Files
chat/tests/test_turn_parse.py
T
Joseph Doherty 49be3cf4b9 fix: parse_turn falls back gracefully + classify logs flapping classifiers
The turn endpoint was 500ing in multi-bot scenes whenever the
classifier provider hiccuped on parse_turn — particularly visible
after a guest was added and bots started exchanging turns. The
traceback was 'classify failed for schema ParsedTurn with no default'
because parse_turn was the only classify caller without a default.

Two changes:

- chat/services/turn_parse.py: parse_turn now passes a default that
  wraps the whole prose as one 'dialogue' segment. The narrative
  still fires on the prose; we lose finer-grained segment kinds
  (action vs dialogue vs ooc) on this turn, but the request returns
  cleanly. Updated the existing test that pinned the old
  RuntimeError contract.

- chat/llm/classify.py: when retries are exhausted, log a WARNING
  with the schema name, last error type, and a snippet of the last
  raw text the model returned. Surfaces flapping classifiers in the
  uvicorn log for diagnosis without taking down the request.

Suite: 471 passed in 11.7s.
2026-04-27 15:07:39 -04:00

98 lines
3.2 KiB
Python

import json
import pytest
from chat.llm.mock import MockLLMClient
from chat.services.turn_parse import (
ParsedTurn,
TurnSegment,
parse_turn,
)
@pytest.mark.asyncio
async def test_parse_turn_three_segment_happy_path():
canned = json.dumps(
{
"segments": [
{"kind": "action", "text": "walks over"},
{"kind": "dialogue", "text": "Hey."},
{"kind": "ooc", "text": "player note"},
]
}
)
mock = MockLLMClient(canned=[canned])
result = await parse_turn(
mock,
model="m",
prose='*walks over* "Hey." ((player note))',
)
assert isinstance(result, ParsedTurn)
assert len(result.segments) == 3
kinds = [s.kind for s in result.segments]
assert kinds == ["action", "dialogue", "ooc"]
texts = [s.text for s in result.segments]
assert texts == ["walks over", "Hey.", "player note"]
assert all(isinstance(s, TurnSegment) for s in result.segments)
@pytest.mark.asyncio
async def test_parse_turn_pure_dialogue_single_segment():
canned = json.dumps(
{
"segments": [
{"kind": "dialogue", "text": "Hello there"},
]
}
)
mock = MockLLMClient(canned=[canned])
result = await parse_turn(
mock,
model="m",
prose='"Hello there"',
)
assert isinstance(result, ParsedTurn)
assert len(result.segments) == 1
assert result.segments[0].kind == "dialogue"
assert result.segments[0].text == "Hello there"
@pytest.mark.asyncio
async def test_parse_turn_empty_prose_short_circuits_without_classifier_call():
# No canned responses provided — if classify() is invoked it will raise
# IndexError on the empty list. The short-circuit must prevent that.
mock = MockLLMClient(canned=[])
result = await parse_turn(mock, model="m", prose="")
assert isinstance(result, ParsedTurn)
assert result.segments == []
# Whitespace-only prose must also short-circuit.
result_ws = await parse_turn(mock, model="m", prose=" \n\t ")
assert isinstance(result_ws, ParsedTurn)
assert result_ws.segments == []
@pytest.mark.asyncio
async def test_parse_turn_falls_back_to_whole_prose_when_classifier_fails():
"""A flapping classifier (3 invalid responses) no longer 500s the
request. ``parse_turn`` returns the original prose as a single
``dialogue`` segment so the turn flow can keep moving — the
narrative will still fire on the prose, just without finer-grained
segment classification.
The old contract was ``RuntimeError`` (no default), but in
production that took down the whole turn endpoint with a 500 the
moment any classifier provider hiccuped — particularly painful in
multi-bot scenes where every user turn pays the parse_turn cost.
"""
mock = MockLLMClient(canned=["nope", "still nope", "nope3"])
result = await parse_turn(
mock,
model="m",
prose='*shrugs* "whatever"',
)
assert len(result.segments) == 1
assert result.segments[0].kind == "dialogue"
assert result.segments[0].text == '*shrugs* "whatever"'
assert result.intent == "narrative"