use multiple apis
This commit is contained in:
622
app/agents.py
622
app/agents.py
@@ -15,6 +15,94 @@ logging.basicConfig(level=logging.INFO)
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
def _parse_json(content: str, error_prefix: str) -> dict:
|
||||
"""解析 LLM 返回的 JSON，解析失败时尝试正则兜底。"""
|
||||
try:
|
||||
return json.loads(content)
|
||||
except json.JSONDecodeError:
|
||||
import re
|
||||
m = re.search(r'\{.*\}', content, re.DOTALL)
|
||||
if m:
|
||||
return json.loads(m.group())
|
||||
raise ValueError(f"{error_prefix}: {content}")
|
||||
|
||||
|
||||
class ClarifyAgent:
|
||||
"""需求澄清 Agent —— 判断需求是否完整，并追问用户补充信息"""
|
||||
|
||||
def __init__(self):
|
||||
self.settings = get_settings()
|
||||
self.client = OpenAI(api_key=self.settings.api_key, base_url=self.settings.base_url)
|
||||
|
||||
def start(self, raw_requirement: str) -> dict:
|
||||
"""
|
||||
用户第一次提交需求时调用。
|
||||
返回 {"ready": bool, "question": str, "clarified_requirement": str}
|
||||
"""
|
||||
prompt = f"""你是一个资深的产品经理助手，负责在正式分析需求之前确认需求的完整性。
|
||||
|
||||
用户提交的需求：
|
||||
{raw_requirement}
|
||||
|
||||
请判断该需求是否足够清晰，可以直接开始产品分析。
|
||||
|
||||
返回 JSON：
|
||||
{{
|
||||
"ready": true 或 false,
|
||||
"question": "如果 ready=false，给用户一个简洁的追问；如果 ready=true，此字段为空字符串",
|
||||
"clarified_requirement": "整合后的完整需求描述（即使 ready=false 也要输出当前已有的描述）"
|
||||
}}
|
||||
|
||||
判断标准：
|
||||
- 需求描述清楚要做什么
|
||||
- 有基本的场景或目标
|
||||
- 不需要追问过多细节，够 PM 开始分析即可
|
||||
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个产品经理助手，负责判断需求完整性，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=500,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
return _parse_json(response.choices[0].message.content, "ClarifyAgent.start 解析失败")
|
||||
|
||||
def continue_clarify(self, clarify_history: list[dict], user_reply: str) -> dict:
|
||||
"""
|
||||
用户补充信息后继续澄清。
|
||||
clarify_history 格式：[{"role": "assistant"|"user", "content": str}, ...]
|
||||
返回同 start() 的格式。
|
||||
"""
|
||||
messages = [
|
||||
{"role": "system", "content": "你是一个产品经理助手，负责判断需求完整性，输出必须是严格的 JSON 格式。"}
|
||||
] + clarify_history + [
|
||||
{"role": "user", "content": user_reply}
|
||||
]
|
||||
|
||||
suffix = """\n\n根据以上对话，判断需求现在是否足够清晰。返回 JSON：
|
||||
{{
|
||||
"ready": true 或 false,
|
||||
"question": "如果 ready=false，继续追问；如果 ready=true，此字段为空字符串",
|
||||
"clarified_requirement": "整合所有对话后的完整需求描述"
|
||||
}}
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
messages.append({"role": "user", "content": suffix})
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=messages,
|
||||
temperature=0.2,
|
||||
max_tokens=500,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
return _parse_json(response.choices[0].message.content, "ClarifyAgent.continue_clarify 解析失败")
|
||||
|
||||
|
||||
class PMAgent:
|
||||
"""产品经理Agent - 完善和扩展需求"""
|
||||
|
||||
@@ -22,20 +110,10 @@ class PMAgent:
|
||||
self.settings = get_settings()
|
||||
self.client = OpenAI(api_key=self.settings.api_key, base_url=self.settings.base_url)
|
||||
|
||||
def analyze_requirement(self, simple_requirement: str) -> RequirementAnalysis:
|
||||
"""
|
||||
分析和完善简单的需求描述
|
||||
def _build_prompt(self, requirement: str) -> str:
|
||||
return f"""你是一个资深的产品经理。请根据以下简单的需求描述，进行深入分析和完善。
|
||||
|
||||
Args:
|
||||
simple_requirement: 用户提供的简单需求描述
|
||||
|
||||
Returns:
|
||||
RequirementAnalysis: 包含完善后的需求信息
|
||||
"""
|
||||
|
||||
prompt = f"""你是一个资深的产品经理。请根据以下简单的需求描述，进行深入分析和完善。
|
||||
|
||||
需求描述：{simple_requirement}
|
||||
需求描述：{requirement}
|
||||
|
||||
请按以下格式返回JSON结果（必须是有效的JSON格式）：
|
||||
{{
|
||||
@@ -55,6 +133,104 @@ class PMAgent:
|
||||
|
||||
返回ONLY JSON内容，不要有其他文字。"""
|
||||
|
||||
def analyze_requirement(self, simple_requirement: str) -> RequirementAnalysis:
|
||||
"""
|
||||
分析和完善简单的需求描述
|
||||
|
||||
Args:
|
||||
simple_requirement: 用户提供的简单需求描述
|
||||
|
||||
Returns:
|
||||
RequirementAnalysis: 包含完善后的需求信息
|
||||
"""
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的产品经理，擅长需求分析和拆解，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(simple_requirement)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=2000,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
result = _parse_json(response.choices[0].message.content, "PMAgent 解析失败")
|
||||
return RequirementAnalysis(**result)
|
||||
|
||||
def stream_analyze(self, simple_requirement: str):
|
||||
"""
|
||||
流式版需求分析。yield 文本块(str)，最后 yield (None, RequirementAnalysis) 作为哨兵。
|
||||
"""
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的产品经理，擅长需求分析和拆解，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(simple_requirement)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=2000,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "PMAgent stream 解析失败")
|
||||
yield (None, RequirementAnalysis(**result))
|
||||
|
||||
def stream_refine(self, previous: RequirementAnalysis, feedback: str):
|
||||
"""
|
||||
流式版需求分析修改。yield 文本块(str)，最后 yield (None, RequirementAnalysis) 作为哨兵。
|
||||
"""
|
||||
prompt = f"""你是一个资深的产品经理。以下是你之前输出的需求分析结果，用户对此有修改意见，请根据意见调整输出。
|
||||
|
||||
之前的需求分析：
|
||||
{json.dumps(previous, ensure_ascii=False, indent=2)}
|
||||
|
||||
用户的修改意见：
|
||||
{feedback}
|
||||
|
||||
请在原有基础上修改，保持 JSON 格式不变，返回完整的修改后结果，返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的产品经理，擅长需求分析和拆解，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=2000,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "PMAgent stream_refine 解析失败")
|
||||
yield (None, RequirementAnalysis(**result))
|
||||
|
||||
