks(TOPIC, "alpha")}
|
||
await _run(ctx, files, monkeypatch, {1: ["Alpha Grundlagen"], 2: ["Alpha Grundlagen"]})
|
||
second = {r["sub_norm"]: r["mentions"] for r in await db.list_subblocks(TOPIC, "alpha")}
|
||
assert first == second
|
||
|
||
|
||
async def test_catchup_adds_and_stops(sub_env, monkeypatch, tmp_path):
|
||
"""Block unter SUBBLOCK_MIN: Nachfass-Runde findet Neues → eigenes Final-File,
|
||
Konsens wächst; zweite Runde ohne Neues → Ende."""
|
||
db, ctx, files = sub_env
|
||
fake, prompts = _mk_race({1: ["Alpha Grundlagen"], 2: ["Alpha Grundlagen"]})
|
||
base = fake
|
||
hit = {"n": 0}
|
||
|
||
async def with_catchup(topic, label, slots, *a, **k):
|
||
if any("-subblock-x" in s["key"] for s in slots):
|
||
hit["n"] += 1
|
||
if hit["n"] == 1: # first catch-up round: both agents agree on one new sub
|
||
text = "<!-- block: Alpha -->\n- Vertiefung der Konzepte"
|
||
return [slot["payload"]((0, text, "")) for slot in slots[:2]]
|
||
return None
|
||
return await base(topic, label, slots, *a, **k)
|
||
|
||
monkeypatch.setattr(blx, "_race", with_catchup)
|
||
raw = await blx._subblocks_block(ctx, lambda *a, **k: None, files, {1: "Alpha — Grundkonzept"},
|
||
"", wipe=False, ns="x-")
|
||
assert set(raw["Alpha"]) == {"Alpha Grundlagen", "Vertiefung der Konzepte"}
|
||
assert (tmp_path / "subblock-final-c1-x1.md").exists()
|
||
assert hit["n"] == 2 # round 2 ran, found nothing, loop ended
|
||
|
||
|
||
# ── Outline-Review ───────────────────────────────────────────────────────────────────
|
||
|
||
def test_outline_review_schema():
|
||
valid = {1, 2, 3, 4, 5, 6}
|
||
ok = blx._outline_review_schema({"moves": {"3": 2}}, valid, 2, 6)
|
||
assert ok == {3: 2}
|
||
assert blx._outline_review_schema({"moves": {}}, valid, 2, 6) == {}
|
||
assert blx._outline_review_schema({"moves": {"9": 1}}, valid, 2, 6) is None # unknown block
|
||
assert blx._outline_review_schema({"moves": {"1": 5}}, valid, 2, 6) is None # chapter range
|
||
assert blx._outline_review_schema({"moves": {"1": 2, "2": 2, "3": 2}}, valid, 2, 6) is None # mass move
|
||
assert blx._outline_review_schema({"chapters": []}, valid, 2, 6) is None
|
||
|
||
|
||
async def test_outline_review_moves_block(testdb, tmp_path, monkeypatch):
|
||
"""Review verschiebt einen fehlplatzierten Block; kaputtes Review lässt den Plan unverändert."""
