dingo-python 2.2.1__py3-none-any.whl → 2.2.2__py3-none-any.whl

dingo/config/input_args.py

@@ -102,10 +102,11 @@ class EmbeddingConfigArgs(BaseModel):
 
 
 class EvaluatorLLMArgs(BaseModel):
+    model_config = {"extra": "allow"}
+
     model: Optional[str] = None
     key: Optional[str] = None
     api_url: Optional[str] = None
-    parameters: Optional[dict] = None
     embedding_config: Optional[EmbeddingConfigArgs] = None
 
 

dingo/model/llm/agent/agent_article_fact_checker.py

@@ -343,21 +343,21 @@ class ArticleFactChecker(BaseAgent):
             "config": {
                 "key": "your-openai-api-key",
                 "model": "gpt-4o-mini",
-                "parameters": {
-                    "agent_config": {
-                        "max_iterations": 10,
-                        "tools": {
-                            "claims_extractor": {
-                                "api_key": "your-openai-api-key",
-                                "max_claims": 50,
-                                "claim_types": ["factual", "institutional", "statistical", "attribution"]
-                            },
-                            "tavily_search": {
-                                "api_key": "your-tavily-api-key",
-                                "max_results": 5
-                            },
-                            "arxiv_search": {"max_results": 5}
-                        }
+                "agent_config": {
+                    "max_iterations": 10,
+                    "overall_timeout": 900,
+                    "max_concurrent_claims": 5,
+                    "tools": {
+                        "claims_extractor": {
+                            "api_key": "your-openai-api-key",
+                            "max_claims": 50,
+                            "claim_types": ["factual", "institutional", "statistical", "attribution"]
+                        },
+                        "tavily_search": {
+                            "api_key": "your-tavily-api-key",
+                            "max_results": 5
+                        },
+                        "arxiv_search": {"max_results": 5}
                     }
                 }
             }
@@ -372,6 +372,9 @@ class ArticleFactChecker(BaseAgent):
     ]
     max_iterations = 10  # Allow more iterations for comprehensive checking
     max_concurrent_claims = 5  # Default parallel claim verification slots
+    overall_timeout = 900       # 15-minute wall-clock timeout for entire evaluation
+    _MIN_OVERALL_TIMEOUT = 30   # Floor: 30 seconds
+    _MAX_OVERALL_TIMEOUT = 7200  # Ceiling: 2 hours
 
     _required_fields = [RequiredField.CONTENT]  # Article text
 
@@ -394,8 +397,8 @@ class ArticleFactChecker(BaseAgent):
         Returns:
             Output directory path (created if needed), or None if saving is disabled.
         """
-        params = cls.dynamic_config.parameters or {}
-        agent_cfg = params.get('agent_config') or {}
+        extra_params = cls.dynamic_config.model_extra
+        agent_cfg = extra_params.get('agent_config') or {}
 
         explicit_path = agent_cfg.get('output_path')
         if explicit_path:
@@ -816,24 +819,42 @@ class ArticleFactChecker(BaseAgent):
         output_dir = cls._get_output_dir()
 
         if cls.dynamic_config:
-            if cls.dynamic_config.parameters is None:
-                cls.dynamic_config.parameters = {}
-            cls.dynamic_config.parameters.setdefault("temperature", 0)
+            if 'temperature' not in cls.dynamic_config.model_extra:
+                cls.dynamic_config.temperature = 0
 
         if output_dir and input_data.content:
             cls._save_article_content(output_dir, input_data.content)
 
+        timeout = cls._get_overall_timeout()
+
+        async def _run_with_timeout() -> EvalDetail:
+            return await asyncio.wait_for(
+                cls._async_eval(input_data, start_time, output_dir),
+                timeout=timeout,
+            )
+
         try:
-            return asyncio.run(cls._async_eval(input_data, start_time, output_dir))
+            return asyncio.run(_run_with_timeout())
+        except asyncio.TimeoutError:
+            elapsed = time.time() - start_time
+            log.warning(f"ArticleFactChecker: overall timeout exceeded ({elapsed:.1f}s / {timeout:.0f}s limit)")
+            return cls._create_overall_timeout_result(elapsed, timeout)
         except RuntimeError as e:
             # Fallback when called inside an already-running event loop (e.g. Jupyter, tests)
             if "cannot run" in str(e).lower() or "already running" in str(e).lower():
                 import concurrent.futures
                 with concurrent.futures.ThreadPoolExecutor(max_workers=1) as pool:
-                    future = pool.submit(
-                        lambda: asyncio.run(cls._async_eval(input_data, start_time, output_dir))
-                    )
-                    return future.result()
+                    future = pool.submit(lambda: asyncio.run(_run_with_timeout()))
+                    try:
+                        # Extra margin so asyncio.wait_for fires before this outer timeout
+                        return future.result(timeout=timeout + 30)
+                    except (asyncio.TimeoutError, concurrent.futures.TimeoutError):
+                        elapsed = time.time() - start_time
+                        log.warning(
+                            f"ArticleFactChecker: overall timeout exceeded "
+                            f"({elapsed:.1f}s / {timeout:.0f}s limit, fallback path)"
+                        )
+                        return cls._create_overall_timeout_result(elapsed, timeout)
             raise
 
     # --- Two-Phase Async Architecture Methods ---
@@ -922,8 +943,8 @@ class ArticleFactChecker(BaseAgent):
         """
         from dingo.model.llm.agent.tools.claims_extractor import ClaimsExtractor, ClaimsExtractorConfig
 
-        params = cls.dynamic_config.parameters or {}
-        agent_cfg = params.get('agent_config') or {}
+        extra_params = cls.dynamic_config.model_extra
+        agent_cfg = extra_params.get('agent_config') or {}
         extractor_cfg = agent_cfg.get('tools', {}).get('claims_extractor', {})
 
         config_kwargs: Dict[str, Any] = {
@@ -1019,10 +1040,30 @@ class ArticleFactChecker(BaseAgent):
     @classmethod
     def _get_max_concurrent_claims(cls) -> int:
         """Read max_concurrent_claims from agent_config or use class default."""
-        params = cls.dynamic_config.parameters or {}
-        agent_cfg = params.get('agent_config') or {}
+        extra_params = cls.dynamic_config.model_extra
+        agent_cfg = extra_params.get('agent_config') or {}
         return agent_cfg.get('max_concurrent_claims', cls.max_concurrent_claims)
 
