check_data_quality
Validate CSV data quality by checking completeness, consistency, duplicates, uniqueness, data types, and outliers using predefined or custom rules.
Instructions
Check data quality based on predefined or custom rules.
Returns: DataQualityResult with comprehensive quality assessment results
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| rules | No | List of quality rules to check (None = use default rules) |
Implementation Reference
- Main handler function implementing the check_data_quality tool. Loads session data, applies configurable quality rules (or defaults), computes scores, detects issues, and returns comprehensive results with recommendations.
def check_data_quality( ctx: Annotated[Context, Field(description="FastMCP context for session access")], rules: Annotated[ list[QualityRuleType] | None, Field(description="List of quality rules to check (None = use default rules)"), ] = None, ) -> DataQualityResult: """Check data quality based on predefined or custom rules. Returns: DataQualityResult with comprehensive quality assessment results """ session_id = ctx.session_id _session, df = get_session_data(session_id) settings = get_settings() rule_results: list[QualityRuleResult] = [] quality_issues: list[QualityIssue] = [] recommendations: list[str] = [] # Apply resource management for large datasets logger.info("Checking data quality for %d rows, %d columns", len(df), len(df.columns)) if len(df) > settings.max_anomaly_sample_size: logger.warning( "Large dataset (%d rows), using sample of %d for quality check", len(df), settings.max_anomaly_sample_size, ) df = sample_large_dataset(df, settings.max_anomaly_sample_size, "Data quality check") # Use default rules if none provided if rules is None: rules = [ CompletenessRule(threshold=0.95), DuplicatesRule(threshold=0.01), DataTypesRule(), OutliersRule(threshold=0.05), ConsistencyRule(), ] total_score: float = 0 score_count = 0 for rule in rules: if isinstance(rule, CompletenessRule): # Check data completeness threshold = rule.threshold columns = rule.columns if rule.columns is not None else df.columns.tolist() for col in columns: if col in df.columns: completeness = 1 - (df[col].isna().sum() / len(df)) passed = completeness >= threshold score = completeness * 100 # Create issue if failed rule_issues = [] if not passed: issue = QualityIssue( type="incomplete_data", severity="high" if completeness < settings.data_completeness_threshold else "medium", column=col, message=f"Column '{col}' is only {round(completeness * 100, 2)}% complete", affected_rows=int(df[col].isna().sum()), metric_value=completeness, threshold=float(threshold), ) rule_issues.append(issue) quality_issues.append(issue) # Add rule result rule_results.append( QualityRuleResult( rule_type="completeness", passed=passed, score=round(score, 2), issues=rule_issues, column=col, ), ) total_score += score score_count += 1 elif isinstance(rule, DuplicatesRule): # Check for duplicate rows threshold = rule.threshold subset = rule.columns duplicates = df.duplicated(subset=subset) duplicate_ratio = duplicates.sum() / len(df) passed = duplicate_ratio <= threshold score = (1 - duplicate_ratio) * 100 # Create issue if failed rule_issues = [] if not passed: issue = QualityIssue( type="duplicate_rows", severity="high" if duplicate_ratio > settings.outlier_detection_threshold else "medium", message=f"Found {duplicates.sum()} duplicate rows ({round(duplicate_ratio * 100, 2)}%)", affected_rows=int(duplicates.sum()), metric_value=duplicate_ratio, threshold=float(threshold), ) rule_issues.append(issue) quality_issues.append(issue) recommendations.append( "Consider removing duplicate rows using the remove_duplicates tool", ) # Add rule result rule_results.append( QualityRuleResult( rule_type="duplicates", passed=passed, score=round(score, 2), issues=rule_issues, ), ) total_score += score score_count += 1 elif isinstance(rule, UniquenessRule): # Check column uniqueness column = rule.column if column in df.columns: unique_ratio = df[column].nunique() / len(df) expected_unique = rule.expected_unique if expected_unique: passed = unique_ratio >= settings.uniqueness_threshold score = unique_ratio * 100 else: passed = True score = 100.0 # Create issue if failed rule_issues = [] if not passed and expected_unique: duplicate_count = len(df) - df[column].nunique() issue = QualityIssue( type="non_unique_values", severity="high", column=str(column), message=f"Column '{column}' expected to be unique but has duplicates", affected_rows=duplicate_count, metric_value=unique_ratio, threshold=settings.uniqueness_threshold, ) rule_issues.append(issue) quality_issues.append(issue) # Add rule result rule_results.append( QualityRuleResult( rule_type="uniqueness", passed=passed, score=round(score, 2), issues=rule_issues, column=str(column), ), ) total_score += score score_count += 1 elif isinstance(rule, DataTypesRule): # Check data type consistency for col in df.columns: col_data = df[col].dropna() if len(col_data) > 0: # Check for mixed types types = col_data.apply(lambda x: type(x).__name__).unique() mixed_types = len(types) > 1 # Check for numeric strings if col_data.dtype == object: numeric_strings = col_data.astype(str).str.match(r"^-?\d+\.?