routed-confidence 0.1.0__tar.gz

routed_confidence-0.1.0/PKG-INFO

@@ -0,0 +1,195 @@
+Metadata-Version: 2.4
+Name: routed-confidence
+Version: 0.1.0
+Summary: Field-level confidence evaluator SDK for routed extraction results
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: dotenv>=0.9.9
+Requires-Dist: fastapi>=0.116.1
+Requires-Dist: httpx[socks]>=0.28.1
+Requires-Dist: openai>=1.107.1
+Requires-Dist: pydantic>=2.11.7
+Requires-Dist: pyyaml>=6.0.2
+Requires-Dist: uvicorn>=0.35.0
+
+# routed-confidence
+
+字段级置信度评估 SDK。
+
+这个包只负责一件事：输入一条提取结果，输出每个字段的置信度分数。它不计算检出率、误伤率、日报统计，也不包含 Web、API 服务或自动迭代逻辑。
+
+## 位置
+
+源码目录：
+
+```text
+routed_confidence/
+```
+
+主要入口：
+
+```text
+routed_confidence/evaluator.py
+```
+
+内部评分组件：
+
+```text
+routed_confidence/internal/confidence_evaluators/
+```
+
+默认规则文件：
+
+```text
+routed_confidence/rules/meeting_ruler.json
+```
+
+## 安装
+
+在当前仓库本地开发安装：
+
+```bash
+pip install -e .
+```
+
+从本地 wheel 安装：
+
+```bash
+python3 -m pip wheel . --no-deps --no-build-isolation -w dist
+pip install dist/routed_confidence-0.1.0-py3-none-any.whl
+```
+
+如果发布到 PyPI 或私有 PyPI 后：
+
+```bash
+pip install routed-confidence
+```
+
+Python 导入名使用下划线：
+
+```python
+from routed_confidence import ConfidenceEvaluator
+```
+
+## 使用
+
+```python
+from routed_confidence import ConfidenceEvaluator
+
+record = {
+    "extraction_result": {
+        "会议召开时间": {
+            "value": "2026-05-20 14:00:00",
+            "content": "会议召开时间为2026年5月20日14:00",
+            "type": "text",
+            "title_path": ["一、会议基本情况"],
+        }
+    },
+    "input_text": "会议召开时间为2026年5月20日14:00。",
+}
+
+evaluator = ConfidenceEvaluator()
+result = evaluator.evaluate(record)
+print(result)
+```
+
+也可以使用一次性函数：
+
+```python
+from routed_confidence import evaluate_confidence
+
+result = evaluate_confidence(record)
+```
+
+## 返回格式
+
+```json
+{
+  "field_scores": {
+    "会议召开时间": {
+      "source_value": "2026-05-20 14:00:00",
+      "total_score": 0.75,
+      "dimension_scores": {
+        "schema": 1.0,
+        "similarity": 0.82,
+        "historical": 0.5,
+        "relation": 1.0
+      },
+      "reason": "字段符合规则要求",
+      "violations": [],
+      "suggestions": []
+    }
+  },
+  "summary": {
+    "total_fields": 1,
+    "average_score": 0.75
+  }
+}
+```
+
+注意：SDK 不返回 `is_correct`。`is_correct` 属于“标注数据 vs 提取数据”的评测逻辑，不属于置信度 SDK。
+
+## 自定义规则和权重
+
+```python
+from routed_confidence import ConfidenceEvaluator
+
+evaluator = ConfidenceEvaluator(
+    rules_file="/path/to/meeting_ruler.json",
+    weights={
+        "schema": 0.1,
+        "sim": 0.2,
+        "wilson": 0.1,
+        "relation": 0.6,
+    },
+)
+```
+
+如果不传 `rules_file`，SDK 会优先寻找当前项目里的 `schema/ruler/meeting_ruler.json`，找不到时使用包内置的默认规则。
+
+## 发布
+
+安装构建工具：
+
+```bash
+pip install build twine
+```
+
+构建：
+
+```bash
+python -m build
+```
+
+会生成：
+
+```text
+dist/routed_confidence-0.1.0-py3-none-any.whl
+dist/routed_confidence-0.1.0.tar.gz
+```
+
+上传到 PyPI：
+
+```bash
+twine upload dist/*
+```
+
+上传到私有 PyPI：
+
+```bash
+twine upload --repository-url https://your-private-pypi/simple/ dist/*
+```
+
+发布后安装：
+
+```bash
+pip install routed-confidence
+```
+
+## 当前边界
+
+- 只做单条/字段级置信度评估。
+- 不计算检出率。
+- 不计算误伤率。
+- 不依赖 `methods/api`、`methods/web_conf`、`methods/reflect_rule`。
+- 不要求用户导入内部四维组件。