def refine(self, previous: RequirementAnalysis, feedback: str) -> RequirementAnalysis:
|
||||
"""
|
||||
根据用户反馈修改已有的需求分析结果。
|
||||
|
||||
Returns:
|
||||
RequirementAnalysis: 修改后的需求分析结果
|
||||
"""
|
||||
prompt = f"""你是一个资深的产品经理。以下是你之前输出的需求分析结果，用户对此有修改意见，请根据意见调整输出。
|
||||
|
||||
之前的需求分析：
|
||||
{json.dumps(previous, ensure_ascii=False, indent=2)}
|
||||
|
||||
用户的修改意见：
|
||||
{feedback}
|
||||
|
||||
请在原有基础上修改，保持 JSON 格式不变，返回完整的修改后结果，返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
@@ -65,22 +241,8 @@ class PMAgent:
|
||||
max_tokens=2000,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
|
||||
# 提取响应内容
|
||||
content = response.choices[0].message.content
|
||||
|
||||
# 解析JSON
|
||||
try:
|
||||
result = json.loads(content)
|
||||
return RequirementAnalysis(**result)
|
||||
except json.JSONDecodeError:
|
||||
# 如果JSON解析失败，尝试提取JSON部分
|
||||
import re
|
||||
json_match = re.search(r'\{.*\}', content, re.DOTALL)
|
||||
if json_match:
|
||||
result = json.loads(json_match.group())
|
||||
return RequirementAnalysis(**result)
|
||||
raise ValueError(f"无法解析Agent响应: {content}")
|
||||
result = _parse_json(response.choices[0].message.content, "PMAgent.refine 解析失败")
|
||||
return RequirementAnalysis(**result)
|
||||
|
||||
|
||||
class QAAgent:
|
||||
@@ -90,18 +252,7 @@ class QAAgent:
|
||||
self.settings = get_settings()
|
||||
self.client = OpenAI(api_key=self.settings.api_key, base_url=self.settings.base_url)
|
||||
|
||||
def generate_test_cases(self, requirement_analysis: RequirementAnalysis) -> TestCaseResult:
|
||||
"""
|
||||
基于需求分析生成测试用例
|
||||
|
||||
Args:
|
||||
requirement_analysis: PM Agent的分析结果
|
||||
|
||||
Returns:
|
||||
TestCaseResult: 包含测试用例的结果
|
||||
"""
|
||||
|
||||
# 构建需求信息
|
||||
def _build_prompt(self, requirement_analysis: RequirementAnalysis) -> str:
|
||||
requirement_text = f"""
|
||||
功能需求：
|
||||
{chr(10).join(f"- {req}" for req in requirement_analysis["functional_requirements"])}
|
||||
@@ -118,8 +269,7 @@ class QAAgent:
|
||||
需求总结：
|
||||
{requirement_analysis["summary"]}
|
||||
"""
|
||||
|
||||
prompt = f"""你是一个资深的Java QA工程师。根据以下需求信息，生成全面的Java测试用例和测试策略。所有测试用例必须基于Java语言，步骤和预期结果要符合Java的类型系统、异常机制和JUnit测试框架。
|
||||
return f"""你是一个资深的Python QA工程师。根据以下需求信息，生成全面的Python测试用例和测试策略。所有测试用例必须基于Python语言，步骤和预期结果要符合Python的类型系统、异常机制和pytest测试框架。
|
||||
|
||||
{requirement_text}
|
||||
|
||||
@@ -145,61 +295,140 @@ class QAAgent:
|
||||
2. 为每个边缘情况生成1个测试用例
|
||||
3. 生成至少1个性能测试用例
|
||||
4. 生成至少1个安全测试用例
|
||||
5. 测试用例要包含明确的步骤和预期结果，步骤和预期结果必须符合Java语言特性（不要出现Python或其他语言的描述）
|
||||
5. 测试用例要包含明确的步骤和预期结果，步骤和预期结果必须符合Python语言特性（不要出现Java或其他语言的描述）
|
||||
6. 步骤和预期结果必须用自然语言描述，不得包含任何代码片段或代码块，不要出现 ```、assert、assertEquals 等代码语法
|
||||
7. 测试策略不要出现JUnit或者Java这种字眼，应该是针对需求的测试方法论和思路描述，覆盖计划要说明如何确保测试覆盖所有功能和边界情况
|
||||
7. 测试策略不要出现pytest或者Python这种字眼，应该是针对需求的测试方法论和思路描述，覆盖计划要说明如何确保测试覆盖所有功能和边界情况
|
||||
|
||||
返回ONLY JSON内容，不要有其他文字。"""
|
||||
|
||||
def generate_test_cases(self, requirement_analysis: RequirementAnalysis) -> TestCaseResult:
|
||||
"""
|
||||
基于需求分析生成测试用例
|
||||
|
||||
Args:
|
||||
requirement_analysis: PM Agent的分析结果
|
||||
|
||||
Returns:
|
||||
TestCaseResult: 包含测试用例的结果
|
||||
"""
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python QA 工程师，擅长为 Python 应用程序设计测试用例，所有测试步骤和预期结果必须基于 Python 语言特性（如 pytest、动态类型、异常机制等），输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(requirement_analysis)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=3000,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
result = _parse_json(response.choices[0].message.content, "QAAgent 解析失败")
|
||||
return TestCaseResult(**result)
|
||||
|
||||
def stream_generate_test_cases(self, requirement_analysis: RequirementAnalysis):
|
||||
"""
|
||||
流式版测试用例生成。yield 文本块(str)，最后 yield (None, TestCaseResult) 作为哨兵。
|
||||
"""
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python QA 工程师，擅长为 Python 应用程序设计测试用例，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(requirement_analysis)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=3000,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "QAAgent stream 解析失败")
|
||||
yield (None, TestCaseResult(**result))
|
||||
|
||||
def stream_refine(self, previous: TestCaseResult, feedback: str):
|
||||
"""
|
||||
流式版测试用例修改。yield 文本块(str)，最后 yield (None, TestCaseResult) 作为哨兵。
|
||||
"""
|
||||
prompt = f"""你是一个资深的Python QA工程师。以下是你之前输出的测试用例，用户对此有修改意见，请根据意见调整输出。
|
||||
|
||||
之前的测试用例：
|
||||
{json.dumps(previous, ensure_ascii=False, indent=2)}
|
||||
|
||||
用户的修改意见：
|
||||
{feedback}
|
||||
|
||||
请在原有基础上修改，保持 JSON 格式不变，返回完整的修改后结果。
|
||||
步骤和预期结果必须用自然语言描述，不得包含代码片段。
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python QA 工程师，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=3000,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "QAAgent stream_refine 解析失败")
|
||||
yield (None, TestCaseResult(**result))
|
||||
|
||||
def refine(self, previous: TestCaseResult, feedback: str) -> TestCaseResult:
|
||||
"""
|
||||
根据用户反馈修改已有的测试用例结果。
|
||||
|
||||
Args:
|
||||
previous: 上一次的测试用例结果
|
||||
feedback: 用户的修改意见
|
||||
|
||||
Returns:
|
||||
TestCaseResult: 修改后的测试用例结果
|
||||
"""
|
||||
prompt = f"""你是一个资深的Python QA工程师。以下是你之前输出的测试用例，用户对此有修改意见，请根据意见调整输出。
|
||||
|
||||
之前的测试用例：
|
||||
{json.dumps(previous, ensure_ascii=False, indent=2)}
|
||||
|
||||
用户的修改意见：
|
||||
{feedback}
|
||||
|
||||
请在原有基础上修改，保持 JSON 格式不变，返回完整的修改后结果。
|
||||
步骤和预期结果必须用自然语言描述，不得包含代码片段。
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Java QA 工程师，擅长为 Java 应用程序设计测试用例，所有测试步骤和预期结果必须基于 Java 语言特性（如 JUnit、强类型系统、异常机制等），输出必须是严格的 JSON 格式。"},