+    @classmethod
+    def _get_overall_timeout(cls) -> float:
+        """Read overall_timeout from agent_config or use class default (900s).
+
+        Returns:
+            Positive timeout in seconds, clamped to [30, 7200].
+        """
+        extra_params = cls.dynamic_config.model_extra
+        agent_cfg = extra_params.get('agent_config') or {}
+        raw = agent_cfg.get('overall_timeout', cls.overall_timeout)
+        try:
+            timeout = float(raw)
+        except (TypeError, ValueError):
+            log.warning(f"Invalid overall_timeout={raw!r}, using default {cls.overall_timeout}s")
+            return float(cls.overall_timeout)
+        clamped = max(cls._MIN_OVERALL_TIMEOUT, min(timeout, cls._MAX_OVERALL_TIMEOUT))
+        if clamped != timeout:
+            log.warning(f"overall_timeout={timeout} out of range, clamped to {clamped}s")
+        return float(clamped)
+
     @classmethod
     def _parse_claim_json_robust(cls, output: Optional[str]) -> Dict[str, Any]:
         """
@@ -1795,6 +1836,38 @@ Begin your systematic fact-checking process now.
         ]
         return result
 
+    @classmethod
+    def _create_overall_timeout_result(cls, elapsed: float, timeout: float) -> EvalDetail:
+        """
+        Create error result when overall wall-clock timeout is exceeded.
+
+        Args:
+            elapsed: Actual elapsed time in seconds
+            timeout: Configured timeout limit in seconds
+
+        Returns:
+            EvalDetail with timeout error status
+        """
+        minutes, seconds = divmod(int(timeout), 60)
+        limit_str = f"{minutes}m{seconds}s" if minutes else f"{int(timeout)}s"
+        result = EvalDetail(metric=cls.__name__)
+        result.status = True
+        result.label = [f"{QualityLabel.QUALITY_BAD_PREFIX}AGENT_OVERALL_TIMEOUT"]
+        result.reason = [
+            "Article Fact-Checking Failed: Overall Timeout Exceeded",
+            "=" * 70,
+            f"Execution exceeded the {int(timeout)}s ({limit_str}) wall-clock limit.",
+            f"Elapsed time: {elapsed:.1f}s",
+            "",
+            "Recommendations:",
+            f"  1. Increase overall_timeout (current: {int(timeout)}s) in agent_config",
+            "  2. Reduce max_claims in claims_extractor config (e.g., 50 -> 20)",
+            "  3. Use a faster model (e.g., gpt-4o-mini instead of gpt-4o)",
+            "  4. Reduce max_concurrent_claims to lower API rate-limit pressure",
+            "  5. Split long articles into shorter sections",
+        ]
+        return result
+
     @classmethod
     def plan_execution(cls, input_data: Data) -> List[Dict[str, Any]]:
         """

dingo/model/llm/agent/agent_fact_check.py

@@ -70,15 +70,13 @@ class AgentFactCheck(BaseAgent):
             "key": "your-openai-api-key",
             "api_url": "https://api.openai.com/v1",
             "model": "gpt-4.1-mini-2025-04-14",
-            "parameters": {
-                "agent_config": {
-                    "max_iterations": 5,
-                    "tools": {
-                        "tavily_search": {
-                            "api_key": "your-tavily-api-key",
-                            "max_results": 5,
-                            "search_depth": "advanced"
-                        }
+            "agent_config": {
+                "max_iterations": 5,
+                "tools": {
+                    "tavily_search": {
+                        "api_key": "your-tavily-api-key",
+                        "max_results": 5,
+                        "search_depth": "advanced"
                     }
                 }
             }

dingo/model/llm/agent/agent_hallucination.py

@@ -82,15 +82,13 @@ class AgentHallucination(BaseAgent):
             "key": "your-openai-api-key",
             "api_url": "https://api.openai.com/v1",
             "model": "gpt-4.1-mini-2025-04-14",
-            "parameters": {
-                "agent_config": {
-                    "max_iterations": 3,
-                    "tools": {
-                        "tavily_search": {
-                            "api_key": "your-tavily-api-key",
-                            "max_results": 5,
-                            "search_depth": "advanced"
-                        }
+            "agent_config": {
+                "max_iterations": 3,
+                "tools": {
+                    "tavily_search": {
+                        "api_key": "your-tavily-api-key",
+                        "max_results": 5,
+                        "search_depth": "advanced"
                     }
                 }
             }

dingo/model/llm/agent/agent_wrapper.py

@@ -327,22 +327,22 @@ class AgentWrapper:
             )
 
         # Extract parameters
-        params = dynamic_config.parameters or {}
+        extra_params = dynamic_config.model_extra
 
         # Create ChatOpenAI instance
         llm = ChatOpenAI(
             api_key=dynamic_config.key,
             base_url=dynamic_config.api_url,
             model=dynamic_config.model or "gpt-4.1-mini",
-            temperature=params.get("temperature", 0.3),
-            max_tokens=params.get("max_tokens", 4096),
-            top_p=params.get("top_p", 1.0),
-            timeout=params.get("timeout", 30)
+            temperature=extra_params.get("temperature", 0.3),
+            max_tokens=extra_params.get("max_tokens", 4096),
+            top_p=extra_params.get("top_p", 1.0),
+            timeout=extra_params.get("timeout", 30)
         )
 
         log.debug(
             f"Created ChatOpenAI: model={dynamic_config.model}, "
-            f"temp={params.get('temperature', 0.3)}"
+            f"temp={extra_params.get('temperature', 0.3)}"
         )
 
         return llm

dingo/model/llm/agent/base_agent.py

@@ -146,7 +146,7 @@ class BaseAgent(BaseOpenAI):
         Extract tool configuration from agent's dynamic_config.
 
         Configuration is expected in:
-        dynamic_config.parameters.agent_config.tools.{tool_name}
+        dynamic_config.agent_config.tools.{tool_name}
 
         Args:
             tool_name: Name of the tool
@@ -154,8 +154,8 @@ class BaseAgent(BaseOpenAI):
         Returns:
             Dict of configuration values for the tool
         """
-        params = cls.dynamic_config.parameters or {}
-        agent_config = params.get('agent_config', {})
+        extra_params = cls.dynamic_config.model_extra
+        agent_config = extra_params.get('agent_config', {})
         tools_config = agent_config.get('tools', {})
         return tools_config.get(tool_name, {})
 
@@ -184,8 +184,8 @@ class BaseAgent(BaseOpenAI):
         Returns:
             Maximum number of iterations allowed
         """
-        params = cls.dynamic_config.parameters or {}
-        agent_config = params.get('agent_config', {})
+        extra_params = cls.dynamic_config.model_extra
+        agent_config = extra_params.get('agent_config', {})
         return agent_config.get('max_iterations', cls.max_iterations)
 
     @classmethod

dingo/model/llm/base_openai.py

@@ -82,22 +82,18 @@ class BaseOpenAI(BaseLLM):
         else:
             model_name = cls.client.models.list().data[0].id
 
