189 lines
6.5 KiB
Python
189 lines
6.5 KiB
Python
import dramatiq
|
||
import json
|
||
import time
|
||
from openai import OpenAI
|
||
from django.db import transaction
|
||
from django.utils import timezone
|
||
from utils.shortcuts import get_env, DRAMATIQ_WORKER_ARGS
|
||
from .models import FlowchartSubmission, FlowchartSubmissionStatus
|
||
|
||
@dramatiq.actor(**DRAMATIQ_WORKER_ARGS(max_retries=3))
|
||
def evaluate_flowchart_task(submission_id):
|
||
"""异步AI评分任务"""
|
||
try:
|
||
submission = FlowchartSubmission.objects.get(id=submission_id)
|
||
|
||
# 更新状态为处理中
|
||
submission.status = FlowchartSubmissionStatus.PROCESSING
|
||
submission.save()
|
||
|
||
start_time = time.time()
|
||
|
||
# 使用固定评分标准
|
||
system_prompt = build_evaluation_prompt(submission.problem)
|
||
|
||
# 构建用户提示词，包含标准答案对比
|
||
user_prompt = f"""
|
||
请对以下Mermaid流程图进行评分：
|
||
|
||
学生提交的流程图：
|
||
```mermaid
|
||
{submission.mermaid_code}
|
||
```
|
||
|
||
标准答案参考：
|
||
```mermaid
|
||
{submission.problem.mermaid_code}
|
||
```
|
||
"""
|
||
# 如果有流程图提示，添加到提示词中
|
||
if submission.problem.flowchart_hint:
|
||
user_prompt += f"""
|
||
|
||
设计提示：{submission.problem.flowchart_hint}
|
||
"""
|
||
|
||
user_prompt += """
|
||
|
||
请按照评分标准进行详细评估，并给出0-100的分数。
|
||
"""
|
||
|
||
# 调用AI进行评分
|
||
api_key = get_env("AI_KEY")
|
||
if not api_key:
|
||
raise Exception("AI_KEY is not set")
|
||
|
||
client = OpenAI(api_key=api_key, base_url="https://api.deepseek.com")
|
||
|
||
response = client.chat.completions.create(
|
||
model="deepseek-chat",
|
||
messages=[
|
||
{"role": "system", "content": system_prompt},
|
||
{"role": "user", "content": user_prompt}
|
||
],
|
||
temperature=0.3,
|
||
)
|
||
|
||
ai_response = response.choices[0].message.content
|
||
score_data = parse_ai_evaluation_response(ai_response)
|
||
|
||
processing_time = time.time() - start_time
|
||
|
||
# 保存评分结果
|
||
with transaction.atomic():
|
||
submission.ai_score = score_data['score']
|
||
submission.ai_grade = score_data['grade']
|
||
submission.ai_feedback = score_data['feedback']
|
||
submission.ai_suggestions = score_data.get('suggestions', '')
|
||
submission.ai_criteria_details = score_data.get('criteria_details', {})
|
||
submission.ai_provider = 'deepseek'
|
||
submission.ai_model = 'deepseek-chat'
|
||
submission.processing_time = processing_time
|
||
submission.status = FlowchartSubmissionStatus.COMPLETED
|
||
submission.evaluation_time = timezone.now()
|
||
submission.save()
|
||
|
||
# 推送评分完成通知
|
||
from utils.websocket import push_flowchart_evaluation_update
|
||
push_flowchart_evaluation_update(
|
||
submission_id=str(submission.id),
|
||
user_id=submission.user_id,
|
||
data={
|
||
"type": "flowchart_evaluation_completed",
|
||
"submission_id": str(submission.id),
|
||
"score": score_data['score'],
|
||
"grade": score_data['grade'],
|
||
"feedback": score_data['feedback'],
|
||
"suggestions": score_data['suggestions'],
|
||
"criteria_details": score_data['criteria_details'],
|
||
}
|
||
)
|
||
|
||
except Exception as e:
|
||
# 处理失败
|
||
submission.status = FlowchartSubmissionStatus.FAILED
|
||
submission.save()
|
||
|
||
# 推送错误通知
|
||
from utils.websocket import push_flowchart_evaluation_update
|
||
push_flowchart_evaluation_update(
|
||
submission_id=str(submission.id),
|
||
user_id=submission.user_id,
|
||
data={
|
||
"type": "flowchart_evaluation_failed",
|
||
"submission_id": str(submission.id),
|
||
"error": str(e)
|
||
}
|
||
)
|
||
raise e
|
||
|
||
def build_evaluation_prompt(problem):
|
||
"""构建AI评分提示词 - 使用固定标准"""
|
||
|
||
# 使用固定的评分标准
|
||
criteria_text = """
|
||
- 逻辑正确性 (权重: 1.0, 最高分: 40): 检查流程图的逻辑是否正确，包括条件判断、循环结构等
|
||
- 完整性 (权重: 0.8, 最高分: 30): 检查流程图是否包含所有必要的步骤和分支
|
||
- 规范性 (权重: 0.6, 最高分: 20): 检查流程图符号使用是否规范，是否符合标准
|
||
- 清晰度 (权重: 0.4, 最高分: 10): 评估流程图的整体布局和连线情况（不用考虑节点ID是否复杂）
|
||
"""
|
||
|
||
return f"""
|
||
你是一个专业的编程教学助手，负责评估学生提交的Mermaid流程图。
|
||
|
||
评分标准：
|
||
{criteria_text}
|
||
|
||
评分要求：
|
||
1. 仔细分析流程图的逻辑正确性、完整性和清晰度
|
||
2. 检查是否涵盖了题目的所有要求
|
||
3. 评估流程图的规范性和可读性
|
||
4. 给出0-100的分数
|
||
5. 提供详细的反馈和改进建议
|
||
|
||
评分等级：
|
||
- S级 (90-100分): 优秀，逻辑清晰，完全符合要求
|
||
- A级 (80-89分): 良好，基本符合要求，有少量改进空间
|
||
- B级 (70-79分): 及格，基本正确但存在一些问题
|
||
- C级 (0-69分): 需要改进，存在明显问题
|
||
|
||
请以JSON格式返回评分结果：
|
||
{{
|
||
"score": 85,
|
||
"grade": "A",
|
||
"feedback": "详细的反馈内容",
|
||
"suggestions": "改进建议",
|
||
"criteria_details": {{
|
||
"逻辑正确性": {{"score": 35, "max": 40, "comment": "逻辑基本正确"}},
|
||
"完整性": {{"score": 25, "max": 30, "comment": "缺少部分步骤"}},
|
||
"规范性": {{"score": 18, "max": 20, "comment": "符号使用规范"}},
|
||
"清晰度": {{"score": 8, "max": 10, "comment": "布局清晰"}}
|
||
}}
|
||
}}
|
||
"""
|
||
|
||
def parse_ai_evaluation_response(ai_response):
|
||
"""解析AI评分响应"""
|
||
try:
|
||
import re
|
||
json_match = re.search(r'\{.*\}', ai_response, re.DOTALL)
|
||
if json_match:
|
||
data = json.loads(json_match.group())
|
||
else:
|
||
data = {
|
||
"score": 60,
|
||
"grade": "C",
|
||
"feedback": "AI评分解析失败，请重新提交",
|
||
"suggestions": "",
|
||
"criteria_details": {}
|
||
}
|
||
return data
|
||
except Exception:
|
||
return {
|
||
"score": 60,
|
||
"grade": "C",
|
||
"feedback": "AI评分解析失败，请重新提交",
|
||
"suggestions": "",
|
||
"criteria_details": {}
|
||
}
|