|
||
entries = {i: f"Block {i} — d" for i in range(1, 7)}
|
||
slots = [tmp_path / f"outline-{i}.json" for i in (1, 2, 3)]
|
||
plan_a = {"chapters": [{"title": "K1", "numbers": [1, 2, 6]}, {"title": "K2", "numbers": [3, 4, 5]}]}
|
||
for p in slots[:2]:
|
||
p.write_text(json.dumps(plan_a), encoding="utf-8")
|
||
files = {"arbeit": tmp_path, "outline": tmp_path / "outline.json", "outline_slots": slots,
|
||
"facts": tmp_path / "facts.json"}
|
||
|
||
review_out = {"val": {"moves": {"6": 2}}}
|
||
|
||
async def fake_slot(ctx, label, *, key, prompt, role, capabilities, payload, timeout, on_line=None):
|
||
out = None
|
||
if key.endswith("outline-prereqs"):
|
||
out = {"prereqs": {}}
|
||
elif key.endswith("outline-judge"):
|
||
out = plan_a
|
||
elif key.endswith("outline-review"):
|
||
out = review_out["val"]
|
||
if out is not None: # neue Semantik: Antwort als TEXT, der Sink persistiert
|
||
return "ok", payload((0, json.dumps(out), ""))
|
||
return "ok", payload(None)
|
||
|
||
monkeypatch.setattr(blx, "run_single_slot", fake_slot)
|
||
ctx = GenContext(topic=TOPIC, provider="claude", is_cancelled=lambda: False)
|
||
plan = await blx._outline_block(ctx, lambda *a, **k: None, files, entries, "")
|
||
assert plan["chapters"][0]["numbers"] == [1, 2]
|
||
assert plan["chapters"][1]["numbers"] == [3, 4, 5, 6]
|
||
|
||
# broken review (mass move) → schema rejects, plan unchanged
|
||
review_out["val"] = {"moves": {"1": 2, "2": 2, "3": 1}}
|
||
(tmp_path / "outline-review.json").unlink()
|
||
(tmp_path / "outline.json").unlink()
|
||
plan2 = await blx._outline_block(ctx, lambda *a, **k: None, files, entries, "")
|
||
assert plan2["chapters"][0]["numbers"] == [1, 2, 6]
|
||
|
||
async def test_paraphrase_saturation_stops_early(sub_env, monkeypatch):
|
||
"""Runde 2 liefert nur eine Paraphrase → zählt nicht als neu, Schleife endet ohne r3.
|
||
Die Paraphrase liegt trotzdem in der DB (Mention fürs Cluster-Voting)."""
|
||
db, ctx, files = sub_env
|
||
base_fake, prompts = _mk_race({1: ["Umbruch braucht Marker"], 2: ["Umbruch braucht Marker"]})
|
||
|
||
async def with_r2(topic, label, slots, *a, **k):
|
||
if any("-r2-" in s["key"] for s in slots):
|
||
text = "<!-- block: Alpha -->\n- Umbruch erfordert explizite Marker!"
|
||
return [slot["payload"]((0, text, "")) for slot in slots[:2]]
|
||
return await base_fake(topic, label, slots, *a, **k)
|
||
|
||
monkeypatch.setattr(blx, "_race", with_r2)
|
||
raw = await blx._subblocks_block(ctx, lambda *a, **k: None, files, {1: "Alpha — Grundkonzept"},
|
||
"", wipe=False, ns="x-")
|
||
assert raw["Alpha"] # Konsens steht
|
||
assert not any("-r3-" in k for k, _ in prompts) # Paraphrase hielt die Schleife NICHT am Leben
|
||
rows = await db.list_subblocks(TOPIC, "alpha")
|
||
assert any(r["sub_title"] == "Umbruch erfordert explizite Marker!" for r in rows)
|
||
|
||
|
||
async def test_round_cap_stops_endless_finders(sub_env, monkeypatch):
|
||
"""Jede Runde ein echt neues Konzept → hartes Cap stoppt bei SUBBLOCK_MAX_ROUNDS."""