-        params = cls.dynamic_config.parameters
-        cls.validate_config(params)
+        extra_params = cls.dynamic_config.model_extra
+        cls.validate_config(extra_params)
 
         completions = cls.client.chat.completions.create(
             model=model_name,
             messages=messages,
-            temperature=params.get("temperature", 0.3) if params else 0.3,
-            top_p=params.get("top_p", 1) if params else 1,
-            max_tokens=params.get("max_tokens", 4000) if params else 4000,
-            presence_penalty=params.get("presence_penalty", 0) if params else 0,
-            frequency_penalty=params.get("frequency_penalty", 0) if params else 0,
+            **extra_params,
         )
 
         if completions.choices[0].finish_reason == "length":
             raise ExceedMaxTokens(
-                f"Exceed max tokens: {params.get('max_tokens', 4000) if params else 4000}"
+                f"Exceed max tokens: {extra_params.get('max_tokens', 4000)}"
             )
 
         return str(completions.choices[0].message.content)

dingo/model/llm/compare/llm_html_extract_compare_v2.py

@@ -25,9 +25,17 @@ class LLMHtmlExtractCompareV2(BaseOpenAI):
     输入数据要求：
     - input_data.prompt: 工具A提取的文本
     - input_data.content: 工具B提取的文本
-    - input_data.raw_data.get("language", "en"): 语言类型 ("zh" 或 "en")
+    - language: 可选，来自 input_data.language 或 raw_data["language"]，缺省为 "en"（"zh" / "en"）
     """
 
+    _metric_info = {
+        'category': 'Pretrain Text Quality Assessment Metrics',
+        'metric_name': 'LLMHtmlExtractCompareV2',
+        'description': 'Compares two HTML main-content extraction tools by computing text diffs and using LLM to judge which preserves more core information',
+        'paper_title': '',
+        'paper_url': '',
+    }
+
     _required_fields = [RequiredField.CONTENT, RequiredField.PROMPT]
     prompt = {
         "content_en": r"""Please compare the following two texts, each extracted from the same webpage using different HTML parsing methods. Your task is to determine whether there is a difference in the core informational content between them.
@@ -174,7 +182,8 @@ C. Text A 包含的核心信息内容少于 Text B
         text_tool_b = input_data.content
 
         # 获取配置参数
-        language = input_data.raw_data.get("language", "en")
+        raw_data = getattr(input_data, 'raw_data', {}) or {}
+        language = raw_data.get("language", getattr(input_data, 'language', "en"))
 
         # 计算文本差异
         diff_result = cls.extract_text_diff(text_tool_a, text_tool_b)

dingo/model/llm/instruction_quality/llm_instruction_clarity.py

@@ -283,8 +283,8 @@ Output:
 
             # 判断是否通过（默认阈值 6.0）
             threshold = 6.0
-            if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-                threshold = cls.dynamic_config.parameters.get('threshold', 6.0)
+            if hasattr(cls, 'dynamic_config'):
+                threshold = cls.dynamic_config.model_extra.get('threshold', 6.0)
 
             if score >= threshold:
                 result.status = False

dingo/model/llm/instruction_quality/llm_task_difficulty.py

@@ -321,14 +321,14 @@ Output:
 
             # 难度评估没有"通过/不通过"的概念，只是描述性的
             # 但为了兼容框架，我们设置一个合理的默认行为
-            # 可以通过 parameters 配置 min_difficulty 和 max_difficulty
+            # 可以通过 config 中的 min_difficulty 和 max_difficulty 配置难度范围
             result.status = False  # 默认不标记为问题
             result.label = [f"TASK_DIFFICULTY.{difficulty_level.upper()}"]
 
             # 如果配置了难度范围要求，进行检查
-            if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-                min_difficulty = cls.dynamic_config.parameters.get('min_difficulty', 0)
-                max_difficulty = cls.dynamic_config.parameters.get('max_difficulty', 10)
+            if hasattr(cls, 'dynamic_config'):
+                min_difficulty = cls.dynamic_config.model_extra.get('min_difficulty', 0)
+                max_difficulty = cls.dynamic_config.model_extra.get('max_difficulty', 10)
 
                 if difficulty_score < min_difficulty:
                     result.status = True

dingo/model/llm/rag/llm_rag_answer_relevancy.py

@@ -242,14 +242,8 @@ class LLMRAGAnswerRelevancy(BaseOpenAI):
 
         try:
             # 增加温度参数以提高问题生成的随机性
-            if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-                if 'temperature' not in cls.dynamic_config.parameters:
-                    cls.dynamic_config.parameters['temperature'] = 0.7
-            else:
-                # 如果没有parameters，创建一个包含temperature的parameters
-                current_params = cls.dynamic_config.parameters or {}
-                current_params['temperature'] = 0.7
-                cls.dynamic_config.parameters = current_params
+            if hasattr(cls, 'dynamic_config') and 'temperature' not in cls.dynamic_config.model_extra:
+                cls.dynamic_config.temperature = 0.7
 
             # 生成多个相关问题
             generated_questions = cls.generate_multiple_questions(input_data, cls.strictness)
@@ -263,10 +257,9 @@ class LLMRAGAnswerRelevancy(BaseOpenAI):
 
             # 根据分数判断是否通过，默认阈值为5
             threshold = 5
-            if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-                threshold = cls.dynamic_config.parameters.get('threshold', 5)
-                # 检查是否有自定义的strictness参数
-                cls.strictness = cls.dynamic_config.parameters.get('strictness', 3)
+            if hasattr(cls, 'dynamic_config'):
+                threshold = cls.dynamic_config.model_extra.get('threshold', 5)
+                cls.strictness = cls.dynamic_config.model_extra.get('strictness', 3)
 
             # 构建详细的reason文本
             all_reasons = []

dingo/model/llm/rag/llm_rag_context_precision.py

@@ -256,8 +256,8 @@ class LLMRAGContextPrecision(BaseOpenAI):
 
         # 根据分数判断是否通过，默认阈值为5
         threshold = 5
-        if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-            threshold = cls.dynamic_config.parameters.get('threshold', 5)
+        if hasattr(cls, 'dynamic_config'):
+            threshold = cls.dynamic_config.model_extra.get('threshold', 5)
 
         if score >= threshold:
             result.status = False

dingo/model/llm/rag/llm_rag_context_recall.py

@@ -215,8 +215,8 @@ class LLMRAGContextRecall(BaseOpenAI):
 
         # 根据分数判断是否通过，默认阈值为5
         threshold = 5
-        if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-            threshold = cls.dynamic_config.parameters.get('threshold', 5)
+        if hasattr(cls, 'dynamic_config'):
+            threshold = cls.dynamic_config.model_extra.get('threshold', 5)
 
         if score >= threshold:
             result.status = False

dingo/model/llm/rag/llm_rag_context_relevancy.py

@@ -206,8 +206,8 @@ class LLMRAGContextRelevancy(BaseOpenAI):
 
         # 根据分数判断是否通过，默认阈值为5
         threshold = 5
-        if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-            threshold = cls.dynamic_config.parameters.get('threshold', 5)
+        if hasattr(cls, 'dynamic_config'):
+            threshold = cls.dynamic_config.model_extra.get('threshold', 5)
 
         if score >= threshold:
             result.status = False

dingo/model/llm/rag/llm_rag_faithfulness.py

@@ -290,8 +290,8 @@ class LLMRAGFaithfulness(BaseOpenAI):
 