routed_confidence-0.1.0/README.md

@@ -0,0 +1 @@
+# RouteD - 数据提取置信度评估API

routed_confidence-0.1.0/pyproject.toml

@@ -0,0 +1,25 @@
+[project]
+name = "routed-confidence"
+version = "0.1.0"
+description = "Field-level confidence evaluator SDK for routed extraction results"
+readme = "routed_confidence/README.md"
+requires-python = ">=3.10"
+dependencies = [
+    "dotenv>=0.9.9",
+    "fastapi>=0.116.1",
+    "httpx[socks]>=0.28.1",
+    "openai>=1.107.1",
+    "pydantic>=2.11.7",
+    "pyyaml>=6.0.2",
+    "uvicorn>=0.35.0",
+]
+
+[build-system]
+requires = ["setuptools>=68", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[tool.setuptools.packages.find]
+include = ["routed_confidence*"]
+
+[tool.setuptools.package-data]
+routed_confidence = ["config.yaml", "rules/*.json"]

routed_confidence-0.1.0/routed_confidence/README.md

@@ -0,0 +1,181 @@
+# routed-confidence
+
+字段级置信度评估 SDK。
+
+这个包只负责一件事：输入一条提取结果，输出每个字段的置信度分数。它不计算检出率、误伤率、日报统计，也不包含 Web、API 服务或自动迭代逻辑。
+
+## 位置
+
+源码目录：
+
+```text
+routed_confidence/
+```
+
+主要入口：
+
+```text
+routed_confidence/evaluator.py
+```
+
+内部评分组件：
+
+```text
+routed_confidence/internal/confidence_evaluators/
+```
+
+默认规则文件：
+
+```text
+routed_confidence/rules/meeting_ruler.json
+```
+
+## 安装
+
+在当前仓库本地开发安装：
+
+```bash
+pip install -e .
+```
+
+从本地 wheel 安装：
+
+```bash
+python3 -m pip wheel . --no-deps --no-build-isolation -w dist
+pip install dist/routed_confidence-0.1.0-py3-none-any.whl
+```
+
+如果发布到 PyPI 或私有 PyPI 后：
+
+```bash
+pip install routed-confidence
+```
+
+Python 导入名使用下划线：
+
+```python
+from routed_confidence import ConfidenceEvaluator
+```
+
+## 使用
+
+```python
+from routed_confidence import ConfidenceEvaluator
+
+record = {
+    "extraction_result": {
+        "会议召开时间": {
+            "value": "2026-05-20 14:00:00",
+            "content": "会议召开时间为2026年5月20日14:00",
+            "type": "text",
+            "title_path": ["一、会议基本情况"],
+        }
+    },
+    "input_text": "会议召开时间为2026年5月20日14:00。",
+}
+
+evaluator = ConfidenceEvaluator()
+result = evaluator.evaluate(record)
+print(result)
+```
+
+也可以使用一次性函数：
+
+```python
+from routed_confidence import evaluate_confidence
+
+result = evaluate_confidence(record)
+```
+
+## 返回格式
+
+```json
+{
+  "field_scores": {
+    "会议召开时间": {
+      "source_value": "2026-05-20 14:00:00",
+      "total_score": 0.75,
+      "dimension_scores": {
+        "schema": 1.0,
+        "similarity": 0.82,
+        "historical": 0.5,
+        "relation": 1.0
+      },
+      "reason": "字段符合规则要求",
+      "violations": [],
+      "suggestions": []
+    }
+  },
+  "summary": {
+    "total_fields": 1,
+    "average_score": 0.75
+  }
+}
+```
+
+注意：SDK 不返回 `is_correct`。`is_correct` 属于“标注数据 vs 提取数据”的评测逻辑，不属于置信度 SDK。
+
+## 自定义规则和权重
+
+```python
+from routed_confidence import ConfidenceEvaluator
+
+evaluator = ConfidenceEvaluator(
+    rules_file="/path/to/meeting_ruler.json",
+    weights={
+        "schema": 0.1,
+        "sim": 0.2,
+        "wilson": 0.1,
+        "relation": 0.6,
+    },
+)
+```
+
+如果不传 `rules_file`，SDK 会优先寻找当前项目里的 `schema/ruler/meeting_ruler.json`，找不到时使用包内置的默认规则。
+
+## 发布
+
+安装构建工具：
+
+```bash
+pip install build twine
+```
+
+构建：
+
+```bash
+python -m build
+```
+
+会生成：
+
+```text
+dist/routed_confidence-0.1.0-py3-none-any.whl
+dist/routed_confidence-0.1.0.tar.gz
+```
+
+上传到 PyPI：
+
+```bash
+twine upload dist/*
+```
+
+上传到私有 PyPI：
+
+```bash
+twine upload --repository-url https://your-private-pypi/simple/ dist/*
+```
+
+发布后安装：
+
+```bash
+pip install routed-confidence
+```
+
+## 当前边界
+
+- 只做单条/字段级置信度评估。
+- 不计算检出率。
+- 不计算误伤率。
+- 不依赖 `methods/api`、`methods/web_conf`、`methods/reflect_rule`。
+- 不要求用户导入内部四维组件。

