| 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105 |
- """plan_classifier.py
- 用于鉴定上传的 Markdown 文件是否为训练计划。
- 采用双层过滤架构:
- 1. Level 1 - 关键字快筛(正则表达式 / 词频)
- 2. Level 2 - LLM (Lite) 语义鉴定
- """
- import os
- import json
- import re
- from typing import Any
- LLM_LITE_API_KEY: str | None = os.getenv("LLM_LITE_API_KEY") or os.getenv("LLM_API_KEY")
- LLM_LITE_BASE_URL: str = (os.getenv("LLM_LITE_BASE_URL") or "https://api.deepseek.com").rstrip("/")
- LLM_LITE_MODE: str = os.getenv("LLM_LITE_MODE_ID", "deepseek-chat")
- # 关键词仅用于快速排除明显无关的文件;不得据此直接接收文件。
- TRAINING_KEYWORDS = [
- "组间歇", "重量", "卧推", "硬拉", "深蹲", "划船", "哑铃", "杠铃", "RM", "组数",
- "力竭", "热身", "拉伸", "核心肌群", "训练", "计划", "运动", "健身", "锻炼", "动作",
- "次", "分钟",
- ]
- def _level1_keyword_check(text: str) -> tuple[bool, str] | None:
- """仅拒绝不含任何训练内容的文件,其余全部交给语义鉴定。"""
- if not any(keyword.lower() in text.lower() for keyword in TRAINING_KEYWORDS):
- return False, "完全不包含任何运动、健身相关的常见词汇"
- return None
- def _extract_json(text: str) -> dict[str, Any] | None:
- md_match = re.search(r"```(?:json)?\s*(\{[\s\S]+?\})\s*```", text)
- if md_match:
- try:
- return json.loads(md_match.group(1))
- except json.JSONDecodeError:
- pass
- brace_match = re.search(r"\{[\s\S]+?\}", text)
- if brace_match:
- try:
- return json.loads(brace_match.group(0))
- except json.JSONDecodeError:
- pass
- return None
- def _level2_llm_check(text: str) -> tuple[bool, str]:
- if not LLM_LITE_API_KEY:
- return False, "未配置 LLM API KEY 且无法通过规则验证"
-
- system_prompt = (
- "判断以下 Markdown 文本是否为一份可执行的健身/运动训练计划。\n"
- "仅仅提及动作名、重量、训练感受、健身知识、训练日志、复盘或文章,不是训练计划。\n"
- "只有包含面向未来执行的明确训练安排(例如动作、次数/时长/强度、顺序或频率)时才是训练计划。\n"
- "返回严格 JSON 格式:\n"
- "{\n"
- ' "is_plan": true 或 false,\n'
- ' "reason": "一句话说明理由"\n'
- "}"
- )
-
- try:
- import requests
- url = f"{LLM_LITE_BASE_URL}/chat/completions"
- headers = {
- "Authorization": f"Bearer {LLM_LITE_API_KEY}",
- "Content-Type": "application/json",
- }
- body = {
- "model": LLM_LITE_MODE,
- "messages": [
- {"role": "system", "content": system_prompt},
- {"role": "user", "content": text[:2000]} # 截断防止超长
- ],
- "max_tokens": 200,
- "temperature": 0.0,
- "response_format": {"type": "json_object"}
- }
- resp = requests.post(url, headers=headers, json=body, timeout=10)
- resp.raise_for_status()
- raw_text = resp.json()["choices"][0]["message"]["content"]
- parsed = _extract_json(raw_text)
- if parsed is None:
- return False, "LLM 返回无法解析"
- return bool(parsed.get("is_plan", False)), str(parsed.get("reason", ""))
- except Exception as exc:
- return False, f"LLM 调用异常: {exc}"
- def validate_training_plan(
- text: str, *, allow_external_models: bool = True
- ) -> dict[str, Any]:
- """主入口
- 返回: {"is_plan": bool, "reason": str, "stage": str}
- """
- res1 = _level1_keyword_check(text)
- if res1 is not None:
- return {"is_plan": res1[0], "reason": res1[1], "stage": "level1_keywords"}
- if not allow_external_models:
- return {
- "is_plan": False,
- "reason": "已关闭外部模型,无法对模糊的计划内容进行语义鉴定",
- "stage": "external_models_disabled",
- }
- res2 = _level2_llm_check(text)
- return {"is_plan": res2[0], "reason": res2[1], "stage": "level2_llm"}
|