         # 根据分数判断是否通过，默认阈值为5
         threshold = 5
-        if hasattr(cls, 'dynamic_config') and cls.dynamic_config.parameters:
-            threshold = cls.dynamic_config.parameters.get('threshold', 5)
+        if hasattr(cls, 'dynamic_config'):
+            threshold = cls.dynamic_config.model_extra.get('threshold', 5)
 
         if score >= threshold:
             result.status = False

dingo/model/llm/text_quality/base_text_quality.py

@@ -47,16 +47,11 @@ class BaseTextQuality(BaseOpenAI):
         response_json = json.loads(response)
         response_model = ResponseScoreTypeNameReason(**response_json)
 
-        # Create EvalDetail with all required fields
-        # status = False for Good quality (no issues found)
-        # status = True for Bad quality (issues found)
-        is_good = response_model.type == "Good"
-
         result = EvalDetail(
             metric=cls.__name__,
-            status=not is_good,  # True if Bad (issues found), False if Good
+            status=False if response_model.score == 1 else True,
             score=response_model.score,
-            label=["QUALITY_GOOD"] if is_good else [f"{response_model.type}.{response_model.name}"],
+            label=["QUALITY_GOOD"] if response_model.score == 1 else [f"{response_model.type}.{response_model.name}"],
             reason=[response_model.reason]
         )
 

dingo/model/llm/text_quality/llm_text_equation.py

@@ -0,0 +1,68 @@
+from dingo.io.input import RequiredField
+from dingo.model import Model
+from dingo.model.llm.text_quality.base_text_quality import BaseTextQuality
+
+
+@Model.llm_register("LLMTextEquation")
+class LLMTextEquation(BaseTextQuality):
+    # Metadata for documentation generation
+    _metric_info = {
+        "category": "Pretrain Text Quality Assessment Metrics",
+        "metric_name": "LLMTextQualityV5",
+        "description": "Impact-driven text quality evaluation for LLM pretraining, focusing on structural completeness, readability, diversity, and safety with quantitative thresholds",
+        "paper_title": "WanJuanSiLu: A High-Quality Open-Source Webtext Dataset for Low-Resource Languages",
+        "paper_url": "https://arxiv.org/abs/2501.14506",
+        "paper_authors": "Yu et al., 2025",
+        "examples": "examples/llm_and_rule/llm_local.py",
+        "evaluation_results": "docs/eval/prompt/redpajama_data_evaluated_by_prompt.md"
+    }
+    _required_fields = [RequiredField.CONTENT]
+    prompt = r"""
+你是一个专业的数学、化学等学科的公式质检员。我会给你一个从文档中提取的 equation 类型元素（JSON 格式），请对其 text 字段进行质量检测。
+
+## 检测维度
+
+1. **语法问题**
+   - LaTeX 命令拼写错误（如 \frace 代替 \frac）
+   - 括号未正确配对闭合（{}、[]、()）
+   - 环境标签不匹配（如 \begin{} 与 \end{} 不对应）
+
+2. **识别问题**
+   - 疑似 OCR 识别错误（如字母与符号混淆：x 与 ×、- 与 −、l 与 1、O 与 0 等）
+   - 公式内容明显残缺或截断
+   - 出现乱码或无意义字符
+
+3. **语义问题**
+   - 公式结构不完整，无法表达完整的数学含义
+   - 运算符或符号使用明显不合数学规范
+
+## 一级错误类型（type）
+
+- `syntax`：语法问题
+- `recognition`：识别问题
+- `semantic`：语义问题
+
+## 二级错误类型（name）
+
+- `command_error`：LaTeX 命令拼写错误
+- `bracket_mismatch`：括号未正确配对
+- `env_mismatch`：环境标签不匹配
+- `ocr_error`：OCR 字符识别错误
+- `truncated_content`：公式残缺或截断
+- `garbled_text`：乱码或无意义字符
+- `incomplete_expression`：公式结构不完整
+- `invalid_notation`：数学符号使用不规范
+- `none`：无问题
+
+## Output Format
+
+Return JSON only: {"score": 0/1, "type": "", "name": "", "reason": ""}
+
+score 类型必须为int；
+score 为 1 表示通过，type 填 "Good"，name 填 "None"，reason 说明公式正常的依据；
+score 为 0 表示不通过，type 和 name 填对应的错误类型，reason 说明判断依据并指出具体的问题位置或内容。
+
+## Input content to evaluate:
+
+"""
+    # process_response method is now inherited from BaseTextQuality

dingo/model/llm/text_quality/llm_text_quality_v5.py

@@ -30,7 +30,27 @@ Evaluate whether this text is suitable for LLM pretraining. Focus on issues that
 **Impact**: Broken structures prevent models from learning correct formatting patterns.
 
 **Check for**:
-- **Error_Formula**: Mathematical expressions with **unmatched delimiters** or **unclosed environments**
+- **Error_Formula**: Mathematical content with **broken syntax** OR **systematically stripped symbols/formulas**
+
+  Two failure modes:
+
+  **(A) Broken LaTeX syntax** — delimiters or environments are present but malformed:
+  - Delimiters unmatched: $ without closing $ (LaTeX context, not dollar signs)
+  - Environments unclosed: \\begin{{align}} without \\end{{align}}
+  - Syntax broken: \\frac{{a}}{{b missing closing }}
+  - HTML tags unclosed: <sub>text without </sub>
+  - Impact: Prevents >50% of mainstream parsers from rendering
+
+  **(B) Stripped mathematical content** — symbols/formulas systematically removed during extraction:
+  - Orphan hyphens from stripped Greek letters: "κ-solutions" → "-solutions", "ε-net" → "-net"
+  - Empty positions after connective words: "thus ;" or "the interval ;" where a formula was removed
+  - Sentences referencing variables/expressions that are absent: "a small number" (number missing), "we have ." (equation missing)
+  - Systematic loss: multiple occurrences throughout the text, not just one or two typos
+  - Impact: Mathematical text becomes incoherent; models learn broken academic writing patterns
+
+  Example (BAD — stripped symbols):
+  "Let be a -solution to the Ricci flow which is -noncollapsed. Ancient, in the sense that t ranges on the interval ; Bounded curvature, thus ;"
+  (Greek letters κ stripped from "κ-solution" and "κ-noncollapsed"; interval expression and inequality after "thus" removed entirely)
 
   ⚠️ **Normal patterns (DO NOT flag)**:
   - Mixing inline ($...$) and display ($$...$$) formulas
@@ -38,31 +58,39 @@ Evaluate whether this text is suitable for LLM pretraining. Focus on issues that
   - Line breaks with \\\\ in alignment environments
   - HTML tags: <sub>x</sub>, <sup>2</sup> for subscripts/superscripts
   - Mixing LaTeX and HTML in web-extracted content
-
-  ✅ **Only flag when**:
-  - Delimiters unmatched: $ without closing $ (LaTeX context, not dollar signs)
-  - Environments unclosed: \\begin{{align}} without \\end{{align}}
-  - Syntax broken: \\frac{{a}}{{b missing closing }}
-  - HTML tags unclosed: <sub>text without </sub>
+  - Plain-text math without any LaTeX (e.g., "a^2 + b^2 = c^2" without $ delimiters) — this is fine as long as the expressions are actually present
 