routed_confidence-0.1.0/routed_confidence/__init__.py

@@ -0,0 +1,5 @@
+"""Public API for the routed confidence evaluation package."""
+
+from .evaluator import ConfidenceEvaluator, evaluate_confidence
+
+__all__ = ["ConfidenceEvaluator", "evaluate_confidence"]

routed_confidence-0.1.0/routed_confidence/config.py

@@ -0,0 +1,65 @@
+"""Package configuration accessors for routed_confidence."""
+
+from __future__ import annotations
+
+from importlib import resources
+from pathlib import Path
+from typing import Any
+
+import yaml
+
+
+def _load_config() -> dict[str, Any]:
+    config_resource = resources.files("routed_confidence").joinpath("config.yaml")
+    with config_resource.open("r", encoding="utf-8") as file:
+        return yaml.safe_load(file) or {}
+
+
+_config = _load_config()
+
+PROJECT_ROOT = str(Path.cwd())
+
+wilson_config = _config.get("wilson", {})
+WILSON_Z_SCORES = wilson_config.get("z_scores", {0.95: 1.96})
+DEFAULT_CONFIDENCE_LEVEL = wilson_config.get("default_confidence_level", 0.95)
+DEFAULT_NEUTRAL_CONFIDENCE = wilson_config.get("default_neutral_confidence", 0.5)
+
+scoring_config = _config.get("scoring", {})
+SCORE_THRESHOLD_HIGH = scoring_config.get("threshold_high", 0.8)
+SCORE_THRESHOLD_MEDIUM = scoring_config.get("threshold_medium", 0.5)
+SPECIAL_VALUE_CORRECT_SCORE = scoring_config.get("special_value_correct_score", 0.9)
+SPECIAL_VALUE_WARNING_SCORE = scoring_config.get("special_value_warning_score", 0.7)
+DATE_FORMAT_ERROR_SCORE = scoring_config.get("date_format_error_score", 0.3)
+MAPPING_CORRECT_SCORE = scoring_config.get("mapping_correct_score", 0.9)
+
+weights_config = _config.get("weights", {})
+REQUIRED_FIELD_WEIGHT = weights_config.get("required_field_weight", 2.0)
+OPTIONAL_FIELD_WEIGHT = weights_config.get("optional_field_weight", 1.0)
+CONTEXT_TYPE_WEIGHTS = weights_config.get("context_type_weights", {})
+DIMENSION_WEIGHTS = weights_config.get(
+    "dimension_weights",
+    {"schema": 0.1, "sim": 0.2, "wilson": 0.1, "relation": 0.6},
+)
+
+date_formats_config = _config.get("date_formats", {})
+SUPPORTED_DATE_FORMATS = date_formats_config.get(
+    "supported_formats",
+    ["%Y-%m-%d %H:%M:%S", "%Y-%m-%d"],
+)
+
+paths_config = _config.get("paths", {})
+DEFAULT_RULES_FILE_PATH = paths_config.get("default_rules_file", "")
+WILSON_ENTROPY_STATISTIC_FILE_PATH = paths_config.get(
+    "wilson_entropy_statistic_file",
+    "",
+)
+SIM_SCORE_FILE_PATH = paths_config.get("sim_score_file", "")
+HISTORICAL_DATA_FILE_PATH = paths_config.get("historical_data_file", "")
+STATISTIC_FILE_SUFFIX = paths_config.get("statistic_file_suffix", "_statistic.json")
+
+score_reasons_config = _config.get("score_reasons", {})
+SCORE_REASON_HIGH = score_reasons_config.get("high", "字段符合规则要求")
+SCORE_REASON_MEDIUM = score_reasons_config.get("medium", "字段基本符合要求，有轻微问题")
+SCORE_REASON_LOW = score_reasons_config.get("low", "字段存在严重问题")
+
+ENABLE_VERBOSE_LOGGING = _config.get("logging", {}).get("enable_verbose_logging", False)

