Spaces:
Sleeping
Sleeping
File size: 8,819 Bytes
a8c7066 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 | import unittest
from src.application.dto.ask_question_dto import AskQuestionCommand
from src.application.ports.outbound.observability_port import ScoreInput
from src.application.use_cases.ask_question import AskQuestionService
from src.domain.errors.exceptions import EmptyQuestionError
from src.domain.model.evaluation_models import AnswerEvaluation
from src.domain.model.rag_models import Document, GenerationResult
class FakeRetrievalPort:
def __init__(self, chunks):
self.chunks = chunks
self.calls = []
def retrieve(self, question, top_k, lambda_mult, similarity_threshold):
self.calls.append((question, top_k, lambda_mult, similarity_threshold))
return self.chunks
class FakeLlmPort:
def __init__(self):
self.calls = []
def generate_answer(self, question, chunks):
self.calls.append((question, chunks))
return GenerationResult(
text="answer-from-llm",
prompt="prompt-from-llm",
input_tokens=10,
output_tokens=12,
thoughts_tokens=3,
finish_reason="STOP",
)
def get_model_name(self):
return "gemini-2.5-flash"
def get_generation_config(self):
return {"thinking_budget": 200, "max_output_tokens": 512}
class FakeObservationHandle:
def __init__(self, trace_id, span_id, name):
self.trace_id = trace_id
self.id = span_id
self.name = name
self.updates = []
self.scores = []
self.closed = False
def update(self, **kwargs):
self.updates.append(kwargs)
def score(self, name, value=None, data_type=None):
if isinstance(name, dict):
# Batch mode: iterate over scores dictionary
for score_name, score_input in name.items():
self.scores.append(score_input)
else:
# Single mode
self.scores.append(ScoreInput(name=name, value=value, data_type=data_type))
def close(self):
self.closed = True
class FakeObservabilityPort:
def __init__(self):
self.traces = []
self.scores = []
self.trace_handle = None
self.all_handles = []
def start_trace(self, name, payload, *, trace_name=None, user_id=None, session_id=None):
self.traces.append((name, payload, trace_name, user_id, session_id))
self.trace_handle = FakeObservationHandle("trace-123", "root-123", name)
self.all_handles.append(self.trace_handle)
return self.trace_handle
def start_observation(self, name, as_type=None, model=None):
handle = FakeObservationHandle("trace-123", f"{name}-123", name)
self.all_handles.append(handle)
return handle
def finish_trace(self):
pass
def create_score(self, trace_id, name, value, comment=None, observation_id=None):
self.scores.append((trace_id, name, value, comment, observation_id))
class FakeEvaluationPort:
def __init__(self):
self.items = []
def enqueue(self, evaluation, trace_id, observation_id=None):
self.items.append((evaluation, trace_id, observation_id))
class AskQuestionServiceTests(unittest.TestCase):
def test_raises_on_empty_question(self):
service = AskQuestionService(
retrieval_port=FakeRetrievalPort([]),
llm_port=FakeLlmPort(),
observability_port=FakeObservabilityPort(),
default_answer="fallback",
evaluation_port=FakeEvaluationPort(),
)
with self.assertRaises(EmptyQuestionError):
service.execute(AskQuestionCommand(question=" "))
def test_returns_default_answer_when_no_chunks(self):
observability = FakeObservabilityPort()
evaluation = FakeEvaluationPort()
service = AskQuestionService(
retrieval_port=FakeRetrievalPort([]),
llm_port=FakeLlmPort(),
observability_port=observability,
default_answer="fallback",
evaluation_port=evaluation,
)
result = service.execute(AskQuestionCommand(question="hello"))
self.assertEqual(result.answer, "fallback")
self.assertEqual(result.trace_id, "trace-123")
self.assertEqual(len(result.chunks), 0)
self.assertEqual(observability.traces[0][0], "rag-pipeline")