   ⚠️ **Important**: Distinguish LaTeX $ from dollar signs ($100)
   - Dollar sign: "$100", "$5.99" (followed by numbers) → NOT LaTeX
   - LaTeX delimiter: "$x$", "$\\alpha$" (contains math symbols) → IS LaTeX
-  - Example: "The price is $100 and equation $x=y$ costs $50" has 4 dollar symbols but only 2 are LaTeX delimiters (and they match)
 
-  - Example (BAD): "$x^2 + y^2 is broken here $$a = b$$$"
+  - Example (BAD — broken delimiters): "$x^2 + y^2 is broken here $$a = b$$$"
     (First LaTeX $ never closes, extra $ at end)
   - Example (GOOD): "The item costs $100 and satisfies $x^2 + y^2 = z^2$ where price is $50"
     (Dollar signs for money + proper LaTeX pair)
-  - Impact: Only flag errors that prevent >50% of mainstream parsers (pdflatex, MathJax, KaTeX, Pandoc, Jupyter) from rendering
 
 - **Error_Table**: Table structures that are malformed or unreadable
   - Example (BAD): Misaligned columns, missing headers, or garbled HTML tags
   - Impact: Models cannot learn proper table representation
 
 - **Error_Code**: Code blocks with formatting corruption
-  - Example (BAD): Line numbers mixed with code, broken syntax highlighting markers
-  - Impact: Teaches incorrect code structure
+  **Common corruption patterns**:
+  - Missing code fence (` ``` `): code appears as plain text without language block
+  - Lost indentation: Python/YAML code with all indentation stripped (flat lines)
+  - Broken identifiers: spaces injected into tokens, e.g. `sys .argv`, `pts .append`, `i[ 0]`
+  - Line numbers mixed with code, broken syntax highlighting markers
+  - Keywords wrapped in inline backticks instead of a fenced block, e.g. `` `import` sys ``
+
+  Example (BAD — indentation and identifiers destroyed):
+  ```
+  `import` sys
+  pts = []
+  for i in range( 1,len(sys .argv), 2):
+  pts .append([int(sys .argv[i]), int(sys .argv[i +1])])
+  ```
+  Correct version would have a code fence, proper indentation, and no spaces inside `sys.argv`.
+
+  - Impact: Teaches incorrect code syntax, broken tokenization patterns, and wrong indentation conventions
 
 **Key Question**: "Can the model learn proper formatting from this structure?"
 
@@ -160,10 +188,14 @@ Output: {{"score": 1, "type": "Good", "name": "None", "reason": "Well-formed mul
 Input: "The eigenstate $\\psi_n$ where <sub>n</sub> is quantum number and energy E<sup>2</sup> = m<sup>2</sup>c<sup>4</sup>"
 Output: {{"score": 1, "type": "Good", "name": "None", "reason": "Normal mix of LaTeX and HTML tags from web content"}}
 
-**Example 2 (Bad - Completeness)**:
+**Example 2 (Bad - Completeness, broken delimiters)**:
 Input: "The formula $x^2 + y^2 is broken here $$a = b$$$"
 Output: {"score": 0, "type": "Completeness", "name": "Error_Formula", "reason": "Unmatched delimiters: first $ never closes, extra $ at end"}
 
+**Example 2.5 (Bad - Completeness, stripped math)**:
+Input: "Definition 1.(-solutions) A -solution is a Ricci flow which is -noncollapsed at every scale. Ancient, in the sense that t ranges on the interval ; Bounded curvature, thus ;"
+Output: {{"score": 0, "type": "Completeness", "name": "Error_Formula", "reason": "Mathematical symbols systematically stripped: Greek letters removed ('-solutions' instead of 'κ-solutions'), formulas missing after 'the interval' and 'thus'"}}
+
 **Example 3 (Bad - Effectiveness)**:
 Input: "Theappleisredandtasty�withsomegarbledtext□□"
 Output: {"score": 0, "type": "Effectiveness", "name": "Error_Garbled_Characters", "reason": "Contains encoding corruption (�, □) and missing spaces (>1% of text)"}

dingo/model/llm/text_quality/llm_text_table.py

@@ -0,0 +1,70 @@
+from dingo.io.input import RequiredField
+from dingo.model import Model
+from dingo.model.llm.text_quality.base_text_quality import BaseTextQuality
+
+
+@Model.llm_register("LLMTextTable")
+class LLMTextTable(BaseTextQuality):
+    # Metadata for documentation generation
+    _metric_info = {
+        "category": "Pretrain Text Quality Assessment Metrics",
+        "metric_name": "LLMTextQualityV5",
+        "description": "Impact-driven text quality evaluation for LLM pretraining, focusing on structural completeness, readability, diversity, and safety with quantitative thresholds",
+        "paper_title": "WanJuanSiLu: A High-Quality Open-Source Webtext Dataset for Low-Resource Languages",
+        "paper_url": "https://arxiv.org/abs/2501.14506",
+        "paper_authors": "Yu et al., 2025",
+        "examples": "examples/llm_and_rule/llm_local.py",
+        "evaluation_results": "docs/eval/prompt/redpajama_data_evaluated_by_prompt.md"
+    }
+    _required_fields = [RequiredField.CONTENT]
+    prompt = r"""
+你是一个专业的表格数据质检员。我会给你一段从文档中提取的 HTML 表格（table_body 字段），请判断该表格是否存在质量问题。
+
+## 检测维度
+
+请从以下维度进行检查：
+
+1. **结构问题**
+   - HTML 标签不完整或嵌套错误（<table>、<tr>、<td> 未正确闭合）
+   - 行列结构异常（某行 <td> 数量与其他行差异过大）
+   - 表格内容全部为空
+
+2. **识别问题**
+   - 存在明显乱码或无意义字符
+   - 疑似 OCR 识别错误（如字母/数字混淆：0与O、1与l、S与5等）
+   - 文字截断或内容残缺
+
+3. **语义问题**
+   - 单元格内容语义不连贯，无法理解表格表达的含义
+   - 行列关系混乱，内容错位
+
+## 一级错误类型（type）
+
+- `structure`：结构问题
+- `recognition`：识别问题
+- `semantic`：语义问题
+
+## 二级错误类型（name）
+
+- `tag_error`：标签不完整或嵌套错误
+- `row_col_mismatch`：行列数量不一致
+- `empty_table`：表格内容为空
+- `garbled_text`：乱码或无意义字符
+- `ocr_error`：OCR 字符识别错误
+- `truncated_content`：文字截断或内容残缺
+- `incoherent_semantics`：语义不连贯
+- `misaligned_content`：内容错位
+- `none`：无问题
+
+## Output Format
+
+Return JSON only: {"score": 0/1, "type": "", "name": "", "reason": ""}
+
+score 类型必须为int；
+score 为 1 表示通过，type 填 "Good"，name 填 "None"，reason 说明公式正常的依据；
+score 为 0 表示不通过，type 和 name 填对应的错误类型，reason 说明判断依据并指出具体位置或内容。
+
+## Input content to evaluate:
+
+"""
+    # process_response method is now inherited from BaseTextQuality

dingo/model/llm/vlm_layout_quality.py

@@ -201,8 +201,8 @@ class VLMLayoutQuality(BaseOpenAI):
         else:
             model_name = cls.client.models.list().data[0].id
 
