wogiflow 2.4.2 → 2.4.3

package/scripts/flow-long-input-constants.js

@@ -1,5 +1,3 @@
-'use strict';
-
 /**
  * Long Input Processing - Constants and Pattern Definitions
  *

package/scripts/flow-long-input-contradictions.js

@@ -0,0 +1,351 @@
+/**
+ * Long Input Processing - Contradiction Detection & Resolution
+ *
+ * Extracted from flow-long-input.js
+ * Handles contradiction detection between statements, orphan resolution via
+ * semantic expansion, correction phrase detection, and clarification persistence.
+ */
+
+const fs = require('node:fs');
+const path = require('node:path');
+const { safeJsonParse } = require('./flow-utils');
+
+const {
+  SEMANTIC_EXPANSIONS,
+  CORRECTION_PATTERNS,
+  ADDITIVE_PATTERNS
+} = require('./flow-long-input-constants');
+
+// These are injected via init()
+let _loadActiveDigest = null;
+let _calculateAssociationConfidence = null;
+
+/**
+ * Initialize with core functions from other modules.
+ * @param {Object} deps
+ */
+function init(deps) {
+  _loadActiveDigest = deps.loadActiveDigest;
+  _calculateAssociationConfidence = deps.calculateAssociationConfidence;
+}
+
+/**
+ * Detect contradictions between statements
+ */
+function detectContradictions(statements) {
+  const contradictions = [];
+  const meaningfulStatements = statements.filter(s => s.meaningful);
+
+  // Contradiction patterns
+  const opposites = [
+    ['left', 'right'],
+    ['top', 'bottom'],
+    ['show', 'hide'],
+    ['enable', 'disable'],
+    ['add', 'remove'],
+    ['include', 'exclude'],
+    ['before', 'after'],
+    ['above', 'below']
+  ];
+
+  for (let i = 0; i < meaningfulStatements.length; i++) {
+    for (let j = i + 1; j < meaningfulStatements.length; j++) {
+      const stmt1 = meaningfulStatements[i];
+      const stmt2 = meaningfulStatements[j];
+
+      // Only check statements in same topic
+      if (stmt1.topic_id !== stmt2.topic_id) continue;
+
+      const text1 = stmt1.text.toLowerCase();
+      const text2 = stmt2.text.toLowerCase();
+
+      // Check for opposite words
+      for (const [word1, word2] of opposites) {
+        if ((text1.includes(word1) && text2.includes(word2)) ||
+            (text1.includes(word2) && text2.includes(word1))) {
+          contradictions.push({
+            statement1_id: stmt1.id,
+            statement2_id: stmt2.id,
+            type: 'opposite_values',
+            attribute: `${word1}/${word2}`,
+            resolution: 'pending'
+          });
+        }
+      }
+
+      // Check for number conflicts (same attribute, different values)
+      const numbers1 = text1.match(/\d+/g);
+      const numbers2 = text2.match(/\d+/g);
+      if (numbers1 && numbers2) {
+        // Simple heuristic: if both mention numbers in similar context
+        const commonWords = text1.split(/\s+/).filter(w => text2.includes(w) && w.length > 3);
+        if (commonWords.length > 2 && numbers1[0] !== numbers2[0]) {
+          contradictions.push({
+            statement1_id: stmt1.id,
+            statement2_id: stmt2.id,
+            type: 'quantity_conflict',
+            values: [numbers1[0], numbers2[0]],
+            resolution: 'pending'
+          });
+        }
+      }
+    }
+  }
+
+  return contradictions;
+}
+
+/**
+ * Enhanced confidence calculation with semantic expansion
+ */
+function calculateExpandedConfidence(statement, topic) {
+  let confidence = 0.5;
+  const reasons = [];
+  const statementLower = statement.text.toLowerCase();
+
+  // Standard matching first
+  const { confidence: baseConf, reasons: baseReasons } = _calculateAssociationConfidence(statement, topic);
+  if (baseConf > confidence) {