|
||||
{"role": "system", "content": "你是一个资深的 Python QA 工程师，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=3000,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
|
||||
content = response.choices[0].message.content
|
||||
|
||||
try:
|
||||
result = json.loads(content)
|
||||
return TestCaseResult(**result)
|
||||
except json.JSONDecodeError:
|
||||
import re
|
||||
json_match = re.search(r'\{.*\}', content, re.DOTALL)
|
||||
if json_match:
|
||||
result = json.loads(json_match.group())
|
||||
return TestCaseResult(**result)
|
||||
raise ValueError(f"无法解析QA Agent响应: {content}")
|
||||
result = _parse_json(response.choices[0].message.content, "QAAgent.refine 解析失败")
|
||||
return TestCaseResult(**result)
|
||||
|
||||
|
||||
class DevAgent:
|
||||
"""开发Agent - 生成Java代码和单元测试"""
|
||||
"""开发Agent - 生成Python代码和单元测试"""
|
||||
|
||||
def __init__(self):
|
||||
self.settings = get_settings()
|
||||
self.client = OpenAI(api_key=self.settings.api_key, base_url=self.settings.base_url)
|
||||
|
||||
def generate_code(
|
||||
self,
|
||||
requirement_analysis: RequirementAnalysis,
|
||||
test_cases: TestCaseResult
|
||||
) -> CodeGenerationResult:
|
||||
"""
|
||||
生成Java实现代码和单元测试代码
|
||||
|
||||
Args:
|
||||
requirement_analysis: PM Agent的分析结果
|
||||
test_cases: QA Agent的测试用例
|
||||
|
||||
Returns:
|
||||
CodeGenerationResult: 包含Java代码和单元测试代码
|
||||
"""
|
||||
|
||||
# 构建上下文
|
||||
def _build_prompt(self, requirement_analysis: RequirementAnalysis, test_cases: TestCaseResult) -> str:
|
||||
requirement_text = f"""
|
||||
功能需求：
|
||||
{chr(10).join(f"- {req}" for req in requirement_analysis["functional_requirements"])}
|
||||
@@ -213,7 +442,6 @@ class DevAgent:
|
||||
需求总结：
|
||||
{requirement_analysis["summary"]}
|
||||
"""
|
||||
|
||||
test_cases_text = chr(10).join(
|
||||
f"- [{c['test_id']}] {c['test_name']}（类型：{c.get('test_type', '未分类')}）｜预期结果：{c['expected_result']}"
|
||||
for c in test_cases["test_cases"]
|
||||
@@ -226,8 +454,7 @@ class DevAgent:
|
||||
关键测试用例列表：
|
||||
{test_cases_text}
|
||||
"""
|
||||
|
||||
prompt = f"""你是一个资深的Java开发工程师。根据以下需求和测试用例，生成高质量的Java实现代码和单元测试代码。
|
||||
return f"""你是一个资深的Python开发工程师。根据以下需求和测试用例，生成高质量的Python实现代码和单元测试代码。
|
||||
|
||||
{requirement_text}
|
||||
|
||||
@@ -235,61 +462,246 @@ class DevAgent:
|
||||
|
||||
请返回JSON格式的结果（必须是有效的JSON格式）：
|
||||
{{
|
||||
"java_code": "完整的Java实现代码（包含主类和必要的辅助类）",
|
||||
"unit_tests": "使用JUnit的单元测试代码",
|
||||
"implementation_notes": "实现说明和注意事项",
|
||||
"unit_tests_count": "生成的单元测试总数量（整数）",
|
||||
"passed_tests_count": "基于代码逻辑分析，预期可通过的单元测试数量（整数）"
|
||||
"java_code": "完整的Python实现代码（包含主模块和必要的辅助类/函数）",
|
||||
"unit_tests": "使用pytest的单元测试代码",
|
||||
"implementation_notes": "实现说明和注意事项"
|
||||
}}
|
||||
|
||||
Java代码要求：
|
||||
1. 使用 Java 11 语法和特性（如 var 局部变量类型推断、String::isBlank/strip/lines、Optional、Stream API、List.of/Map.of 等不可变集合工厂方法），不要使用 Java 8 以前的写法
|
||||
2. 包含详细的中文代码注释，所有注释内容必须用中文撰写，所有多行注释必须以 /* 开头、以 */ 结尾，Javadoc注释以 /** 开头、以 */ 结尾，绝对不能用单独的 / 作为注释结尾
|
||||
Python代码要求：
|
||||
1. 使用 Python 3.10+ 语法和特性（如 match/case、类型注解、dataclass、pathlib、f-string、列表推导等），代码风格遵循 PEP 8
|
||||
2. 包含详细的中文代码注释，类和函数必须有中文 docstring
|
||||
3. 包含异常处理
|
||||
4. 支持所有的功能需求
|
||||
5. 考虑非功能需求（性能、安全等）
|
||||
|
||||
单元测试要求：
|
||||
1. 使用 JUnit 5（jupiter），充分利用 @DisplayName、@ParameterizedTest、assertThrows 等特性
|
||||
2. 为每个公共方法生成测试
|
||||
1. 使用 pytest，充分利用 @pytest.mark.parametrize、pytest.raises、fixture 等特性
|
||||
2. 为每个公共函数/方法生成测试
|
||||
3. 包含正常情况、边缘情况和异常情况的测试
|
||||
4. 使用有意义的测试方法名称
|
||||
5. 每个测试类顶部加中文类级注释说明该类的测试范围
|
||||
4. 使用有意义的测试函数名称（如 test_xxx_when_xxx_should_xxx）
|
||||
5. 每个测试函数内用中文注释标注 Given / When / Then 三个阶段
|
||||
6. 测试代码要清晰易读
|
||||
7. 重要：业务代码文件保存名为 implementation.py，测试代码文件保存名为 test_implementation.py，测试文件必须使用 `from implementation import ...` 或 `import implementation` 导入业务代码
|
||||
|
||||
implementation_notes要求返回中文实现说明，内容要具体且有指导意义，不能只是简单的总结性描述，要包含对关键设计决策的解释和对复杂逻辑的说明。
|
||||
|
||||
返回ONLY JSON内容，不要有其他文字。"""
|
||||
|
||||
def generate_code(
|
||||
self,
|
||||
requirement_analysis: RequirementAnalysis,
|
||||
test_cases: TestCaseResult
|
||||
) -> CodeGenerationResult:
|
||||
"""
|
||||
生成Python实现代码和单元测试代码
|
||||
|
||||
Args:
|
||||
requirement_analysis: PM Agent的分析结果
|
||||
test_cases: QA Agent的测试用例
|
||||
|
||||
Returns:
|
||||
CodeGenerationResult: 包含Python代码和单元测试代码
|
||||
"""
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Java 11 开发工程师，擅长使用 Java 11 特性（var、Stream API、Optional、HttpClient、String 新方法等）编写高质量代码和单元测试，所有代码注释用中文，类名和方法名保持英文命名规范，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "system", "content": "你是一个资深的 Python 3.10+ 开发工程师，擅长使用 Python 类型注解、dataclass、pytest 等特性编写高质量代码和单元测试，所有代码注释用中文，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(requirement_analysis, test_cases)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=8192,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
result = _parse_json(response.choices[0].message.content, "DevAgent 解析失败")
|
||||
return CodeGenerationResult(**result)
|
||||
|
||||
def stream_generate_code(
|
||||
self,
|
||||
requirement_analysis: RequirementAnalysis,
|
||||
test_cases: TestCaseResult,
|
||||
):
|
||||
"""
|
||||
流式版代码生成。yield 文本块(str)，最后 yield (None, CodeGenerationResult) 作为哨兵。
|
||||
"""
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python 3.10+ 开发工程师，擅长使用 Python 类型注解、dataclass、pytest 等特性编写高质量代码，所有代码注释用中文，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(requirement_analysis, test_cases)}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=8192,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "DevAgent stream 解析失败")