-        params = cls.dynamic_config.parameters
-        cls.validate_config(params)
+        extra_params = cls.dynamic_config.model_extra
+        cls.validate_config(extra_params)
 
         completions = cls.client.chat.completions.create(
             model=model_name,

{dingo_python-2.2.1.dist-info → dingo_python-2.2.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dingo-python
-Version: 2.2.1
+Version: 2.2.2
 Summary: A Comprehensive AI Data Quality Evaluation Tool for Large Models
 Home-page: https://github.com/MigoXLab/dingo
 Author: Dingo
@@ -41,6 +41,7 @@ Provides-Extra: agent
 Requires-Dist: langchain>=1.0.0; extra == "agent"
 Requires-Dist: langchain-openai>=1.0.0; extra == "agent"
 Requires-Dist: tavily-python>=0.3.0; extra == "agent"
+Requires-Dist: arxiv>=2.4.0; extra == "agent"
 Provides-Extra: hhem
 Requires-Dist: transformers>=4.30.0; extra == "hhem"
 Requires-Dist: torch>=1.12.0; extra == "hhem"
@@ -54,6 +55,7 @@ Requires-Dist: tokenizers>=0.13.0; extra == "all"
 Requires-Dist: langchain>=1.0.0; extra == "all"
 Requires-Dist: langchain-openai>=1.0.0; extra == "all"
 Requires-Dist: tavily-python>=0.3.0; extra == "all"
+Requires-Dist: arxiv>=2.4.0; extra == "all"
 Dynamic: author
 Dynamic: classifier
 Dynamic: description
@@ -94,6 +96,7 @@ Dynamic: summary
   <a href="https://mseep.ai/app/dataeval-dingo"><img src="https://mseep.net/pr/dataeval-dingo-badge.png" alt="MseeP.ai Security Assessment Badge" height="20"></a>
   <a href="https://deepwiki.com/MigoXLab/dingo"><img src="https://deepwiki.com/badge.svg" alt="Ask DeepWiki"></a>
   <a href="https://archestra.ai/mcp-catalog/dataeval__dingo"><img src="https://archestra.ai/mcp-catalog/api/badge/quality/DataEval/dingo" alt="Trust Score"></a>
+  <a href="https://clawhub.ai/e06084/dingo"><img src="https://img.shields.io/badge/ClawHub-Skill-orange?logo=data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCI+PHRleHQgeT0iMTgiIGZvbnQtc2l6ZT0iMTYiPvCfpp48L3RleHQ+PC9zdmc+" alt="ClawHub Skill"></a>
 </p>
 
