MCP Dual-Cycle Reasoner

export const DESCRIPTIONS = { // Common field descriptions LAST_ACTION: 'Latest action name to be added to the accumulated action history', CURRENT_CONTEXT: 'Current environment context or state', GOAL: 'Current goal being pursued', // Loop detection LOOP_DETECTED: 'Whether a loop was detected', LOOP_TYPE: 'Type of loop detected', LOOP_CONFIDENCE: 'Confidence in loop detection', LOOP_DETAILS: 'Details about the detected loop', ACTIONS_INVOLVED: 'Actions involved in the loop', DETECTION_METHOD: 'Loop detection method to use: statistical, pattern or hybrid.', // Statistical metrics ENTROPY_SCORE: 'Statistical entropy score', VARIANCE_SCORE: 'Statistical variance score', TREND_SCORE: 'Statistical trend score', CYCLICITY_SCORE: 'Statistical cyclicity score', // Beliefs CURRENT_BELIEFS: 'Current beliefs as simple strings', CONTRADICTING_EVIDENCE: 'Evidence that contradicts current beliefs', INITIAL_BELIEFS: 'Initial beliefs about the task and environment', REVISED_BELIEFS: 'Updated beliefs as simple strings', REMOVED_BELIEFS: 'Beliefs that were removed', RATIONALE: 'Explanation for the changes', // Experience/Cases PROBLEM_DESCRIPTION: 'Simple description of the problem', SOLUTION: 'What action resolved the issue', OUTCOME: 'Whether the solution was successful', // Configuration WINDOW_SIZE: 'Size of the monitoring window', MAX_RESULTS: 'Maximum number of cases to return', // Configuration parameters PROGRESS_INDICATORS: 'Action patterns that indicate positive task progress (e.g., ["success", "complete", "found"])', MIN_ACTIONS_FOR_DETECTION: 'Minimum number of actions required before loop detection', ALTERNATING_THRESHOLD: 'Threshold for detecting alternating action patterns (0.0-1.0)', REPETITION_THRESHOLD: 'Threshold for detecting repetitive action patterns (0.0-1.0)', PROGRESS_THRESHOLD_ADJUSTMENT: 'How much to increase thresholds when progress indicators are present', SEMANTIC_INTENTS: 'Domain-specific action intents for semantic analysis (e.g., ["navigating", "clicking", "typing"])', // Threshold descriptions ENTROPY_THRESHOLD: 'Threshold for entropy-based anomaly detection', VARIANCE_THRESHOLD: 'Threshold for variance-based stagnation detectio cfn', TREND_THRESHOLD: 'Threshold for trend-based progress detection', CYCLICITY_THRESHOLD: 'Threshold for detecting cyclical patterns', // Action and state descriptions ACTION_NAME: "Action name (e.g., 'scroll_down', 'click_element')", ACTION_RESULT: 'Result or error from the action', ENVIRONMENT_CONTEXT: 'Current environment context or location', } as const;