|
||||
yield (None, CodeGenerationResult(**result))
|
||||
|
||||
def refine(
|
||||
self,
|
||||
previous: CodeGenerationResult,
|
||||
requirement_analysis: RequirementAnalysis,
|
||||
test_cases: TestCaseResult,
|
||||
feedback: str
|
||||
) -> CodeGenerationResult:
|
||||
"""
|
||||
根据用户反馈修改已有的代码生成结果。
|
||||
|
||||
Args:
|
||||
previous: 上一次的代码生成结果
|
||||
requirement_analysis: PM Agent的分析结果（供参考）
|
||||
test_cases: QA Agent的测试用例（供参考）
|
||||
feedback: 用户的修改意见
|
||||
|
||||
Returns:
|
||||
CodeGenerationResult: 修改后的代码生成结果
|
||||
"""
|
||||
refine_prompt = f"""以下是你之前生成的代码结果，用户对此有修改意见，请根据意见调整输出。
|
||||
|
||||
用户的修改意见：
|
||||
{feedback}
|
||||
|
||||
请在原有代码基础上修改，保持 JSON 格式不变，返回完整的修改后结果。
|
||||
继续遵守原有的 Python 3.10+、中文注释、pytest 等所有要求。
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python 3.10+ 开发工程师，所有代码注释用中文，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": self._build_prompt(requirement_analysis, test_cases)},
|
||||
{"role": "assistant", "content": json.dumps(previous, ensure_ascii=False)},
|
||||
{"role": "user", "content": refine_prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=8192,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
result = _parse_json(response.choices[0].message.content, "DevAgent.refine 解析失败")
|
||||
return CodeGenerationResult(**result)
|
||||
|
||||
|
||||
class FixAgent:
|
||||
"""自动修复Agent - 根据pytest失败信息修复Python代码"""
|
||||
|
||||
def __init__(self):
|
||||
self.settings = get_settings()
|
||||
self.client = OpenAI(api_key=self.settings.api_key, base_url=self.settings.base_url)
|
||||
|
||||
def fix(self, code_generation: CodeGenerationResult, test_output: str) -> CodeGenerationResult:
|
||||
"""
|
||||
根据pytest输出自动修复业务代码和/或测试代码。
|
||||
|
||||
Args:
|
||||
code_generation: 之前 DevAgent 的产出
|
||||
test_output: pytest 执行的完整输出
|
||||
|
||||
Returns:
|
||||
CodeGenerationResult: 修复后的完整代码
|
||||
"""
|
||||
prompt = f"""以下Python代码在运行单元测试时出现了失败，请根据pytest失败信息修复代码。
|
||||
|
||||
当前业务代码（implementation.py）：
|
||||
```python
|
||||
{code_generation["java_code"]}
|
||||
```
|
||||
|
||||
当前测试代码（test_implementation.py）：
|
||||
```python
|
||||
{code_generation["unit_tests"]}
|
||||
```
|
||||
|
||||
pytest 执行输出：
|
||||
{test_output[:4000]}
|
||||
|
||||
请分析失败原因并修复（业务代码和/或测试代码中的问题 都可以修复）。
|
||||
层层修复要径直至递推所有测试权都能通过。
|
||||
业务代码文件名为 implementation.py，测试文件名为 test_implementation.py，测试文件导入必须用 from implementation import ...。
|
||||
|
||||
返回 JSON 格式：
|
||||
{{
|
||||
"java_code": "修复后的完整Python业务代码",
|
||||
"unit_tests": "修复后的完整Python测试代码",
|
||||
"implementation_notes": "修复说明：具体说明了哪些问题，如何修复的"
|
||||
}}
|
||||
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
response = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python 3.10+ 开发工程师，擅长根据测试错误信息定位并修复 Bug，所有代码注释用中文，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=8192,
|
||||
response_format={"type": "json_object"}
|
||||
)
|
||||
result = _parse_json(response.choices[0].message.content, "FixAgent 解析失败")
|
||||
return CodeGenerationResult(**result)
|
||||
|
||||
content = response.choices[0].message.content
|
||||
def stream_fix(self, code_generation: CodeGenerationResult, test_output: str):
|
||||
"""
|
||||
流式修复。yield 文本块(str)，最后 yield (None, CodeGenerationResult) 作为哨兵。
|
||||
"""
|
||||
prompt = f"""以下Python代码在运行单元测试时出现了失败，请根据pytest失败信息修复代码。
|
||||
|
||||
当前业务代码（implementation.py）：
|
||||
```python
|
||||
{code_generation["java_code"]}
|
||||
```
|
||||
|
||||
try:
|
||||
result = json.loads(content)
|
||||
return CodeGenerationResult(**result)
|
||||
except json.JSONDecodeError:
|
||||
import re
|
||||
json_match = re.search(r'\{.*\}', content, re.DOTALL)
|
||||
if json_match:
|
||||
result = json.loads(json_match.group())
|
||||
return CodeGenerationResult(**result)
|
||||
raise ValueError(f"无法解析Dev Agent响应: {content}")
|
||||
当前测试代码（test_implementation.py）：
|
||||
```python
|
||||
{code_generation["unit_tests"]}
|
||||
```
|
||||
|
||||
pytest 执行输出：
|
||||
{test_output[:4000]}
|
||||
|
||||
请分析失败原因并修复（业务代码和/或测试代码中的问题 都可以修复）。
|
||||
层层修复要径直至递推所有测试权都能通过。
|
||||
业务代码文件名为 implementation.py，测试文件名为 test_implementation.py，测试文件导入必须用 from implementation import ...。
|
||||
|
||||
返回 JSON 格式：
|
||||
{{
|
||||
"java_code": "修复后的完整Python业务代码",
|
||||
"unit_tests": "修复后的完整Python测试代码",
|
||||
"implementation_notes": "修复说明：具体说明了哪些问题，如何修复的"
|
||||
}}
|
||||
|
||||
返回 ONLY JSON，不要有其他文字。"""
|
||||
|
||||
stream = self.client.chat.completions.create(
|
||||
model=self.settings.model,
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个资深的 Python 3.10+ 开发工程师，擅长根据测试错误信息定位并修复 Bug，所有代码注释用中文，输出必须是严格的 JSON 格式。"},
|
||||
{"role": "user", "content": prompt}
|
||||
],
|
||||
temperature=0.2,
|
||||
max_tokens=8192,
|
||||
response_format={"type": "json_object"},
|
||||
stream=True,
|
||||
)
|
||||
full_text = ""
|
||||
for chunk in stream:
|
||||
delta = chunk.choices[0].delta.content or ""
|
||||
if delta:
|
||||
full_text += delta
|
||||
yield delta
|
||||
result = _parse_json(full_text, "FixAgent stream 解析失败")
|
||||
yield (None, CodeGenerationResult(**result))
|
||||
|
||||
|
||||
async def orchestrate_agents(simple_requirement: str) -> dict:
|
||||
"""
|
||||
编排三个Agent的工作流程
|
||||
编排三个Agent的工作流程（保留原有全量接口）
|
||||
|
||||
Args:
|
||||
simple_requirement: 用户的简单需求描述
|
||||
|
||||
@@ -15,11 +15,12 @@ _TYPE_EMOJI = {
|
||||
|
||||
|