+    confidence = baseConf;
+    reasons.push(...baseReasons);
+  }
+
+  // Semantic expansion - check synonyms
+  for (const [term, synonyms] of Object.entries(SEMANTIC_EXPANSIONS)) {
+    const allTerms = [term, ...synonyms];
+    const topicTitle = topic.title.toLowerCase();
+    const topicKeywords = (topic.keywords || []).map(k => k.toLowerCase());
+
+    for (const syn of allTerms) {
+      // Statement contains synonym and topic contains related term
+      if (statementLower.includes(syn)) {
+        for (const related of allTerms) {
+          if (topicTitle.includes(related) || topicKeywords.includes(related)) {
+            confidence = Math.max(confidence, 0.7);
+            reasons.push(`semantic_expansion:${syn}->${related}`);
+          }
+        }
+      }
+    }
+  }
+
+  return { confidence, reasons };
+}
+
+/**
+ * Try to resolve a single orphan statement
+ */
+function resolveOrphan(orphan, topics) {
+  const candidates = [];
+
+  // Try enhanced matching against all topics
+  for (const topic of topics) {
+    const { confidence, reasons } = calculateExpandedConfidence(orphan, topic);
+    if (confidence >= 0.5) {
+      candidates.push({ topic, confidence, reasons });
+    }
+  }
+
+  // Sort by confidence
+  candidates.sort((a, b) => b.confidence - a.confidence);
+
+  // Resolution decision
+  if (candidates.length === 0) {
+    return {
+      resolved: false,
+      method: 'no_match',
+      confidence: 0
+    };
+  }
+
+  const best = candidates[0];
+
+  // Clear winner
+  if (best.confidence >= 0.6 && (candidates.length === 1 || best.confidence - candidates[1].confidence > 0.15)) {
+    return {
+      resolved: true,
+      method: 'semantic_expansion',
+      topic_id: best.topic.id,
+      confidence: best.confidence,
+      reasons: best.reasons
+    };
+  }
+
+  // Ambiguous - multiple close matches
+  if (candidates.length > 1 && candidates[0].confidence - candidates[1].confidence < 0.1) {
+    return {
+      resolved: false,
+      method: 'ambiguous',
+      possible_topics: candidates.slice(0, 3).map(c => c.topic.id),
+      confidence: best.confidence
+    };
+  }
+
+  // Low confidence winner
+  return {
+    resolved: best.confidence >= 0.5,
+    method: best.confidence >= 0.5 ? 'context_reanalysis' : 'low_confidence',
+    topic_id: best.confidence >= 0.5 ? best.topic.id : null,
+    confidence: best.confidence,
+    reasons: best.reasons
+  };
+}
+
+/**
+ * Check if statement contains a correction phrase
+ */
+function detectCorrectionPhrase(text) {
+  for (const { pattern, name, weight } of CORRECTION_PATTERNS) {
+    if (pattern.test(text)) {
+      return { detected: true, phrase: name, weight };
+    }
+  }
+  return { detected: false, phrase: null, weight: 0 };
+}
+
+/**
+ * Check if statement is additive (not a real contradiction)
+ */
+function isAdditive(text) {
+  return ADDITIVE_PATTERNS.some(pattern => pattern.test(text));
+}
+
+/**
+ * Calculate resolution confidence for a contradiction
+ */
+function calculateResolutionConfidence(stmt1, stmt2, contradiction) {
+  let confidence = 0.5;
+  let reasons = [];
+
+  // Check for correction phrase in stmt2 (later statement)
+  const correction = detectCorrectionPhrase(stmt2.text);
+  if (correction.detected) {
+    confidence += correction.weight;
+    reasons.push(`correction_phrase:${correction.phrase}`);
+  }
+
+  // Same speaker increases confidence
+  if (stmt1.speaker && stmt2.speaker && stmt1.speaker === stmt2.speaker) {
+    confidence += 0.15;
+    reasons.push('same_speaker');
+  }
+
+  // Position difference - later statements typically override
+  const positionDiff = stmt2.position - stmt1.position;