self.assertEqual(observability.traces[0][1]["query"], "hello")
self.assertTrue(any(update.get("status_message") == "DEFAULT_ANSWER" for handle in observability.all_handles for update in handle.updates))
self.assertIsInstance(evaluation.items[0][0], AnswerEvaluation)
self.assertEqual(evaluation.items[0][0].contexts, [])
self.assertEqual(evaluation.items[0][1], "trace-123")
def test_uses_llm_when_chunks_exist(self):
chunk = Document(id="1", content="context", score=0.91)
retrieval = FakeRetrievalPort([chunk])
llm = FakeLlmPort()
observability = FakeObservabilityPort()
evaluation = FakeEvaluationPort()
service = AskQuestionService(
retrieval_port=retrieval,
llm_port=llm,
observability_port=observability,
default_answer="fallback",
evaluation_port=evaluation,
)
result = service.execute(AskQuestionCommand(question="hello", use_mmr=True, lambda_mult=0.5))
self.assertEqual(result.answer, "answer-from-llm")
self.assertEqual(len(result.chunks), 1)
self.assertEqual(retrieval.calls[0][2], 0.5)
self.assertEqual(observability.traces[0][1]["query"], "hello")
self.assertEqual(observability.traces[0][1]["lambda_mult"], 0.5)
self.assertEqual(llm.calls[0][0], "hello")
self.assertEqual(observability.all_handles[1].updates[0]["status_message"], "SUCCESS")
generation_metadata = observability.all_handles[2].updates[0]["metadata"]
self.assertEqual(generation_metadata["thinking_budget"], 200)
self.assertEqual(generation_metadata["max_output_tokens"], 512)
self.assertEqual(evaluation.items[0][0].question, "hello")
self.assertEqual(evaluation.items[0][0].contexts, ["context"])
def test_score_payloads_are_emitted_as_score_input_objects(self):
observability = FakeObservabilityPort()
evaluation = FakeEvaluationPort()
service = AskQuestionService(
retrieval_port=FakeRetrievalPort([Document(id="1", content="context", score=0.91)]),
llm_port=FakeLlmPort(),
observability_port=observability,
default_answer="fallback",
evaluation_port=evaluation,
)
service.execute(AskQuestionCommand(question="hello"))
self.assertTrue(observability.all_handles[1].scores)
self.assertIsInstance(observability.all_handles[1].scores[0], ScoreInput)
self.assertEqual(observability.all_handles[1].scores[0].name, "api_latency")
def test_trace_payload_uses_none_lambda_mult_when_mmr_disabled(self):
observability = FakeObservabilityPort()
evaluation = FakeEvaluationPort()
service = AskQuestionService(
retrieval_port=FakeRetrievalPort([]),
llm_port=FakeLlmPort(),
observability_port=observability,
default_answer="fallback",
evaluation_port=evaluation,
)
service.execute(AskQuestionCommand(question="hello", use_mmr=False, lambda_mult=0.5))
self.assertIsNone(observability.traces[0][1]["lambda_mult"])
def test_passes_trace_name_and_user_id_and_resets_session_id_per_call(self):
observability = FakeObservabilityPort()
evaluation = FakeEvaluationPort()
service = AskQuestionService(
retrieval_port=FakeRetrievalPort([]),
llm_port=FakeLlmPort(),
observability_port=observability,
default_answer="fallback",
evaluation_port=evaluation,
)
service.execute(AskQuestionCommand(question="hello", trace_name="chatbot-search", user_id="user-42"))
service.execute(AskQuestionCommand(question="hello again", trace_name="chatbot-search", user_id="user-42"))
first_trace_name, first_user_id, first_session_id = observability.traces[0][2:5]
second_trace_name, second_user_id, second_session_id = observability.traces[1][2:5]
self.assertEqual(first_trace_name, "chatbot-search")
self.assertEqual(first_user_id, "user-42")
self.assertEqual(second_trace_name, "chatbot-search")
self.assertEqual(second_user_id, "user-42")
self.assertIsNotNone(first_session_id)
self.assertIsNotNone(second_session_id)
self.assertNotEqual(first_session_id, second_session_id)
if __name__ == "__main__":
unittest.main()
|