|
||
db, ctx, files = sub_env
|
||
_, prompts = _mk_race({})
|
||
|
||
async def endless(topic, label, slots, *a, **k):
|
||
if "-subblock-final-" in slots[0]["key"]:
|
||
return None # panel fails → consensus fallback
|
||
outs = []
|
||
import re as _re
|
||
rn = _re.search(r"-r(\d+)-", slots[0]["key"])
|
||
n = rn.group(1) if rn else "x"
|
||
for slot in slots[:2]:
|
||
prompts.append((slot["key"], slot["prompt"]))
|
||
outs.append(slot["payload"]((0, f"<!-- block: Alpha -->\n- Konzept{n} ist eigenständig", "")))
|
||
return outs
|
||
|
||
monkeypatch.setattr(blx, "_race", endless)
|
||
await blx._subblocks_block(ctx, lambda *a, **k: None, files, {1: "Alpha — Grundkonzept"},
|
||
"", wipe=False, ns="x-")
|
||
max_round = max(int(k.split("-r")[1].split("-")[0]) for k, _ in prompts if "-r" in k and "-subblock-c" in k)
|
||
assert max_round == blx.SUBBLOCK_MAX_ROUNDS
|
||
|
||
|
||
# ── Inline-Evidenz für Judges (Token-Umbau) ──────────────────────────────────────────
|
||
|
||
def _corpus(tmp_path):
|
||
d = tmp_path / "korpus"
|
||
d.mkdir()
|
||
(d / "Skript.txt").write_text(
|
||
"Kapitel 1\nAlpha Grundlagen: der Kernbegriff.\nMehr Text dazu.\n\n"
|
||
"Kapitel 2\nGamma Randnotiz ohne Bezug.\n", encoding="utf-8")
|
||
(d / "Aufgaben.txt").write_text("Übung 1\nAlpha Vertiefung der Konzepte.\n", encoding="utf-8")
|
||
return d
|
||
|
||
|
||
def test_evidence_pack_selects_matching_sections(tmp_path):
|
||
d = _corpus(tmp_path)
|
||
pack = blx._evidence_pack(d, None, ["Alpha Grundlagen"])
|
||
assert "── Skript.txt" in pack and "Kernbegriff" in pack
|
||
pack2 = blx._evidence_pack(d, ["Aufgaben.txt"], ["Alpha"]) # genannte Quellen engen ein
|
||
assert "Skript.txt" not in pack2 and "Aufgaben.txt" in pack2
|
||
assert blx._evidence_pack(None, None, ["x"]) == "" # kein Korpus → Selbst-Recherche bleibt
|
||
|
||
|
||
def test_evidence_pack_budget_and_guarantee(tmp_path):
|
||
d = tmp_path / "korpus"
|
||
d.mkdir()
|
||
(d / "A.txt").write_text("Alpha wichtig. " * 50, encoding="utf-8")
|
||
(d / "B.txt").write_text("Beta anderes Thema. " * 50, encoding="utf-8")
|
||
pack = blx._evidence_pack(d, None, ["Alpha"], budget=10)
|
||
assert "Alpha" in pack # Abdeckungs-Garantie schlägt das Budget
|
||
assert "Beta" not in pack # Top-up respektiert das Budget
|
||
|
||
|
||
def test_cite_ref_parses_positions(tmp_path):
|
||
d = _corpus(tmp_path)
|
||
files = blx._corpus_files(d, None)
|
||
f, lo, hi = blx._cite_ref("Skript.txt, Übung 6.47, Z.2-3", files)
|
||
assert f.name == "Skript.txt" and (lo, hi) == (2, 3)
|
||
f2, lo2, hi2 = blx._cite_ref("Aufgaben.txt Zeile 2", files)
|
||
assert f2.name == "Aufgaben.txt" and lo2 == hi2 == 2
|
||
assert blx._cite_ref("Skript.txt, Übung 6.47", files) is None # keine Zeilenangabe
|
||
assert blx._cite_ref("Z.5 irgendwo", files) is None # keine Datei
|
||
# englische Zitierformen (Quellen sind nicht immer deutsch)
|
