Big change

.gitignore

@@ -81,3 +81,4 @@ dmypy.json
 
 /.kilocode/
 /data/whisper-models/
+/.qwen/

app/coding/domain/entities.py

@@ -357,7 +357,7 @@ def start_timer_for_task(
     def with_submit_test_summary(
         self,
         task_row_id: int,
-        summary: dict[str, Any],
+        summary: dict[str, Any] | None,
         *,
         source_code: str,
     ) -> CodingSection:

app/coding/domain/value_objects.py

@@ -32,7 +32,7 @@ class PlannedCodingTask:
 
 
 @dataclass(frozen=True, slots=True)
-class TestCaseRunResult:
+class CaseRunResult:
     """Outcome of executing one public test case via Judge0.
 
     Attributes:
@@ -77,5 +77,5 @@ class CodingRunResult:
     compile_output: str | None
     tests_passed: int
     tests_total: int
-    test_results: tuple[TestCaseRunResult, ...]
+    test_results: tuple[CaseRunResult, ...]
     duration_ms: int | None

app/coding/services/evaluation_persistence.py

@@ -121,11 +121,16 @@ def persist(
             raise CodingSectionNotFoundError(interview_id)
         section.ensure_active()
 
+        existing_task = section.find_task(task_id, round_num)
         updated = section.with_evaluation(
             task_id,
             round_num,
             evaluation.score,
             evaluation.feedback,
+        ).with_submit_test_summary(
+            existing_task.id,
+            submit_test_summary,
+            source_code=existing_task.submitted_code or submitted_source_code,
         )
         follow_up_round: int | None = None
         if follow_up_needed:

app/coding/services/run_execution.py

@@ -13,7 +13,7 @@
     CodingRunLimitExceededError,
     CodingSectionNotFoundError,
 )
-from app.coding.domain.value_objects import CodingRunResult, TestCaseRunResult
+from app.coding.domain.value_objects import CaseRunResult, CodingRunResult
 from app.coding.services.runner import CodingRunnerService
 from app.interview.domain.exceptions import InterviewNotFoundError
 from app.interview.repositories.uow import InterviewUnitOfWork
@@ -50,7 +50,7 @@ def _ensure_interview_active(interview_id: str) -> None:
         aggregate.ensure_active()
 
 
-def _serialize_test_result(result: TestCaseRunResult) -> dict[str, Any]:
+def _serialize_test_result(result: CaseRunResult) -> dict[str, Any]:
     """Convert a domain test result into an API/persistence payload.
 
     Args:

app/coding/services/runner.py

@@ -7,9 +7,9 @@
 from typing import Any
 
 from app.coding.domain.value_objects import (
+    CaseRunResult,
     CodingRunResult,
     RunOutcomeStatus,
-    TestCaseRunResult,
 )
 from app.coding.services.harness import build_python_script
 from app.coding.services.judge0_client import (
@@ -109,7 +109,7 @@ async def run_public_tests(
                 client=judge0,
             )
 
-        results: list[TestCaseRunResult] = []
+        results: list[CaseRunResult] = []
         total_duration_ms = 0
         last_stdout: str | None = None
         last_stderr: str | None = None
@@ -218,7 +218,7 @@ def _case_result_from_submission(
         name: str,
         expected_stdout: str,
         submission: Judge0SubmissionResult,
-    ) -> TestCaseRunResult:
+    ) -> CaseRunResult:
         """Map one Judge0 submission to a public test case result.
 
         Args:
@@ -234,7 +234,7 @@ def _case_result_from_submission(
             submission.status_id == JUDGE0_STATUS_ACCEPTED
             and actual_stdout == expected_stdout
         )
-        return TestCaseRunResult(
+        return CaseRunResult(
             name=name,
             passed=passed,
             expected_stdout=expected_stdout,
@@ -272,7 +272,7 @@ def _status_from_submission(
         return "tests_failed"
 
     @staticmethod
-    def _aggregate_status(results: list[TestCaseRunResult]) -> RunOutcomeStatus:
+    def _aggregate_status(results: list[CaseRunResult]) -> RunOutcomeStatus:
         """Derive the aggregate run status from per-test results.
 
