Upload 13 files

detectors/ambiguity.py

@@ -2,7 +2,7 @@
 import re
 from ..utils.nlp_helpers import tokenize
 
-def evaluate(prompt: str) -> dict:
+def detect_ambiguity(prompt: str) -> dict:
     # 简单检测多义词列表
     ambiguous = [w for w in ["或者","可能","大概"] if w in prompt]
     score = 1.0 - len(ambiguous)*0.2

detectors/clarity.py

@@ -1,7 +1,7 @@
 # detectors/clarity.py
 from ..utils.nlp_helpers import sentence_length
 
-def evaluate(prompt: str) -> dict:
+def detect_clarity(prompt: str) -> dict:
     length = sentence_length(prompt)
     score = 1.0 if length < 50 else max(0.0, 1.0 - (length - 50)/100)
     suggestions = []

detectors/context_completeness.py

@@ -1,6 +1,6 @@
 # detectors/context_completeness.py
 
-def evaluate(prompt: str) -> dict:
+def detect_context_completeness(prompt: str) -> dict:
     # 简易：检测是否包含关键词示例或上下文标签
     has_context = '背景' in prompt or '示例' in prompt
     score = 1.0 if has_context else 0.5

detectors/diversity.py

@@ -2,7 +2,7 @@
 from sklearn.feature_extraction.text import TfidfVectorizer
 import numpy as np
 
-def evaluate(batch_prompts: list) -> dict:
+def detect_diversity(batch_prompts: list) -> dict:
     vec = TfidfVectorizer().fit_transform(batch_prompts)
     sim = (vec * vec.T).A
     avg_sim = np.mean(sim[np.triu_indices_from(sim, k=1)])

detectors/ethic_compliance.py

@@ -1,6 +1,6 @@
 # detectors/ethic_compliance.py
 
-def evaluate(prompt: str) -> dict:
+def detect_ethic_compliance(prompt: str) -> dict:
     # 简易词库检测
     blacklist = ['暴力','歧视','仇恨']
     hits = [w for w in blacklist if w in prompt]

detectors/feasibility.py

@@ -1,7 +1,7 @@
 # detectors/feasibility.py
 from ..utils.nlp_helpers import sentence_length
 
-def evaluate(prompt: str, max_tokens: int = 512) -> dict:
+def detect_feasibility(prompt: str, max_tokens: int = 512) -> dict:
     length = sentence_length(prompt)
     score = 1.0 if length < max_tokens/2 else 0.5
     suggestions = []

detectors/grammar_spelling.py

@@ -1,10 +1,34 @@
-# detectors/grammar_spelling.py
-from language_tool_python import LanguageTool
+"""
+Grammar and spelling detector for PRIVAL.
+"""
 
-tool = LanguageTool('en-US')
+try:
+    from language_tool_python import LanguageTool
+except Exception:
+    LanguageTool = None
 
-def evaluate(prompt: str) -> dict:
+
+def grammar_spelling(prompt: str):
+    """
+    使用 LanguageTool 检测语法和拼写错误，如果不可用则跳过。
+    """
+    # 如果 LanguageTool 不可用，跳过此检测
+    if LanguageTool is None:
+        return {"score": None, "suggestions": ["LanguageTool not available; skipped grammar checks."]}
+
+    # 尝试初始化 LanguageTool 服务
+    try:
+        tool = LanguageTool('en-US')
+    except Exception:
+        return {"score": None, "suggestions": ["Failed to start LanguageTool; skipped grammar checks."]}
+
+    # 执行检查
     matches = tool.check(prompt)
-    score = 1.0 if not matches else max(0.0, 1.0 - len(matches)*0.1)
-    suggestions = [m.message for m in matches]
-    return {"score": round(score,2), "suggestions": suggestions}
+    suggestions = []
+    for match in matches:
+        suggestions.append(f"{match.ruleId}: {match.message} at position {match.offset}")
+
+    # 计算分数：错误越少得分越高
+    length = len(prompt.split()) if prompt else 1
+    score = max(0.0, 1.0 - len(matches) / length)
+    return {"score": score, "suggestions": suggestions}

detectors/injection_risk.py

@@ -1,7 +1,7 @@
 # detectors/injection_risk.py
 import re
 
-def evaluate(prompt: str) -> dict:
+def detect_injection_risk(prompt: str) -> dict:
     patterns = [r"\bignore previous\b", r"\bmalicious\b"]
     hits = [p for p in patterns if re.search(p, prompt, re.IGNORECASE)]
     score = 1.0 - len(hits)*0.5

detectors/length_appropriateness.py

@@ -1,7 +1,7 @@
 # detectors/length_appropriateness.py
 from ..utils.nlp_helpers import sentence_length
 