||
f3, lo3, hi3 = blx._cite_ref("Skript.txt, line 2", files)
|
||
assert f3.name == "Skript.txt" and lo3 == hi3 == 2
|
||
f4, lo4, hi4 = blx._cite_ref("Aufgaben.txt, lines 1-2", files)
|
||
assert f4.name == "Aufgaben.txt" and (lo4, hi4) == (1, 2)
|
||
|
||
|
||
def test_cited_evidence_lines_and_fallback(tmp_path):
|
||
d = _corpus(tmp_path)
|
||
ev = blx._cited_evidence(d, None, ["Skript.txt, Z.2"], ["Alpha"])
|
||
assert "── Skript.txt · Z." in ev and "Kernbegriff" in ev
|
||
ev2 = blx._cited_evidence(d, None, ["ohne Position"], ["Alpha Grundlagen"])
|
||
assert "Kernbegriff" in ev2 # Keyword-Fallback
|
||
|
||
|
||
def test_sink_json_writes_only_valid(tmp_path):
|
||
p = tmp_path / "level-final-c1.json"
|
||
ok = blx._sink_json((0, 'Vorab {"levels": {"1": "beginner"}} nach', ""), p,
|
||
lambda d: blx._levels_schema(d, {1}))
|
||
assert ok == {1: "beginner"}
|
||
assert json.loads(p.read_text(encoding="utf-8"))["levels"]["1"] == "beginner"
|
||
bad = blx._sink_json((0, "kein json", ""), tmp_path / "x.json", lambda d: d)
|
||
assert bad is None and not (tmp_path / "x.json").exists()
|
||
|
||
|
||
async def test_clarify_inline_evidence_no_tools(sub_env, monkeypatch, tmp_path):
|
||
"""Mit Korpus: Judges UND Finder bekommen Auszüge inline und laufen ohne Tools
|
||
(Text-Antwort); die Dateien schreibt die Engine. Der Finder verlor vorher 3–13
|
||
Tool-Runden pro Call mit der Material-Suche via glob/grep/bash."""
|
||
db, ctx, files = sub_env
|
||
d = _corpus(tmp_path)
|
||
monkeypatch.setattr(blx, "source_folder", lambda t: d)
|
||
monkeypatch.setattr(blx, "load_source", lambda t: {"type": "uni"})
|
||
fake, prompts = _mk_race({1: ["Alpha Grundlagen"], 2: ["Alpha Grundlagen"]})
|
||
monkeypatch.setattr(blx, "_race", fake)
|
||
raw = await blx._subblocks_block(ctx, lambda *a, **k: None, files, {1: "Alpha — Grundkonzept"},
|
||
"", wipe=False, ns="x-", sources=["Skript.txt"])
|
||
assert raw == {"Alpha": ["Alpha Grundlagen"]}
|
||
judges = [s for s in fake.slots_seen if "-subblock-final-" in s["key"]]
|
||
finders = [s for s in fake.slots_seen if "-r1-" in s["key"]]
|
||
assert judges and all(s["capabilities"] == "none" for s in judges)
|
||
assert "── Skript.txt" in judges[0]["prompt"]
|
||
assert "ls/find" not in judges[0]["prompt"] # keine Selbst-Recherche-Anweisung mehr
|
||
assert finders and all(s["capabilities"] == "none" for s in finders)
|
||
assert "── Skript.txt" in finders[0]["prompt"] # Auszüge inline statt Dateisystem-Suche
|
||
assert "web search" not in finders[0]["prompt"]
|
||
assert list(tmp_path.glob("subblock-final-*-j*.md")) # Engine persistiert die Judge-Antwort
|
||
|
||
|
||
async def test_panel_2of3_kehrt_bei_einigkeit_zurueck(tmp_path):
|
||
"""Zwei übereinstimmende Verdicts → Rückkehr ohne den langsamen Dritten; sein
|
||
Ergebnis wird detached nachpersistiert (Resume)."""