         Args:

app/interview/services/section_prefetch.py

@@ -38,7 +38,7 @@ async def prefetch_section_feedback(
 
     try:
         provider = ConfigService.create_provider_from_config()
-    except ValueError:
+    except Exception:
         logger.warning(
             "Skipping %s section prefetch for %s: provider not configured",
             section_name,

tests/ai/test_faster_whisper_transcriber.py

@@ -0,0 +1,137 @@
+# Copyright 2026 GrillKit Contributors
+# SPDX-License-Identifier: Apache-2.0
+"""Tests for FasterWhisperTranscriber."""
+
+from unittest.mock import MagicMock, patch
+
+import numpy as np
+import pytest
+
+from app.ai.faster_whisper_transcriber import FasterWhisperTranscriber
+
+
+class FakeSegment:
+    """Fake segment matching faster-whisper segment interface."""
+
+    def __init__(self, text):
+        self.text = text
+
+
+class TestFasterWhisperTranscriber:
+    """Tests for the faster-whisper adapter."""
+
+    @pytest.mark.asyncio
+    async def test_transcribe_calls_model_transcribe(self):
+        """Adapter delegates to WhisperModel.transcribe with correct arguments."""
+        segment = FakeSegment(" hello world")
+        model = MagicMock()
+        model.transcribe.return_value = ([segment], None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(16000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, "en")
+
+        assert text == "hello world"
+        model.transcribe.assert_called_once()
+        call_args, call_kwargs = model.transcribe.call_args
+        assert call_args[0] is audio
+        assert call_kwargs["language"] == "en"
+        assert call_kwargs["task"] == "transcribe"
+
+    @pytest.mark.asyncio
+    async def test_uses_vad_filter_true(self):
+        """Transcription enables VAD filtering."""
+        segment = FakeSegment(" test")
+        model = MagicMock()
+        model.transcribe.return_value = ([segment], None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        await transcriber.transcribe(audio, "ru")
+
+        call_kwargs = model.transcribe.call_args.kwargs
+        assert call_kwargs["vad_filter"] is True
+
+    @pytest.mark.asyncio
+    async def test_normalizes_locale(self):
+        """Locale is normalized before being passed to the model."""
+        segment = FakeSegment(" result")
+        model = MagicMock()
+        model.transcribe.return_value = ([segment], None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, " RU ")
+
+        call_kwargs = model.transcribe.call_args.kwargs
+        assert call_kwargs["language"] == "ru"
+        assert text == "result"
+
+    @pytest.mark.asyncio
+    async def test_joins_multiple_segments(self):
+        """Multiple segments are concatenated and stripped."""
+        segments = [FakeSegment(" Hello"), FakeSegment(" world")]
+        model = MagicMock()
+        model.transcribe.return_value = (segments, None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, "en")
+
+        assert text == "Hello world"
+
+    @pytest.mark.asyncio
+    async def test_returns_empty_string_when_no_segments(self):
+        """Empty segment list produces empty transcript."""
+        model = MagicMock()
+        model.transcribe.return_value = ([], None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, "en")
+
+        assert text == ""
+
+    @pytest.mark.asyncio
+    async def test_runs_in_thread(self):
+        """Transcription runs the blocking model in a thread."""
+        segment = FakeSegment(" async result")
+        model = MagicMock()
+        model.transcribe.return_value = ([segment], None)
+
+        with patch(
+            "asyncio.to_thread", side_effect=lambda f, *a, **k: f(*a, **k)
+        ) as mock_to_thread:  # noqa: E501
+            transcriber = FasterWhisperTranscriber(model)
+            audio = np.zeros(8000, dtype=np.float32)
+            await transcriber.transcribe(audio, "en")
+
+        mock_to_thread.assert_called_once()
+
+    @pytest.mark.asyncio
+    async def test_strips_whitespace(self):
+        """Trailing and leading whitespace is stripped from result."""
+        segments = [FakeSegment("  padded  ")]
+        model = MagicMock()
+        model.transcribe.return_value = (segments, None)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, "en")
+
+        assert text == "padded"
+
+    @pytest.mark.asyncio
+    async def test_ignores_info_return_value(self):
+        """Second return value from transcribe (info) is ignored."""
+        segment = FakeSegment("text")
+        info = MagicMock()
+        info.language = "en"
+        model = MagicMock()
+        model.transcribe.return_value = ([segment], info)
+
+        transcriber = FasterWhisperTranscriber(model)
+        audio = np.zeros(8000, dtype=np.float32)
+        text = await transcriber.transcribe(audio, "en")
+
+        assert text == "text"