dongarandClaude Opus 5.5
e351d0a74e
A pure move; behaviour unchanged. - models (L9, L7): llm, catalog, probe, render, render_anthropic, context - log (L7, L6): canonical, views - state (L8): store (was state), reconcile, worker, cancel, dump - knowledge (L10, L6.4): vault, memory - conversation (L5): transport, matrix - execution (L13): ssh (was execution), testing - authority (P3, P4, L4): actions - gitea, oplog, envfile stay at the top until each gains a second module. Every import, monkeypatch target, script and doc reference rewritten. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01J5Utyfzw7gF7UhjRDEtjTk
59 lines
1.8 KiB
Python
59 lines
1.8 KiB
Python
"""A real consult over the normalized route (L9).
|
|
|
|
The point is the prompt-level contract: a live model must produce something
|
|
that parses against our local schema, since no API-level enforcement exists.
|
|
"""
|
|
|
|
import io
|
|
from pathlib import Path
|
|
|
|
import pytest
|
|
|
|
from nisse_core.models.llm import LiteLLMClient, RoleMap
|
|
from nisse_core.oplog import OpLog
|
|
from nisse_dev.evals import collect
|
|
from nisse_dev.expert import BudgetExhausted, ExpertDesk
|
|
from nisse_dev.state import StateStore
|
|
|
|
pytestmark = pytest.mark.integration
|
|
|
|
|
|
@pytest.fixture
|
|
def desk(tmp_path: Path, litellm: LiteLLMClient, live_model: str):
|
|
def make(budget: int = 2) -> tuple[ExpertDesk, StateStore]:
|
|
model = live_model
|
|
roles = RoleMap(driver=model, helper=model, expert=model)
|
|
store = StateStore(
|
|
tmp_path / "state.db", oplog=OpLog("nisse-dev", io.StringIO())
|
|
)
|
|
return ExpertDesk(litellm, roles, budget=budget), store
|
|
|
|
return make
|
|
|
|
|
|
def test_live_consult_parses_against_the_local_schema(desk) -> None:
|
|
expert, store = desk()
|
|
|
|
answer = expert.ask(
|
|
store.transcript("t", "expert"),
|
|
"For a single-process agent storing a few thousand rows of task state, "
|
|
"is SQLite sufficient, or is PostgreSQL warranted? Answer briefly.",
|
|
)
|
|
|
|
assert answer.confidence in {"high", "medium", "low"}
|
|
assert len(answer.answer) > 20
|
|
# The consult is captured for L9 without any extra bookkeeping.
|
|
[record] = collect(store)
|
|
assert record.role == "expert"
|
|
assert record.exit_status == "answered"
|
|
|
|
|
|
def test_live_budget_stops_a_circling_caller(desk) -> None:
|
|
expert, store = desk(budget=1)
|
|
expert.ask(
|
|
store.transcript("t", "expert"),
|
|
"Name one advantage of WAL mode in SQLite. Be brief.",
|
|
)
|
|
with pytest.raises(BudgetExhausted):
|
|
expert.ask(store.transcript("t", "expert"), "And another?")
|