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
65 lines
2.3 KiB
Python
65 lines
2.3 KiB
Python
"""Live smoke test against the real LiteLLM proxy."""
|
|
|
|
import pytest
|
|
|
|
from nisse_core.log.canonical import AssistantText, StoredEvent, UserMessage
|
|
from nisse_core.models.catalog import ModelCatalog
|
|
from nisse_core.models.context import BudgetedClient
|
|
from nisse_core.models.llm import LiteLLMClient
|
|
from nisse_dev.config import AgentConfig
|
|
|
|
pytestmark = pytest.mark.integration
|
|
|
|
|
|
def test_the_driver_model_completes_via_litellm(
|
|
litellm: LiteLLMClient, live_config: AgentConfig
|
|
) -> None:
|
|
events = [
|
|
StoredEvent(
|
|
id=1,
|
|
ts="t",
|
|
event=UserMessage(
|
|
text="Reply with exactly one word: pong", source="smoke-test"
|
|
),
|
|
)
|
|
]
|
|
result = litellm.complete(events, tools=None, model=live_config.roles.driver)
|
|
texts = [e.text for e in result if isinstance(e, AssistantText)]
|
|
assert texts, f"no assistant text in {result!r}"
|
|
# A plumbing check, so any answer will do: what the model chooses to say
|
|
# is not under test, and asserting the word made CI flaky (it once
|
|
# replied "ping").
|
|
assert texts[0].strip(), f"empty assistant text in {result!r}"
|
|
|
|
|
|
def test_declared_limits_reach_a_budgeted_request(
|
|
litellm: LiteLLMClient, live_config: AgentConfig
|
|
) -> None:
|
|
"""The whole chain: LiteLLM's model_info, the catalog, the budget, and a
|
|
real request through the proxy with the ceiling the budget chose."""
|
|
catalog = ModelCatalog.discover(live_config.llm_base_url, live_config.llm_token)
|
|
roles = live_config.roles.pinned()
|
|
catalog.require_limits(roles) # every test role has a declared window
|
|
|
|
model = live_config.roles.driver
|
|
facts = catalog.facts(model)
|
|
client = BudgetedClient(
|
|
litellm,
|
|
lambda m: (
|
|
catalog.facts(m).max_input_tokens,
|
|
catalog.facts(m).max_output_tokens,
|
|
),
|
|
max_output_tokens=live_config.max_output_tokens,
|
|
)
|
|
assert client.budget_for(model).window == facts.max_input_tokens
|
|
|
|
events = [
|
|
StoredEvent(
|
|
id=1, ts="t", event=UserMessage(text="Reply with exactly: pong", source="t")
|
|
)
|
|
]
|
|
produced = client.complete(events, None, model, system="Be terse.")
|
|
texts = [e.text for e in produced if isinstance(e, AssistantText)]
|
|
assert texts and "pong" in texts[0].lower()
|
|
assert client.last_usage.get("input", 0) > 0
|