|
||
import asyncio as aio
|
||
gesunken = {}
|
||
|
||
async def judge(j, delay, antwort):
|
||
await aio.sleep(delay)
|
||
return (0, antwort, "")
|
||
|
||
tasks = {aio.create_task(judge(1, 0.01, "a")): 1,
|
||
aio.create_task(judge(2, 0.02, "a")): 2,
|
||
aio.create_task(judge(3, 5.0, "b")): 3}
|
||
|
||
def sink(j, r):
|
||
gesunken[j] = r[1]
|
||
|
||
import time
|
||
t0 = time.monotonic()
|
||
await blx._panel_2of3(tasks, sink, lambda: list(gesunken.values()), lambda s: s)
|
||
assert time.monotonic() - t0 < 1.0 # nicht auf j3 gewartet
|
||
assert gesunken == {1: "a", 2: "a"}
|
||
|
||
|
||
async def test_panel_2of3_dissens_wartet_auf_dritten():
|
||
"""Uneinige erste zwei → der dritte wird abgewartet (Mehrheit braucht ihn)."""
|
||
import asyncio as aio
|
||
gesunken = {}
|
||
|
||
async def judge(j, delay, antwort):
|
||
await aio.sleep(delay)
|
||
return (0, antwort, "")
|
||
|
||
tasks = {aio.create_task(judge(1, 0.01, "a")): 1,
|
||
aio.create_task(judge(2, 0.02, "b")): 2,
|
||
aio.create_task(judge(3, 0.1, "a")): 3}
|
||
await blx._panel_2of3(tasks, lambda j, r: gesunken.__setitem__(j, r[1]),
|
||
lambda: list(gesunken.values()), lambda s: s)
|
||
assert gesunken == {1: "a", 2: "b", 3: "a"}
|
||
|
||
|
||
async def test_facts_check_inline_cited_evidence(sub_env, monkeypatch, tmp_path):
|
||
"""Facts-Check: zitierte Zeilenbereiche gehen inline mit, Judge läuft ohne Tools,
|
||
die Check-Datei schreibt die Engine aus der Text-Antwort."""
|
||
db, ctx, files = sub_env
|
||
d = _corpus(tmp_path)
|
||
facts = {"facts": [{"block": "Alpha", "subblock": "Sub Eins", "key_points": ["k"],
|
||
"prerequisites": "", "hurdles": "",
|
||
"cited_facts": [{"text": "T", "source": "Skript.txt, Z.2"}],
|
||
"example_idea": ""}]}
|
||
|
||
sh = blx._subs_hash({"Alpha": ["Sub Eins"]}) # Resume-Dateien tragen den Sub-Satz-Hash
|
||
|
||
async def fake_slot(ctx2, label, *, key, prompt, role, capabilities, payload, timeout, on_line=None):
|
||
if "-facts-erg-" in key:
|
||
return blx.FAILED, None
|
||
(tmp_path / f"facts-{sh}-c0.json").write_text(json.dumps(facts), encoding="utf-8")
|
||
return blx.OK, None
|
||
|
||
seen = []
|
||
|
||
async def fake_agent(key, prompt, timeout, provider="", role="", capabilities="", scope=None, label="", **kw):
|
||
seen.append((key, capabilities, prompt))
|
||
return (0, '{"ok": true}', "")
|
||
|
||
monkeypatch.setattr(blx, "run_single_slot", fake_slot)
|
||
monkeypatch.setattr(blx, "run_agent", fake_agent)
|
||
res = await blx._facts_block(ctx, lambda *a, **k: None, {"arbeit": tmp_path},
|
||
{"Alpha": ["Sub Eins"]}, {"type": "uni"}, d, "", ns="x-")
|
||
assert res is not None
|
||
facts_map, discarded = res
|
||
assert "Alpha" in facts_map and not discarded
|
||
assert len(seen) == blx.FACTS_CHECK_PANEL
|
||
key, caps, prompt = seen[0]
|
||
assert caps == "none" and "── Skript.txt · Z." in prompt
|
||
assert (tmp_path / f"facts-check-{sh}-c0-j1.json").exists() # Engine persistiert die Antwort
|
||
|
||
|
||
async def test_facts_find_inline_evidence_no_tools(sub_env, monkeypatch, tmp_path):
|
||
"""Facts find/erg mit Korpus: Auszüge inline, Agent ohne Tools — Tool-Agenten
|
||
verloren sich in Reasoning-Schleifen und endeten mit leerem Turn (Retry-Wellen)."""