+  if (positionDiff > 500) {  // Significant distance
+    confidence += 0.1;
+    reasons.push('later_position');
+  }
+
+  // Check if stmt2 explicitly references the attribute
+  const attr = contradiction.attribute;
+  if (attr) {
+    const [word1, word2] = attr.split('/');
+    if (stmt2.text.toLowerCase().includes(word1) || stmt2.text.toLowerCase().includes(word2)) {
+      confidence += 0.1;
+      reasons.push('explicit_attribute_reference');
+    }
+  }
+
+  // Check for additive pattern - might not be a real contradiction
+  if (isAdditive(stmt2.text)) {
+    confidence = 0.3;  // Low confidence - likely not a contradiction
+    reasons = ['additive_pattern'];
+  }
+
+  return {
+    confidence: Math.min(confidence, 1.0),
+    reasons,
+    winner: confidence >= 0.5 ? stmt2.id : null,
+    isAdditive: isAdditive(stmt2.text)
+  };
+}
+
+/**
+ * Generate clarification question for unresolved contradiction
+ */
+function generateContradictionQuestion(stmt1, stmt2, contradiction) {
+  const attr = contradiction.attribute || 'value';
+  const [val1, val2] = attr.split('/');
+
+  // Extract the actual values from statements if possible
+  const extractValue = (text, hints) => {
+    for (const hint of hints || []) {
+      if (text.toLowerCase().includes(hint.toLowerCase())) {
+        return hint;
+      }
+    }
+    return text.slice(0, 50);
+  };
+
+  const value1 = val1 || extractValue(stmt1.text, []);
+  const value2 = val2 || extractValue(stmt2.text, []);
+
+  return {
+    question: `You mentioned "${stmt1.text.slice(0, 60)}" but later said "${stmt2.text.slice(0, 60)}". Which do you prefer?`,
+    options: [
+      { id: 'opt-1', text: value1, statement_id: stmt1.id },
+      { id: 'opt-2', text: value2, statement_id: stmt2.id },
+      { id: 'opt-3', text: 'Both are needed', resolution: 'keep_both' }
+    ],
+    attribute: attr
+  };
+}
+
+/**
+ * Save clarifications to file
+ */
+function saveClarifications(clarifications) {
+  const activeDigest = _loadActiveDigest();
+  if (!activeDigest.session.digest_path) {
+    throw new Error('No active digest session');
+  }
+
+  const clarPath = path.join(activeDigest.session.digest_path, 'clarifications.json');
+  fs.writeFileSync(clarPath, JSON.stringify(clarifications, null, 2));
+  return clarifications;
+}
+
+/**
+ * Load clarifications from file
+ */
+function loadClarifications() {
+  const activeDigest = _loadActiveDigest();
+  if (!activeDigest.session.digest_path) {
+    return null;
+  }
+
+  const clarPath = path.join(activeDigest.session.digest_path, 'clarifications.json');
+  const defaultClarifications = {
+    questions: [],
+    contradictions: [],
+    metadata: {
+      total_questions: 0,
+      answered_questions: 0,
+      pending_questions: 0,
+      total_contradictions: 0,
+      resolved_contradictions: 0,
+      auto_resolved_count: 0,
+      user_resolved_count: 0
+    }
+  };
+  const data = safeJsonParse(clarPath, null);
+  if (!data) return defaultClarifications;
+  if (!data.questions) data.questions = [];
+  if (!data.contradictions) data.contradictions = [];
+  if (!data.metadata) data.metadata = defaultClarifications.metadata;
+  return data;
+}
+
+module.exports = {
+  init,
+  detectContradictions,
+  calculateExpandedConfidence,
+  resolveOrphan,
+  detectCorrectionPhrase,
+  isAdditive,
+  calculateResolutionConfidence,
+  generateContradictionQuestion,
+  saveClarifications,
+  loadClarifications
+};

package/scripts/flow-long-input-detection.js

@@ -1,5 +1,3 @@
-'use strict';
-
 /**
  * Long Input Processing - Large Input Detection & Content Classification
  *