-def evaluate(prompt: str, min_len: int = 10, max_len: int = 200) -> dict:
+def detect_length_appropriateness(prompt: str, min_len: int = 10, max_len: int = 200) -> dict:
     length = sentence_length(prompt)
     score = 1.0 if min_len <= length <= max_len else 0.5
     suggestions = []

detectors/relevance.py

@@ -3,7 +3,7 @@ from sentence_transformers import SentenceTransformer, util
 
 model = SentenceTransformer('all-MiniLM-L6-v2')
 
-def evaluate(prompt: str, reference: str = None) -> dict:
+def detect_relevance(prompt: str, reference: str = None) -> dict:
     if reference:
         sim = util.cos_sim(model.encode(prompt), model.encode(reference)).item()
     else:

detectors/step_guidance.py

@@ -1,17 +1,9 @@
 # detectors/step_guidance.py
 from ..utils.nlp_helpers import tokenize
 
-def evaluate(prompt: str) -> dict:
+def detect_step_guidance(prompt: str) -> dict:
     tokens = tokenize(prompt)
     has_step = any(w in ["步骤","首先","然后","最后"] for w in tokens)
     score = 1.0 if has_step else 0.0
     suggestions = [] if has_step else ["建议在 prompt 中添加明确步骤提示，如'首先...'、'然后...'" ]
-    return {"score": score, "suggestions": suggestions}# detectors/step_guidance.py
-from ..utils.nlp_helpers import tokenize
-
-def evaluate(prompt: str) -> dict:
-    tokens = tokenize(prompt)
-    has_step = any(w in ["步骤","首先","然后","最后"] for w in tokens)
-    score = 1.0 if has_step else 0.0
-    suggestions = [] if has_step else ["建议在 prompt 中添加明确步骤提示，如'首先...'、'然后...'" ]
-    return {"score": score, "suggestions": suggestions}
+    return {"score": score, "suggestions": suggestions}# detectors/step_guidance.py

detectors/structural_cleanness.py

@@ -1,10 +1,27 @@
-# detectors/structural_cleanness.py
-from ..utils.nlp_helpers import dependency_depth
-
-def evaluate(prompt: str) -> dict:
-    depth = dependency_depth(prompt)
-    score = 1.0 if depth < 3 else max(0.0, 1.0 - (depth-3)*0.2)
-    suggestions = []
-    if depth > 5:
-        suggestions.append("句子结构过于复杂，建议拆分或简化嵌套。")
-    return {"score": round(score,2), "suggestions": suggestions}
+# prival/prival/detectors/structural_cleanness.py
+
+# 1. 捕获 spacy 不存在的情况
+try:
+    import spacy
+except ImportError:
+    spacy = None
+
+from ..utils.nlp_helpers import dependency_depth, sentence_length
+# …其它 imports…
+
+def structural_cleanness(prompt: str):
+    # 2. 如果没有安装 spaCy，就跳过
+    if spacy is None:
+        return {
+            "score": None,
+            "suggestions": ["spaCy not installed; structural_cleanness skipped."]
+        }
+
+    # 原有的 spaCy 分析逻辑，例如：
+    # nlp = spacy.load("en_core_web_sm")
+    # doc = nlp(prompt)
+    # depth = dependency_depth(doc)
+    # …计算分数、生成 suggestions …
+
+    # 最终返回一个 dict
+    return {"score": computed_score, "suggestions": suggestions}

detectors/verbosity.py

@@ -1,7 +1,7 @@
 # detectors/verbosity.py
 from ..utils.nlp_helpers import sentence_length
 
-def evaluate(prompt: str) -> dict:
+def detect_verbosity(prompt: str) -> dict:
     length = sentence_length(prompt)
     score = 1.0 if length < 60 else max(0.0, 1.0 - (length-60)/200)
     suggestions = []