||||
def _post(data: dict):
|
||||
requests.post(
|
||||
webhook_url,
|
||||
headers={"Content-Type": "application/json"},
|
||||
data=json.dumps(data, ensure_ascii=False)
|
||||
)
|
||||
return
|
||||
# requests.post(
|
||||
# webhook_url,
|
||||
# headers={"Content-Type": "application/json"},
|
||||
# data=json.dumps(data, ensure_ascii=False)
|
||||
# )
|
||||
|
||||
|
||||
def _make_card(title: str, color: str, elements: list) -> dict:
|
||||
@@ -136,32 +137,6 @@ def send_generate_code(code_result: CodeGenerationResult):
|
||||
[{"tag": "markdown", "content": f"```java\n{code_result['unit_tests']}\n```"}]
|
||||
))
|
||||
|
||||
# 单元测试执行结果
|
||||
try:
|
||||
total = int(code_result["unit_tests_count"])
|
||||
passed = int(code_result["passed_tests_count"])
|
||||
failed = total - passed
|
||||
rate = f"{passed / total * 100:.0f}%" if total > 0 else "—"
|
||||
status = "✅ 全部通过" if failed == 0 else f"⚠️ {failed} 个未通过"
|
||||
failed_display = f"{failed} ❌" if failed > 0 else "0"
|
||||
except (ValueError, TypeError):
|
||||
total = code_result["unit_tests_count"]
|
||||
passed = code_result["passed_tests_count"]
|
||||
failed_display, rate, status = "—", "—", "—"
|
||||
|
||||
result_md = (
|
||||
f"**总用例数：** {total}\n\n"
|
||||
f"**通过：** {passed} ✅\n\n"
|
||||
f"**未通过：** {failed_display}\n\n"
|
||||
f"**通过率：** {rate}\n\n"
|
||||
f"---\n\n"
|
||||
f"**整体状态：** {status}"
|
||||
)
|
||||
_post(_make_card(
|
||||
"📊 代码生成结果 — 单元测试执行结果", "green",
|
||||
[{"tag": "markdown", "content": result_md}]
|
||||
))
|
||||
|
||||
|
||||
def send_test_cases(test_case: TestCaseResult):
|
||||
_post(build_full_feishu_card(test_case))
|
||||
|
||||
@@ -22,5 +22,3 @@ class CodeGenerationResult(TypedDict):
|
||||
java_code: str
|
||||
unit_tests: str
|
||||
implementation_notes: str
|
||||
unit_tests_count: str
|
||||
passed_tests_count: str
|
||||
|
||||
0
app/routers/__init__.py
Normal file
0
app/routers/__init__.py
Normal file
514
app/routers/session.py
Normal file
514
app/routers/session.py
Normal file
@@ -0,0 +1,514 @@
|
||||
"""
|
||||
app/routers/session.py - 交互式会话路由
|
||||
"""
|
||||
|
||||
import logging
|
||||
import json
|
||||
from fastapi import APIRouter, HTTPException
|
||||
from fastapi.responses import StreamingResponse
|
||||
from pydantic import BaseModel
|
||||
from typing import Optional
|
||||
|
||||
from app.session import SessionStore, SessionStatus
|
||||
from app.agents import ClarifyAgent, PMAgent, QAAgent, DevAgent, FixAgent
|
||||
from app.test_runner import run_python_tests
|
||||
from app.message import (
|
||||
send_workflow_start,
|
||||
send_requirement_result,
|
||||
send_test_cases,
|
||||
send_generate_code,
|
||||
)
|
||||
|
||||
router = APIRouter(prefix="/session", tags=["session"])
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
# ---------- 请求 / 响应模型 ----------
|
||||
|
||||
class StartRequest(BaseModel):
|
||||
requirement: str
|
||||
|
||||
class ClarifyRequest(BaseModel):
|
||||
message: str
|
||||
|
||||
class RefineRequest(BaseModel):
|
||||
feedback: str
|
||||
|
||||
|
||||
class SessionResponse(BaseModel):
|
||||
session_id: str
|
||||
status: str
|
||||
ready: bool = False
|
||||
question: Optional[str] = None # 当 ready=False 时返回追问
|
||||
data: Optional[dict] = None # 当前阶段产出
|
||||
|
||||
|
||||
# ---------- 工具函数 ----------
|
||||
|
||||
def _get_session_or_404(session_id: str):
|
||||
session = SessionStore.get(session_id)
|
||||
if not session:
|
||||
raise HTTPException(status_code=404, detail="会话不存在或已过期")
|
||||
return session
|
||||
|
||||
def _require_status(session, *allowed: SessionStatus):
|
||||
if session.status not in allowed:
|
||||
raise HTTPException(
|
||||
status_code=400,
|
||||
detail=f"当前状态 [{session.status}] 不允许此操作，允许的状态: {[s.value for s in allowed]}"
|
||||
)
|
||||
|
||||
|
||||
# ---------- 接口 ----------
|
||||
|
||||
@router.post("/start", response_model=SessionResponse)
|
||||
def start_session(body: StartRequest):
|
||||
"""创建会话，AI 判断需求是否完整，不够则追问。"""
|
||||
session = SessionStore.create(body.requirement)
|
||||
agent = ClarifyAgent()
|
||||
result = agent.start(body.requirement)
|
||||
|
||||
q = result.get("question", "")
|
||||
if q:
|
||||
session.clarify_history.append({"role": "assistant", "content": q})
|
||||
session.clarified_requirement = result.get("clarified_requirement", body.requirement)
|
||||
|
||||
if result.get("ready"):
|
||||
session.status = SessionStatus.PM_READY
|
||||
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=result.get("ready", False),
|
||||
question=result.get("question") or None,
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/clarify", response_model=SessionResponse)
|
||||
def clarify(session_id: str, body: ClarifyRequest):
|
||||
"""用户补充需求，AI 继续判断是否够了。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.CLARIFYING)
|
||||
|
||||
session.clarify_history.append({"role": "user", "content": body.message})
|
||||
|
||||
agent = ClarifyAgent()
|
||||
result = agent.continue_clarify(session.clarify_history, body.message)
|
||||
|
||||
q = result.get("question", "")
|
||||
if q:
|
||||
session.clarify_history.append({"role": "assistant", "content": q})
|
||||
session.clarified_requirement = result.get("clarified_requirement", session.clarified_requirement)
|
||||
|
||||
if result.get("ready"):
|
||||
session.status = SessionStatus.PM_READY
|
||||
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=result.get("ready", False),
|
||||