|
||
db, ctx, files = sub_env
|
||
d = _corpus(tmp_path)
|
||
seen = []
|
||
facts = {"facts": [{"block": "Alpha", "subblock": "Sub Eins", "key_points": ["k"],
|
||
"prerequisites": "", "hurdles": "",
|
||
"cited_facts": [{"text": "T", "source": "Skript.txt, Z.2"}],
|
||
"example_idea": ""}]}
|
||
|
||
async def fake_slot(ctx2, label, *, key, prompt, role, capabilities, payload, timeout, on_line=None):
|
||
seen.append((key, capabilities, prompt))
|
||
return blx.OK, payload((0, json.dumps(facts), ""))
|
||
|
||
async def fake_agent(key, prompt, timeout, **kw): # Check-Panel
|
||
return (0, '{"ok": true}', "")
|
||
|
||
monkeypatch.setattr(blx, "run_single_slot", fake_slot)
|
||
monkeypatch.setattr(blx, "run_agent", fake_agent)
|
||
res = await blx._facts_block(ctx, lambda *a, **k: None, {"arbeit": tmp_path},
|
||
{"Alpha": ["Sub Eins"]}, {"type": "uni"}, d, "", ns="x-")
|
||
assert res is not None
|
||
finder = [s for s in seen if "-facts-c0" in s[0] or "-facts-erg-" in s[0]]
|
||
assert finder and all(caps == "none" for _, caps, _ in finder)
|
||
assert all("── Skript.txt" in prompt for _, _, prompt in finder) # Auszüge inline
|
||
assert all("ls/find" not in prompt for _, _, prompt in finder)
|
||
|
||
|
||
def test_sub_key_resolves_short_titles():
|
||
"""Artefakt-Agenten echoen den Kurztitel; der Sub-Key heißt 'kurztitel: beschreibung'.
|
||
Eindeutiger Präfix wird aufgelöst, Mehrdeutiges und Fehlendes bleibt unverändert."""
|
||
import board_artefacts as ba
|
||
existing = {"autolink mit url: erzeugt link", "bilder: bindet bilder ein",
|
||
"doppel: eins", "doppel: zwei", "exakt"}
|
||
assert ba._sub_key(existing, "exakt") == "exakt"
|
||
assert ba._sub_key(existing, "autolink mit url") == "autolink mit url: erzeugt link"
|
||
assert ba._sub_key(existing, "doppel") == "doppel" # mehrdeutig → unverändert
|
||
assert ba._sub_key(existing, "fehlt") == "fehlt" # kein Treffer → unverändert
|
||
# Fuzzy: Paraphrase/Kürzung ohne Doppelpunkt-Präfix löst eindeutig auf
|
||
lang = {"der backslash selbst muss mit escaped werden, um literal zu erscheinen"}
|
||
assert ba._sub_key(lang, "der backslash selbst muss mit escaped werden") == next(iter(lang))
|
||
assert ba._sub_key(lang | {"der backslash am zeilenende"}, "der backslash") == "der backslash" # mehrdeutig
|
||
|
||
|
||
def test_subs_hash_invalidiert_bei_neuem_zuschnitt():
|
||
"""Gleicher Sub-Satz → gleicher Hash (Resume greift); geänderter → neuer Hash.
|
||
raw-Form (Strings) und sidecar-Form (dicts) hashen identisch."""
|
||
a = {"Block": ["s1", "s2"]}
|
||
assert blx._subs_hash(a) == blx._subs_hash({"Block": ["s1", "s2"]})
|
||
assert blx._subs_hash(a) != blx._subs_hash({"Block": ["s1", "s3"]})
|
||
assert blx._subs_hash(a) == blx._subs_hash({"Block": [{"title": "s1"}, {"title": "s2"}]})
|