routed_confidence-0.1.0/routed_confidence/config.yaml

@@ -0,0 +1,94 @@
+# 置信度评估系统配置文件
+
+# Wilson置信区间配置
+wilson:
+  # Wilson置信区间的z分数映射
+  z_scores:
+    0.90: 1.645
+    0.95: 1.96
+    0.99: 2.576
+  
+  # 默认置信水平
+  default_confidence_level: 0.95
+  
+  # 无数据时的默认中性置信度
+  default_neutral_confidence: 0.5
+
+# 规则评分配置
+scoring:
+  # 评分阈值
+  threshold_high: 0.8    # 高分阈值（字段符合规则要求）
+  threshold_medium: 0.5  # 中等分阈值（字段基本符合要求）
+  
+  # 特殊值处理分数
+  special_value_correct_score: 0.9   # 特殊值正确处理的分数
+  special_value_warning_score: 0.7   # 特殊值可能处理不当的分数
+  
+  # 日期格式错误的惩罚分数
+  date_format_error_score: 0.3
+  
+  # 值映射正确的分数
+  mapping_correct_score: 0.9
+
+# 权重配置
+weights:
+  # 字段重要性权重
+  required_field_weight: 2.0      # 必填字段权重
+  optional_field_weight: 1.0      # 非必填字段权重
+  
+  # 上下文类型权重
+  context_type_weights:
+    table: 0.1
+    text: 0.05
+    title: 0.02
+  # 不同维度权重
+  dimension_weights:
+    schema: 0.10
+    sim: 0.2
+    wilson: 0.10
+    relation: 0.60
+# 日期格式配置
+date_formats:
+  # 支持的日期格式列表
+  supported_formats:
+    - "%Y-%m-%d %H:%M:%S"  # 2025-05-16 14:30:00
+    - "%Y-%m-%d"           # 2025-05-16
+    - "%Y/%m/%d %H:%M:%S"  # 2025/05/16 14:30:00
+    - "%Y/%m/%d"           # 2025/05/16
+
+# 文件路径配置
+paths:
+  # 项目根目录将在运行时动态计算
+  # 默认规则文件路径
+  default_rules_file: "schema/ruler/meeting_ruler.json"
+  
+  # wilson&entropy 统计文件路径
+  wilson_entropy_statistic_file: "methods/data/wilson&ent"
+  
+  # 历史文件路径
+  historical_data_file: "data/wrong_pdf/10.22/test_data_fusion.json"
+  
+  # 统计文件后缀
+  statistic_file_suffix: "_statistic.json"
+
+  # 测试数据文件路径
+  test_data_file: "data/wrong_pdf/ori/10.24_before.json"
+  
+  # 相似度分数文件路径
+  sim_score_file: "methods/data/sim"
+  
+  # schema文件路径（用于AI评估）
+  schema_file: "schema/meeting.json"
+
+
+
+# 评分理由模板
+score_reasons:
+  high: "字段符合规则要求"
+  medium: "字段基本符合要求，有轻微问题"
+  low: "字段存在严重问题"
+
+# 日志配置
+logging:
+  # 是否启用详细日志
+  enable_verbose_logging: true