question=result.get("question") or None,
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/pm/stream")
|
||||
def pm_stream(session_id: str):
|
||||
"""流式返回 PM Agent 分析过程和结果（SSE）。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_READY)
|
||||
|
||||
send_workflow_start(session.clarified_requirement)
|
||||
agent = PMAgent()
|
||||
simple_requirement = session.clarified_requirement
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_analyze(simple_requirement):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.requirement_analysis = result
|
||||
send_requirement_result(result)
|
||||
session.status = SessionStatus.PM_DONE
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'requirement_analysis': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/pm/run", response_model=SessionResponse)
|
||||
def pm_run(session_id: str):
|
||||
"""触发 PM Agent 分析需求。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_READY)
|
||||
|
||||
send_workflow_start(session.clarified_requirement)
|
||||
|
||||
agent = PMAgent()
|
||||
session.requirement_analysis = agent.analyze_requirement(session.clarified_requirement)
|
||||
send_requirement_result(session.requirement_analysis)
|
||||
|
||||
session.status = SessionStatus.PM_DONE
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"requirement_analysis": session.requirement_analysis},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/pm/refine/stream")
|
||||
def pm_refine_stream(session_id: str, feedback: str):
|
||||
"""流式修改 PM 产出（SSE），feedback 经由 query param 传入。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_DONE)
|
||||
|
||||
agent = PMAgent()
|
||||
previous = session.requirement_analysis
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_refine(previous, feedback):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.requirement_analysis = result
|
||||
send_requirement_result(result)
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'requirement_analysis': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/pm/refine", response_model=SessionResponse)
|
||||
def pm_refine(session_id: str, body: RefineRequest):
|
||||
"""根据反馈修改 PM 产出。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_DONE)
|
||||
|
||||
agent = PMAgent()
|
||||
session.requirement_analysis = agent.refine(session.requirement_analysis, body.feedback)
|
||||
send_requirement_result(session.requirement_analysis)
|
||||
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"requirement_analysis": session.requirement_analysis},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/qa/stream")
|
||||
def qa_stream(session_id: str):
|
||||
"""流式返回 QA Agent 测试用例生成过程（SSE）。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_DONE)
|
||||
|
||||
if not session.requirement_analysis:
|
||||
raise HTTPException(status_code=400, detail="PM Agent 产出不存在")
|
||||
|
||||
agent = QAAgent()
|
||||
req_analysis = session.requirement_analysis
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_generate_test_cases(req_analysis):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.test_cases = result
|
||||
send_test_cases(result)
|
||||
session.status = SessionStatus.QA_DONE
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'test_cases': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/qa/run", response_model=SessionResponse)
|
||||
def qa_run(session_id: str):
|
||||
"""触发 QA Agent 生成测试用例。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.PM_DONE)
|
||||
|
||||
if not session.requirement_analysis:
|
||||
raise HTTPException(status_code=400, detail="PM Agent 产出不存在")
|
||||
|
||||
agent = QAAgent()
|
||||
session.test_cases = agent.generate_test_cases(session.requirement_analysis)
|
||||
send_test_cases(session.test_cases)
|
||||
|
||||
session.status = SessionStatus.QA_DONE
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"test_cases": session.test_cases},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/qa/refine/stream")
|
||||
def qa_refine_stream(session_id: str, feedback: str):
|
||||
"""流式修改 QA 产出（SSE），feedback 经由 query param 传入。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.QA_DONE)
|
||||
|
||||
agent = QAAgent()
|
||||
previous = session.test_cases
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_refine(previous, feedback):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.test_cases = result
|
||||
send_test_cases(result)
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'test_cases': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/qa/refine", response_model=SessionResponse)
|
||||
def qa_refine(session_id: str, body: RefineRequest):
|
||||
"""根据反馈修改 QA 产出。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.QA_DONE)
|
||||
|
||||
agent = QAAgent()
|
||||
session.test_cases = agent.refine(session.test_cases, body.feedback)
|
||||
send_test_cases(session.test_cases)
|
||||
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"test_cases": session.test_cases},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/dev/stream")
|
||||
def dev_stream(session_id: str):
|
||||
"""流式返回 Dev Agent 代码生成过程（SSE）。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.QA_DONE)
|
||||
|
||||
if not session.requirement_analysis or not session.test_cases:
|
||||
raise HTTPException(status_code=400, detail="PM / QA Agent 产出不完整")
|
||||
|
||||
agent = DevAgent()
|
||||
req_analysis = session.requirement_analysis
|
||||
test_cases = session.test_cases
