68f4f01cd7
更新了 .env.example，新增聊天模型配置，以提升对话处理能力。 增强了 README.md，反映了包括需求澄清、代码复用和自动重试在内的新功能。 重构了 agent.py，以支持多模型交互，并为无法在本地执行的任务新增了引导处理逻辑。 改进了 SandboxRunner，增加了任务执行成功校验，并加入了工作区清理功能。 扩展了 HistoryManager，支持任务摘要生成以及记录的批量删除。 优化了 chat_view.py 和 history_view.py 中的 UI 组件，提升用户体验，包括 Markdown 渲染和任务管理选项。
97 lines
3.6 KiB
Python
97 lines
3.6 KiB
Python
"""
|
||
意图分类器单元测试
|
||
"""
|
||
|
||
import unittest
|
||
import sys
|
||
from pathlib import Path
|
||
|
||
# 添加项目根目录到路径
|
||
sys.path.insert(0, str(Path(__file__).parent.parent))
|
||
|
||
from intent.labels import CHAT, EXECUTION, GUIDANCE, VALID_LABELS, EXECUTION_CONFIDENCE_THRESHOLD
|
||
|
||
|
||
class TestIntentLabels(unittest.TestCase):
|
||
"""意图标签测试"""
|
||
|
||
def test_labels_defined(self):
|
||
"""测试标签已定义"""
|
||
self.assertEqual(CHAT, "chat")
|
||
self.assertEqual(EXECUTION, "execution")
|
||
self.assertEqual(GUIDANCE, "guidance")
|
||
|
||
def test_valid_labels(self):
|
||
"""测试有效标签集合"""
|
||
self.assertIn(CHAT, VALID_LABELS)
|
||
self.assertIn(EXECUTION, VALID_LABELS)
|
||
self.assertIn(GUIDANCE, VALID_LABELS)
|
||
self.assertEqual(len(VALID_LABELS), 3)
|
||
|
||
def test_confidence_threshold(self):
|
||
"""测试置信度阈值"""
|
||
self.assertGreater(EXECUTION_CONFIDENCE_THRESHOLD, 0)
|
||
self.assertLessEqual(EXECUTION_CONFIDENCE_THRESHOLD, 1)
|
||
|
||
|
||
class TestIntentClassifierParsing(unittest.TestCase):
|
||
"""意图分类器解析测试（不需要 API）"""
|
||
|
||
def setUp(self):
|
||
from intent.classifier import IntentClassifier
|
||
self.classifier = IntentClassifier()
|
||
|
||
def test_parse_valid_chat_response(self):
|
||
"""测试解析有效的 chat 响应"""
|
||
response = '{"label": "chat", "confidence": 0.95, "reason": "这是一个问答"}'
|
||
result = self.classifier._parse_response(response)
|
||
self.assertEqual(result.label, CHAT)
|
||
self.assertEqual(result.confidence, 0.95)
|
||
self.assertEqual(result.reason, "这是一个问答")
|
||
|
||
def test_parse_valid_execution_response(self):
|
||
"""测试解析有效的 execution 响应"""
|
||
response = '{"label": "execution", "confidence": 0.9, "reason": "需要复制文件"}'
|
||
result = self.classifier._parse_response(response)
|
||
self.assertEqual(result.label, EXECUTION)
|
||
self.assertEqual(result.confidence, 0.9)
|
||
|
||
def test_parse_low_confidence_execution(self):
|
||
"""测试低置信度的 execution 降级为 chat"""
|
||
response = '{"label": "execution", "confidence": 0.5, "reason": "不太确定"}'
|
||
result = self.classifier._parse_response(response)
|
||
# 低于阈值应该降级为 chat
|
||
self.assertEqual(result.label, CHAT)
|
||
|
||
def test_parse_invalid_label(self):
|
||
"""测试无效标签降级为 chat"""
|
||
response = '{"label": "unknown", "confidence": 0.9, "reason": "测试"}'
|
||
result = self.classifier._parse_response(response)
|
||
self.assertEqual(result.label, CHAT)
|
||
|
||
def test_parse_invalid_json(self):
|
||
"""测试无效 JSON 降级为 chat"""
|
||
response = 'not a json'
|
||
result = self.classifier._parse_response(response)
|
||
self.assertEqual(result.label, CHAT)
|
||
self.assertEqual(result.confidence, 0.0)
|
||
|
||
def test_extract_json_with_prefix(self):
|
||
"""测试从带前缀的文本中提取 JSON"""
|
||
text = 'Here is the result: {"label": "chat", "confidence": 0.8, "reason": "test"}'
|
||
json_str = self.classifier._extract_json(text)
|
||
self.assertTrue(json_str.startswith('{'))
|
||
self.assertTrue(json_str.endswith('}'))
|
||
|
||
def test_extract_json_with_suffix(self):
|
||
"""测试从带后缀的文本中提取 JSON"""
|
||
text = '{"label": "chat", "confidence": 0.8, "reason": "test"} That is my answer.'
|
||
json_str = self.classifier._extract_json(text)
|
||
self.assertTrue(json_str.startswith('{'))
|
||
self.assertTrue(json_str.endswith('}'))
|
||
|
||
|
||
if __name__ == '__main__':
|
||
unittest.main()
|
||
|