 </div>

{dingo_python-2.2.1.dist-info → dingo_python-2.2.2.dist-info}/RECORD

@@ -1,6 +1,6 @@
 dingo/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dingo/config/__init__.py,sha256=SaeOmGEUG0Hp5lqHxnHUTE_9ysN5KzA_Icilb9xY2mQ,349
-dingo/config/input_args.py,sha256=shJu6o7ZtiF5QTNDkpoi_HrP64urHmI-Qf5EKe8vOkg,4146
+dingo/config/input_args.py,sha256=48LEVcj7qIwIIVcPM7I29Mpl2YWbCGwv-nRdtVeTYcc,4147
 dingo/data/__init__.py,sha256=reCw4XQoInUTtvRW6c1wY_LH1EWJ7XpZDQcBCW61Lf8,214
 dingo/data/converter/__init__.py,sha256=1MiG4H8Sg2sYHQmYdg0F9_1okP_YoMNHyQorPEAf6zw,91
 dingo/data/converter/base.py,sha256=_WXa_plKj83iFgQyHABchGbX-dv3d17QuODua-bd83w,12820
@@ -38,7 +38,7 @@ dingo/model/model.py,sha256=4Y73hETATJVzwb9p62D0NV7STpDUIJo4Sx0_NYak68w,6106
 dingo/model/llm/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dingo/model/llm/base.py,sha256=n5ZHJNoJ0XSeG2i6ydN3W6pUYSAJaQgirjT_CiaJUlU,384
 dingo/model/llm/base_lmdeploy_apiclient.py,sha256=fTfSyqynGH-C29IijIU0euIWpc3BxoYXEEj-UEJWtCA,3457
-dingo/model/llm/base_openai.py,sha256=3RXXExlZQec5IF2XHL6qjxRjMccxkhtSoxr75JWKSXs,8240
+dingo/model/llm/base_openai.py,sha256=_qu96G0nVQcJdyashrbXGbTCTEuz6WkL1Uh4TjXNS28,7894
 dingo/model/llm/llm_classify_qr.py,sha256=CpaXg1jaRRxBHJEpznV8qr4thVrK93vd5K65ubKp1QI,2568
 dingo/model/llm/llm_classify_topic.py,sha256=AjSXi6KR5sVrkuxPfTYF-HoQlIMa-OshlJjFgdIE48w,5185
 dingo/model/llm/llm_dataman_assessment.py,sha256=yxHn3wc106XqKKyAaslJ-ZQhMinFYnEPmaHLOcGsqss,5574
@@ -55,14 +55,14 @@ dingo/model/llm/llm_text_chaos.py,sha256=tuvhSyatcImj8ZRB6-Ah2F7lmfe41sEHVxHMEdr
 dingo/model/llm/llm_text_code_list_issue.py,sha256=hEa6L-_uc-lp_0cTMl3KmlHjKxJiChcC4acGg9bQGCY,3392
 dingo/model/llm/llm_text_kaoti.py,sha256=8-MAALqF-iBNSE3Qukh5Tt9LhM8BF4ywR7KhjbllItk,8047
 dingo/model/llm/vlm_image_relevant.py,sha256=nuOqMSxfqHPdYzCXy-le3baX9-RIKnSdCoy1IFwK1Bk,4209
-dingo/model/llm/vlm_layout_quality.py,sha256=1MBevCUvetrPwC0lZclDCKIjn69U0QeVRj_-3J9hU5w,14147
+dingo/model/llm/vlm_layout_quality.py,sha256=hILcPWmxls0dr9XF1idSxmE4KjSfamJ-CJGp4sBfIVU,14160
 dingo/model/llm/vlm_ocr_understanding.py,sha256=_fmcYWeoh4rNx7WrkVx8PhxJ8JEw-nNm-nxFunuiMt4,7810
 dingo/model/llm/agent/__init__.py,sha256=gPo09JDUrctXbiqruFlR_rs1et0Nz1_Au3N_xAOTLTg,718
-dingo/model/llm/agent/agent_article_fact_checker.py,sha256=gkeFHXrQ43qcLfdSntj56ZG_kuEIn8Da1yC7vYM8ZCk,76515
-dingo/model/llm/agent/agent_fact_check.py,sha256=kLgCt4KhMkXo0vl3rDB6seooNvJnv1hiU2YhV6scRPg,16024
-dingo/model/llm/agent/agent_hallucination.py,sha256=IzjC0DXlZHDfwIWSdJlWuAKZvAqnBDCnhiFzhia1tIY,16219
-dingo/model/llm/agent/agent_wrapper.py,sha256=04dLx6CeDkHbSOWEYGRbeG2fXFRykASMeqqmViXauxk,12386
-dingo/model/llm/agent/base_agent.py,sha256=Z1qVVG35ZZUYgdRlmycmUZO1Lo3_q-vVVeygUTd0Ez4,17076
+dingo/model/llm/agent/agent_article_fact_checker.py,sha256=oS5O-LOYjc3CsTzqL6ui7jTxB0_amNHYVrFTV6NFvq8,80030
+dingo/model/llm/agent/agent_fact_check.py,sha256=fRvDoULR-bL6VvXpuKWkGH3nKHyDFoXEMsEGnW3bOOQ,15942
+dingo/model/llm/agent/agent_hallucination.py,sha256=p0OX2SSwoo2cQJouIjZJiHmFJdlNRvAqD7Bpwx6RrDo,16137
+dingo/model/llm/agent/agent_wrapper.py,sha256=EI-nYEM0Jwn3wJZUYSiEy-820G_vKTHItSpInrH5qVg,12417
+dingo/model/llm/agent/base_agent.py,sha256=Gedq9WrFVFkrS-jzjg_EZj6QeR6Bj-xABqIiOSozMnI,17079
 dingo/model/llm/agent/langchain_adapter.py,sha256=lVYr5yRha3vIRPwqne1wzz0fB0hjsN-rC0NWExJHZ2c,7438
 dingo/model/llm/agent/tools/__init__.py,sha256=HRUXnmkOdoC8XOq_YRg4yuhfeANNrDL5kI01L4C1cSc,590
 dingo/model/llm/agent/tools/arxiv_search.py,sha256=TqkJfqsBTmxmmzL-Ijw5rBBZGCqr3rsvXoY77lzs-fc,19645
@@ -76,7 +76,7 @@ dingo/model/llm/compare/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG
 dingo/model/llm/compare/llm_code_compare.py,sha256=ihWgJQ22Ycsa3kcY6K2_vN0dZzCR2HTyXv2LJnwPpiI,7077
 dingo/model/llm/compare/llm_html_extract_compare.py,sha256=23sK61QbEe3ap6LQh9NmxmjrVpl_0G5qfQH2oYQuYr8,6443
 dingo/model/llm/compare/llm_html_extract_compare_en.py,sha256=siPbaCcCWqvHNyHduul3wCI9lfSWC08f81JMZH-Ebh4,5427
-dingo/model/llm/compare/llm_html_extract_compare_v2.py,sha256=8oMgwNIhiTcOYue-_rk1HtycznyrVJPP77wFRk9jziA,11471
+dingo/model/llm/compare/llm_html_extract_compare_v2.py,sha256=-HWm4NLBTaQmF9fcjEs2V8IOHPo4f_hegdPxjZckpzA,11940
 dingo/model/llm/compare/llm_math_compare.py,sha256=Jseu9i6cCQ0uXxEp7SovraKDxDZkRdsGwI0hfadbdoQ,7662
 dingo/model/llm/compare/llm_table_compare.py,sha256=zw7JhFK1v-NZitOIRmTdbw-GdWkUGMyah5N1gFsdMF4,7564
 dingo/model/llm/hhh/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -85,8 +85,8 @@ dingo/model/llm/hhh/llm_text_3h_harmless.py,sha256=BEllrmFzBfGJvC_gN4TOeit9FXAAT
 dingo/model/llm/hhh/llm_text_3h_helpful.py,sha256=OrK3chIL6KOTnIHDZciICQNM5pURhv3KhtzuXlcjCWE,2397
 dingo/model/llm/hhh/llm_text_3h_honest.py,sha256=4xgHwzbqfuK_HNB8qqwYI-YVXfaGy2c6U5i-wltnVdY,2151
 dingo/model/llm/instruction_quality/__init__.py,sha256=PRFsZUG1oBA6EP74b5eQCs-gaLkG5_W2isOeX6TCcxM,708
-dingo/model/llm/instruction_quality/llm_instruction_clarity.py,sha256=6NWzSc8HagBapt-wTawzCc38Ma8e2XnBx2qdftRgJyg,11714
-dingo/model/llm/instruction_quality/llm_task_difficulty.py,sha256=9xBYZq_Rxhf6ijGLlMmQEBR4Qze1tKNjOMEYzmDXcPU,14729