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_generate_code(req_analysis, test_cases):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.code_generation = result
|
||||
send_generate_code(result)
|
||||
session.status = SessionStatus.DEV_DONE
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'code_generation': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/dev/run", response_model=SessionResponse)
|
||||
def dev_run(session_id: str):
|
||||
"""触发 Dev Agent 生成代码。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.QA_DONE)
|
||||
|
||||
if not session.requirement_analysis or not session.test_cases:
|
||||
raise HTTPException(status_code=400, detail="PM / QA Agent 产出不完整")
|
||||
|
||||
agent = DevAgent()
|
||||
session.code_generation = agent.generate_code(session.requirement_analysis, session.test_cases)
|
||||
send_generate_code(session.code_generation)
|
||||
|
||||
session.status = SessionStatus.DEV_DONE
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"code_generation": session.code_generation},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/dev/refine", response_model=SessionResponse)
|
||||
def dev_refine(session_id: str, body: RefineRequest):
|
||||
"""根据反馈修改 Dev 产出。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.DEV_DONE)
|
||||
|
||||
agent = DevAgent()
|
||||
session.code_generation = agent.refine(
|
||||
session.code_generation,
|
||||
session.requirement_analysis,
|
||||
session.test_cases,
|
||||
body.feedback,
|
||||
)
|
||||
send_generate_code(session.code_generation)
|
||||
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"code_generation": session.code_generation},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/test/run", response_model=SessionResponse)
|
||||
def test_run(session_id: str):
|
||||
"""在临时目录中真实执行 pytest，返回测试结果。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
|
||||
if not session.code_generation:
|
||||
raise HTTPException(status_code=400, detail="Dev Agent 产出不存在")
|
||||
|
||||
result = run_python_tests(
|
||||
session.code_generation["java_code"],
|
||||
session.code_generation["unit_tests"],
|
||||
)
|
||||
session.test_execution = result
|
||||
session.status = SessionStatus.TEST_DONE
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"test_execution": result},
|
||||
)
|
||||
|
||||
|
||||
@router.post("/{session_id}/test/fix", response_model=SessionResponse)
|
||||
def test_fix(session_id: str):
|
||||
"""调用 FixAgent 根据测试失败信息自动修复代码。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.TEST_DONE)
|
||||
|
||||
if not session.test_execution:
|
||||
raise HTTPException(status_code=400, detail="尚未执行测试")
|
||||
if session.test_execution.get("success"):
|
||||
raise HTTPException(status_code=400, detail="测试已全部通过，无需修复")
|
||||
|
||||
agent = FixAgent()
|
||||
session.code_generation = agent.fix(
|
||||
session.code_generation,
|
||||
session.test_execution["output"],
|
||||
)
|
||||
session.status = SessionStatus.DEV_DONE # 修复后重置为 dev_done，可再次测试
|
||||
session.touch()
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=True,
|
||||
data={"code_generation": session.code_generation},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}/test/fix/stream")
|
||||
def test_fix_stream(session_id: str):
|
||||
"""流式返回 FixAgent 代码修复过程（SSE）。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
_require_status(session, SessionStatus.TEST_DONE)
|
||||
|
||||
if not session.test_execution:
|
||||
raise HTTPException(status_code=400, detail="尚未执行测试")
|
||||
if session.test_execution.get("success"):
|
||||
raise HTTPException(status_code=400, detail="测试已全部通过，无需修复")
|
||||
|
||||
agent = FixAgent()
|
||||
code_generation = session.code_generation
|
||||
test_output = session.test_execution["output"]
|
||||
|
||||
def generate():
|
||||
try:
|
||||
result = None
|
||||
for item in agent.stream_fix(code_generation, test_output):
|
||||
if isinstance(item, tuple):
|
||||
_, result = item
|
||||
else:
|
||||
yield f"data: {json.dumps({'type': 'chunk', 'text': item}, ensure_ascii=False)}\n\n"
|
||||
session.code_generation = result
|
||||
session.status = SessionStatus.DEV_DONE
|
||||
session.touch()
|
||||
yield f"data: {json.dumps({'type': 'done', 'status': session.status, 'data': {'code_generation': result}}, ensure_ascii=False)}\n\n"
|
||||
except Exception as e:
|
||||
yield f"data: {json.dumps({'type': 'error', 'message': str(e)})}\n\n"
|
||||
|
||||
return StreamingResponse(
|
||||
generate(),
|
||||
media_type="text/event-stream",
|
||||
headers={"Cache-Control": "no-cache", "X-Accel-Buffering": "no"},
|
||||
)
|
||||
|
||||
|
||||
@router.get("/{session_id}", response_model=SessionResponse)
|
||||
def get_session(session_id: str):
|
||||
"""获取当前会话状态和所有产出。"""
|
||||
session = _get_session_or_404(session_id)
|
||||
return SessionResponse(
|
||||
session_id=session.session_id,
|
||||
status=session.status,
|
||||
ready=session.status != SessionStatus.CLARIFYING,
|
||||
data={
|
||||
"raw_requirement": session.raw_requirement,
|
||||
"clarify_history": session.clarify_history,
|
||||
"requirement_analysis": session.requirement_analysis,
|
||||
"test_cases": session.test_cases,
|
||||
"code_generation": session.code_generation,
|
||||
"test_execution": session.test_execution,
|
||||
},
|
||||
)
|
||||
68
app/session.py
Normal file
68
app/session.py
Normal file
@@ -0,0 +1,68 @@
|
||||
"""
|
||||
app/session.py - 会话状态管理
|
||||
"""
|
||||
|
||||
import uuid
|
||||
import time
|
||||
from enum import Enum
|
||||
from typing import Optional
