Buckets:

download
raw
3.39 kB
# tests/test_analyzer.py
"""Tests for veronica.analyzer -- RuleAnalyzer with 3 rules."""
from __future__ import annotations
import time
from veronica.analyzer import RuleAnalyzer
from veronica.types import HistoryView, StepIntent, StepOutcome
def _intent(kind: str = "llm", model: str = "gpt-4") -> StepIntent:
return StepIntent(
step_id="s1",
request_id="r1",
chain_id="c1",
kind=kind,
model=model,
tool_name=None,
timeout_ms=30_000,
metadata={},
)
def _outcome(status: str = "ok", cost: float = 0.005) -> StepOutcome:
return StepOutcome(
step_id="s1",
request_id="r1",
chain_id="c1",
kind="llm",
status=status,
cost_usd=cost,
tokens_in=100,
tokens_out=50,
elapsed_ms=100.0,
model="gpt-4",
events=(),
timestamp_ms=int(time.time() * 1000),
)
def _history(
failure_streak: int = 0,
depth: int = 1,
rolling_cost: float = 0.01,
) -> HistoryView:
return HistoryView(
chain_id="c1",
last_n=(),
rolling_cost_usd=rolling_cost,
failure_streak=failure_streak,
depth=depth,
loop_score=0.0,
)
class TestRuleAnalyzer:
def test_nominal_clean_run(self) -> None:
"""Rule 2: clean run -> loosen."""
analyzer = RuleAnalyzer()
result = analyzer.analyze(_intent(), _outcome("ok"), _history())
assert result.risk_level == "nominal"
assert result.recommendation == "loosen"
def test_halt_triggers_tighten(self) -> None:
"""Rule 1: halt -> tighten."""
analyzer = RuleAnalyzer()
result = analyzer.analyze(_intent(), _outcome("halted"), _history())
assert result.recommendation == "tighten"
def test_depth_guard(self) -> None:
"""Rule 3: depth >= 8 -> halt recommendation."""
analyzer = RuleAnalyzer()
result = analyzer.analyze(
_intent(),
_outcome("ok"),
_history(depth=8),
)
assert result.recommendation == "halt"
assert result.risk_level == "critical"
signals = [s.kind for s in result.signals]
assert "depth_anomaly" in signals
def test_failure_streak_elevated(self) -> None:
"""Consecutive failures -> elevated risk."""
analyzer = RuleAnalyzer()
result = analyzer.analyze(
_intent(),
_outcome("error"),
_history(failure_streak=3),
)
assert result.risk_level in ("elevated", "critical")
assert result.recommendation == "tighten"
def test_intent_model_mismatch(self) -> None:
"""Intent kind != outcome kind -> intent_deviation signal."""
analyzer = RuleAnalyzer()
outcome = StepOutcome(
step_id="s1",
request_id="r1",
chain_id="c1",
kind="tool",
status="ok",
cost_usd=0.0,
tokens_in=0,
tokens_out=0,
elapsed_ms=10.0,
model=None,
events=(),
timestamp_ms=int(time.time() * 1000),
)
result = analyzer.analyze(
_intent(kind="llm"),
outcome,
_history(),
)
signals = [s.kind for s in result.signals]
assert "intent_deviation" in signals

Xet Storage Details

Size:
3.39 kB
·
Xet hash:
ed0081b94c097bf6099a1600f34266f21c558d4be62c8772d72fb0e1561c0a7d

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.