+dingo/model/llm/instruction_quality/llm_instruction_clarity.py,sha256=sq_UdY_y0zRsow3KtyTDTlPgITSDZ5ludtVjhTKS-pA,11681
+dingo/model/llm/instruction_quality/llm_task_difficulty.py,sha256=iSwYrbRYD33Gqu_2KEr7zSrJRAuge7qsuWuhQxXVZ1U,14712
 dingo/model/llm/meta_rater/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dingo/model/llm/meta_rater/llm_meta_rater_cleanliness.py,sha256=PzpHBzuVZJ4FRuBEsWdyJtFErW09-K3fpBU6GFp81Gk,6879
 dingo/model/llm/meta_rater/llm_meta_rater_professionalism.py,sha256=o_J9KpFzYujMzvnYTNVdJD1Ka-icq-yvUwivmSnOgKc,7435
@@ -105,22 +105,24 @@ dingo/model/llm/minor_lan/llm_text_language_sr.py,sha256=ff-cV1XJHTy7UI27LbWN7b2
 dingo/model/llm/minor_lan/llm_text_language_th.py,sha256=4t5yqwGysGFmqwYOvfNydsqAvJ4ZLe__iFyqYfPib6k,1232
 dingo/model/llm/minor_lan/llm_text_language_vi.py,sha256=IihiHbONc1dCumJb1FI4c79sSP1kkcImbdSgopd-oBU,1244
 dingo/model/llm/rag/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-dingo/model/llm/rag/llm_rag_answer_relevancy.py,sha256=7KpACS_6la4Kltj05DvHPrgyduu2_JzaTO86-WoBbY0,12003
-dingo/model/llm/rag/llm_rag_context_precision.py,sha256=V1Kdse_dDOB_NnhaA3yoOmDHVvsnfyoslNa3wsELy-U,11327
-dingo/model/llm/rag/llm_rag_context_recall.py,sha256=Bwv4nRJZaCr1XumFXkXKha41ePr64ztEZ7For33Ezns,10518
-dingo/model/llm/rag/llm_rag_context_relevancy.py,sha256=v_OluJPEglb4wewZ02h_8m4FieFtmmFMCiknRLlhWdg,7671
-dingo/model/llm/rag/llm_rag_faithfulness.py,sha256=fLU0UuCXY3PvUHQAivKuJOuf51LjiUqVmwLWxv9xo_s,9444
+dingo/model/llm/rag/llm_rag_answer_relevancy.py,sha256=GAs0-vUURALZL825RHc8d6agrD08C5nO289KB1iJ7sE,11554
+dingo/model/llm/rag/llm_rag_context_precision.py,sha256=7oX1hMDLlp4oD4aobATYl4TPCcJd6IVAR8RZO_3psmI,11294
+dingo/model/llm/rag/llm_rag_context_recall.py,sha256=SkZx1UByl35y50d5Ckwk-6LPTOSCehgmlgVOycFhXPE,10485
+dingo/model/llm/rag/llm_rag_context_relevancy.py,sha256=OjtFKufsbgDL2aIHKzuM3hHdCrYpdXKy9tsQbuBhbwU,7638
+dingo/model/llm/rag/llm_rag_faithfulness.py,sha256=x2BkMu5SKybjUh0b3yiC6ZhHNXf_LM3JdK8gprf-R8U,9411
 dingo/model/llm/security/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 dingo/model/llm/security/llm_security.py,sha256=f2PvPUGfm3hhvI4FFlnz_0G5a2pGZeA7-3mk8FLe0i8,1194
 dingo/model/llm/security/llm_security_politics.py,sha256=dYMQagmIGJqnAPXY_LCHLH6R7JcJxDSnzxNtFzWbEF0,1314
 dingo/model/llm/security/llm_security_prohibition.py,sha256=arPI0jIvHbMibrGReBBAadEiXVeLYXR_aVRukf3lZ0Y,591
 dingo/model/llm/text_quality/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-dingo/model/llm/text_quality/base_text_quality.py,sha256=s2DrpWHNlmxqq7d3uOYFOR8sb5PjBi5GjphqeCvrIYM,2069
+dingo/model/llm/text_quality/base_text_quality.py,sha256=dI1Byf2jR3KENX0ieoLVSnWysuLeL_lb5gomxbQ4Ozo,1858
+dingo/model/llm/text_quality/llm_text_equation.py,sha256=zTsAO0NwSuCjUJNEPJ4mN2lfYrhcsac5AQOsDuRIgT0,2753
 dingo/model/llm/text_quality/llm_text_quality_v2.py,sha256=Kgu-hkvhXDT5PzG4a2TSkzQZBb8fvlYHy188GY2XzCI,3523
 dingo/model/llm/text_quality/llm_text_quality_v3.py,sha256=Tqg6AcdSmnTsyra5wQ_LzBmmfYEoi0NPhlRWdajtN_4,5341
 dingo/model/llm/text_quality/llm_text_quality_v4.py,sha256=EEn3CDTnp44cW3cmvXs2AGIV_PoK9_ySLA-F3ToQDKs,4659
-dingo/model/llm/text_quality/llm_text_quality_v5.py,sha256=UJF8K5vINAcyuBP9EmfTvsrLTg54bQR7zWZanIgSz6I,7776
+dingo/model/llm/text_quality/llm_text_quality_v5.py,sha256=zGXV6yodZnNonjtXXPLH1NUBOkKeLWYug3i1klmDiQc,10010
 dingo/model/llm/text_quality/llm_text_repeat.py,sha256=4vVO1L2jIXMIRT7UOAW5R2wMN4KQfGdXC9H23wE6VvQ,1764
+dingo/model/llm/text_quality/llm_text_table.py,sha256=E09Ye2pkCGn1NO050_PuFjmXREW7NK3JqJAbv2U1WpQ,2720
 dingo/model/llm/text_quality/llm_text_unread_issue.py,sha256=wklbD5Znt7LVTiXB-DH8-Yvsb7OQnh25ThA2mqDVYkQ,3669
 dingo/model/llm/text_quality/llm_text_word_stick.py,sha256=yqGAu3WHPXiECrpmPJd1GJproBrqzlZLcXpS9HkzWx4,3551
 dingo/model/response/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -146,9 +148,9 @@ dingo/utils/__init__.py,sha256=d8nJluje6i4z_Bb1rcXJSmEoAhyn1mkqEXJEOEdaMy4,51
 dingo/utils/exception.py,sha256=fh58dSLSmYSnwW4MQXg-Jfai2QcZfDruTaYGbaWk7Wc,446
 dingo/utils/log_util/__init__.py,sha256=VfzAAHUV8RuN-QaySahfAPfhM__-myigUlKx7ywVerA,717
 dingo/utils/log_util/logger.py,sha256=spGK0w22UgXsCcArd1rpt2teLPy7QPlIuvBaKYioHdY,1414
-dingo_python-2.2.1.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-dingo_python-2.2.1.dist-info/METADATA,sha256=KhY2l0LRzJErpLDrRsbK8oe93xr_Ukh69-MUYqzpaGA,26694
-dingo_python-2.2.1.dist-info/WHEEL,sha256=SmOxYU7pzNKBqASvQJ7DjX3XGUF92lrGhMb3R6_iiqI,91
-dingo_python-2.2.1.dist-info/entry_points.txt,sha256=Vo_p8qSVnOENdy1uubqxJRppZIpiQ753JG3WPAUeYps,45
-dingo_python-2.2.1.dist-info/top_level.txt,sha256=gSXQSLowu_WOQRi75wK3qyjbHxeN5PqsaA4ChGmJdek,6
-dingo_python-2.2.1.dist-info/RECORD,,
+dingo_python-2.2.2.dist-info/licenses/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+dingo_python-2.2.2.dist-info/METADATA,sha256=tBPvDUcDxzjjBcM170bnA7fYVeWvWuW7M4IbIMK054Y,27086
+dingo_python-2.2.2.dist-info/WHEEL,sha256=SmOxYU7pzNKBqASvQJ7DjX3XGUF92lrGhMb3R6_iiqI,91
+dingo_python-2.2.2.dist-info/entry_points.txt,sha256=Vo_p8qSVnOENdy1uubqxJRppZIpiQ753JG3WPAUeYps,45
+dingo_python-2.2.2.dist-info/top_level.txt,sha256=gSXQSLowu_WOQRi75wK3qyjbHxeN5PqsaA4ChGmJdek,6
+dingo_python-2.2.2.dist-info/RECORD,,