|
||||
from app.models import RequirementAnalysis, TestCaseResult, CodeGenerationResult
|
||||
|
||||
|
||||
class SessionStatus(str, Enum):
|
||||
CLARIFYING = "clarifying" # 正在澄清需求
|
||||
PM_READY = "pm_ready" # 需求澄清完毕，可运行 PM Agent
|
||||
PM_DONE = "pm_done" # PM Agent 已完成
|
||||
QA_READY = "qa_ready" # 可运行 QA Agent
|
||||
QA_DONE = "qa_done" # QA Agent 已完成
|
||||
DEV_READY = "dev_ready" # 可运行 Dev Agent
|
||||
DEV_DONE = "dev_done" # Dev Agent 已完成
|
||||
TEST_DONE = "test_done" # 单元测试执行完成
|
||||
|
||||
class Session:
|
||||
def __init__(self, session_id: str, raw_requirement: str):
|
||||
self.session_id: str = session_id
|
||||
self.raw_requirement: str = raw_requirement # 用户最原始的需求
|
||||
self.clarified_requirement: str = raw_requirement # 经过澄清补充后的完整需求
|
||||
self.clarify_history: list[dict] = [] # 澄清对话历史
|
||||
self.status: SessionStatus = SessionStatus.CLARIFYING
|
||||
self.created_at: float = time.time()
|
||||
self.updated_at: float = time.time()
|
||||
|
||||
# 各 Agent 产出
|
||||
self.requirement_analysis: Optional[RequirementAnalysis] = None
|
||||
self.test_cases: Optional[TestCaseResult] = None
|
||||
self.code_generation: Optional[CodeGenerationResult] = None
|
||||
self.test_execution: Optional[dict] = None # pytest 执行结果
|
||||
|
||||
def touch(self):
|
||||
self.updated_at = time.time()
|
||||
|
||||
|
||||
class SessionStore:
|
||||
_store: dict[str, Session] = {}
|
||||
_TTL_SECONDS = 7200 # 2小时过期
|
||||
|
||||
@classmethod
|
||||
def create(cls, raw_requirement: str) -> Session:
|
||||
session_id = str(uuid.uuid4())
|
||||
session = Session(session_id, raw_requirement)
|
||||
cls._store[session_id] = session
|
||||
cls._evict_expired()
|
||||
return session
|
||||
|
||||
@classmethod
|
||||
def get(cls, session_id: str) -> Optional[Session]:
|
||||
session = cls._store.get(session_id)
|
||||
if session and time.time() - session.updated_at > cls._TTL_SECONDS:
|
||||
del cls._store[session_id]
|
||||
return None
|
||||
return session
|
||||
|
||||
@classmethod
|
||||
def _evict_expired(cls):
|
||||
now = time.time()
|
||||
expired = [sid for sid, s in cls._store.items()
|
||||
if now - s.updated_at > cls._TTL_SECONDS]
|
||||
for sid in expired:
|
||||
del cls._store[sid]
|
||||
80
app/test_runner.py
Normal file
80
app/test_runner.py
Normal file
@@ -0,0 +1,80 @@
|
||||
"""
|
||||
app/test_runner.py - 在临时目录中真实执行 pytest 单元测试
|
||||
"""
|
||||
import os
|
||||
import re
|
||||
import sys
|
||||
import tempfile
|
||||
import subprocess
|
||||
|
||||
|
||||
def run_python_tests(python_code: str, test_code: str) -> dict:
|
||||
"""
|
||||
将业务代码写入 implementation.py，测试代码写入 test_implementation.py，
|
||||
在隔离的临时目录中用 pytest 执行，返回结构化测试结果。
|
||||
|
||||
Returns:
|
||||
dict: {
|
||||
success: bool,
|
||||
passed: int,
|
||||
failed: int,
|
||||
errors: int,
|
||||
total: int,
|
||||
output: str # pytest 完整输出
|
||||
}
|
||||
"""
|
||||
with tempfile.TemporaryDirectory() as tmpdir:
|
||||
impl_path = os.path.join(tmpdir, "implementation.py")
|
||||
test_path = os.path.join(tmpdir, "test_implementation.py")
|
||||
conftest_path = os.path.join(tmpdir, "conftest.py")
|
||||
|
||||
with open(impl_path, "w", encoding="utf-8") as f:
|
||||
f.write(python_code)
|
||||
|
||||
with open(test_path, "w", encoding="utf-8") as f:
|
||||
f.write(test_code)
|
||||
|
||||
# conftest.py 确保 tmpdir 在 sys.path 首位，解决模块导入问题
|
||||
with open(conftest_path, "w", encoding="utf-8") as f:
|
||||
f.write("import sys, os\nsys.path.insert(0, os.path.dirname(__file__))\n")
|
||||
|
||||
try:
|
||||
proc = subprocess.run(
|
||||
[sys.executable, "-m", "pytest", "test_implementation.py",
|
||||
"-v", "--tb=short", "--no-header", "--color=no"],
|
||||
cwd=tmpdir,
|
||||
capture_output=True,
|
||||
text=True,
|
||||
timeout=60,
|
||||
env={**os.environ, "PYTHONPATH": tmpdir},
|
||||
)
|
||||
output = proc.stdout
|
||||
if proc.stderr.strip():
|
||||
output += "\n--- stderr ---\n" + proc.stderr
|
||||
except subprocess.TimeoutExpired:
|
||||
return {
|
||||
"success": False,
|
||||
"passed": 0, "failed": 0, "errors": 1, "total": 0,
|
||||
"output": "❌ 测试执行超时（超过 60 秒）",
|
||||
}
|
||||
except FileNotFoundError:
|
||||
return {
|
||||
"success": False,
|
||||
"passed": 0, "failed": 0, "errors": 1, "total": 0,
|
||||
"output": "❌ 未找到 pytest，请确保已安装：pip install pytest",
|
||||
}
|
||||
|
||||
passed = int(m.group(1)) if (m := re.search(r"(\d+) passed", output)) else 0
|
||||
failed = int(m.group(1)) if (m := re.search(r"(\d+) failed", output)) else 0
|
||||
errors = int(m.group(1)) if (m := re.search(r"(\d+) error", output)) else 0
|
||||
total = passed + failed + errors
|
||||
success = (failed == 0 and errors == 0 and total > 0)
|
||||
|
||||
return {
|
||||
"success": success,
|
||||
"passed": passed,
|
||||
"failed": failed,
|
||||
"errors": errors,
|
||||
"total": total,
|
||||
"output": output,
|
||||
}
|
||||
4
main.py
4
main.py
@@ -9,6 +9,7 @@ from pydantic import BaseModel
|
||||
|
||||
from app.agents import orchestrate_agents
|
||||
from app.config import get_settings
|
||||
from app.routers import session as session_router
|
||||
|
||||
# 初始化日志
|
||||
logging.basicConfig(level=logging.INFO)
|
||||
@@ -33,6 +34,9 @@ app.add_middleware(
|
||||
allow_headers=["*"],
|
||||
)
|
||||
|
||||
# 注册交互式会话路由
|
||||
app.include_router(session_router.router)
|
||||
|
||||
|
||||
class FullWorkflowResponse(BaseModel):
|
||||
"""完整工作流响应"""
|
||||
|
||||
Reference in New Issue
Block a user