\d*$").sum() numeric_ratio = numeric_strings / len(col_data) else: numeric_ratio = 0 score = 100.0 if not mixed_types else 50.0 # Create recommendations for numeric strings if numeric_ratio > settings.high_quality_threshold: recommendations.append( f"Column '{col}' appears to contain numeric data stored as strings. " f"Consider converting to numeric type using change_column_type tool", ) # Add rule result rule_results.append( QualityRuleResult( rule_type="data_type_consistency", passed=not mixed_types, score=score, issues=[], column=col, ), ) total_score += score score_count += 1 elif isinstance(rule, OutliersRule): # Check for outliers in numeric columns threshold = rule.threshold numeric_cols = df.select_dtypes(include=[np.number]).columns for col in numeric_cols: q1 = df[col].quantile(0.25) q3 = df[col].quantile(0.75) iqr = q3 - q1 lower_bound = q1 - 1.5 * iqr upper_bound = q3 + 1.5 * iqr outliers = ((df[col] < lower_bound) | (df[col] > upper_bound)).sum() outlier_ratio = outliers / len(df) passed = outlier_ratio <= threshold score = (1 - min(outlier_ratio, 1)) * 100 # Create issue if failed rule_issues = [] if not passed: issue = QualityIssue( type="outliers", severity="medium", column=col, message=f"Column '{col}' has {outliers} outliers ({round(outlier_ratio * 100, 2)}%)", affected_rows=int(outliers), metric_value=outlier_ratio, threshold=float(threshold), ) rule_issues.append(issue) quality_issues.append(issue) # Add rule result rule_results.append( QualityRuleResult( rule_type="outliers", passed=passed, score=round(score, 2), issues=rule_issues, column=col, ), ) total_score += score score_count += 1 elif isinstance(rule, ConsistencyRule): # Check data consistency columns = rule.columns # Date consistency check date_cols = df.select_dtypes(include=["datetime64"]).columns if len(date_cols) >= settings.min_statistical_sample_size and not columns: columns = date_cols.tolist() if len(columns) >= settings.min_statistical_sample_size: col1, col2 = str(columns[0]), str(columns[1]) if ( col1 in df.columns and col2 in df.columns and pd.api.types.is_datetime64_any_dtype(df[col1]) and pd.api.types.is_datetime64_any_dtype(df[col2]) ): inconsistent = (df[col1] > df[col2]).sum() consistency_ratio = 1 - (inconsistent / len(df)) passed = consistency_ratio >= settings.uniqueness_threshold score = consistency_ratio * 100 # Create issue if failed rule_issues = [] if not passed: issue = QualityIssue( type="data_inconsistency", severity="high", message=f"Found {inconsistent} rows where {col1} > {col2}", affected_rows=int(inconsistent), metric_value=consistency_ratio, threshold=settings.uniqueness_threshold, ) rule_issues.append(issue) quality_issues.append(issue) # Add rule result rule_results.append( QualityRuleResult( rule_type="consistency", passed=passed, score=round(score, 2), issues=rule_issues, ), ) total_score += score score_count += 1 # Calculate overall score overall_score = round(total_score / score_count, 2) if score_count > 0 else 100.0 # Add general recommendations if not recommendations and overall_score < settings.character_score_threshold: recommendations.append( "Consider running profile_data to get a comprehensive overview of data issues", ) # Count passed/failed rules passed_rules = sum(1 for rule in rule_results if rule.passed) failed_rules = len(rule_results) - passed_rules # Apply limits to quality issues to prevent resource exhaustion limited_issues, was_truncated = apply_violation_limits( quality_issues, settings.max_validation_violations, "Data quality check" ) if was_truncated: logger.warning( "Quality check found %d issues, limited to %d", len(quality_issues), settings.max_validation_violations, ) # Create QualityResults quality_results = QualityResults( overall_score=overall_score, passed_rules=passed_rules, failed_rules=failed_rules, total_issues=len(limited_issues), rule_results=rule_results, issues=limited_issues, recommendations=recommendations, ) # No longer recording operations (simplified MCP architecture) return DataQualityResult( quality_results=quality_results, ) - src/databeak/servers/validation_server.py:1308-1310 (registration)Registers the check_data_quality function as an MCP tool on the DataBeak-Validation FastMCP server instance.
validation_server.tool(name="validate_schema")(validate_schema) validation_server.tool(name="check_data_quality")(check_data_quality) validation_server.tool(name="find_anomalies")(find_anomalies) - Output schema model for the tool response, containing QualityResults with overall score, rule results, issues, and recommendations.
class DataQualityResult(BaseModel): """Response model for data quality check operations.""" quality_results: QualityResults = Field(description="Comprehensive quality assessment results") - Input schema: Discriminated union type for quality rules (completeness, duplicates, etc.) used to configure custom checks.
QualityRuleType = Annotated[ CompletenessRule | DuplicatesRule | UniquenessRule | DataTypesRule | OutliersRule | ConsistencyRule, Field(discriminator="type"), ] - Default quality rules applied when no custom rules are provided.
rules = [ CompletenessRule(threshold=0.95), DuplicatesRule(threshold=0.01), DataTypesRule(), OutliersRule(threshold=0.05), ConsistencyRule(),