Clear Thought 1.5

/** * Notebook Presets for Reasoning Patterns * * Pre-configured notebook templates for each reasoning pattern * based on research implementations. */ export const TREE_OF_THOUGHT_PRESET = { name: 'Tree of Thought', description: 'Systematic exploration of multiple reasoning paths', cells: [ { type: 'markdown', source: `# Tree of Thought Reasoning This notebook implements Tree of Thought (ToT) reasoning for systematic exploration of solution paths. ## Problem Statement Define your problem here...` }, { type: 'code', source: `// Tree of Thought Implementation class TreeNode { constructor(state, thought, value = null) { this.state = state; this.thought = thought; this.value = value; this.children = []; this.parent = null; } addChild(childNode) { childNode.parent = this; this.children.push(childNode); return childNode; } evaluate() { // Placeholder for evaluation logic return Math.random(); } } class TreeOfThought { constructor(rootPrompt, maxDepth = 3, branchFactor = 3) { this.root = new TreeNode('', rootPrompt); this.maxDepth = maxDepth; this.branchFactor = branchFactor; this.bestPath = []; } explore(node, depth = 0) { if (depth >= this.maxDepth) { return node.evaluate(); } // Generate child thoughts for (let i = 0; i < this.branchFactor; i++) { const childThought = \`Branch \${i + 1} from: \${node.thought.substring(0, 20)}...\`; const childNode = node.addChild(new TreeNode( node.state + '\\n' + childThought, childThought )); // Recursively explore childNode.value = this.explore(childNode, depth + 1); } // Select best child if (node.children.length > 0) { const bestChild = node.children.reduce((best, child) => child.value > best.value ? child : best ); node.value = bestChild.value; } return node.value; } solve() { this.explore(this.root); return this.extractBestPath(this.root); } extractBestPath(node) { const path = [node.thought]; while (node.children.length > 0) { node = node.children.reduce((best, child) => child.value > best.value ? child : best ); path.push(node.thought); } return path; } } // Example usage const problem = "How to optimize database queries?"; const tot = new TreeOfThought(problem, 3, 2); const solution = tot.solve(); console.log("Best solution path:", solution);`, language: 'javascript' }, { type: 'markdown', source: `## Visualization \`\`\`mermaid graph TD A[Root: Problem] --> B[Branch 1] A --> C[Branch 2] B --> D[Sub-branch 1.1] B --> E[Sub-branch 1.2] C --> F[Sub-branch 2.1] C --> G[Sub-branch 2.2] style B fill:#90EE90 style D fill:#90EE90 \`\`\` Green nodes indicate the selected path.` } ] }; export const BEAM_SEARCH_PRESET = { name: 'Beam Search', description: 'Parallel exploration with periodic pruning', cells: [ { type: 'markdown', source: `# Beam Search Reasoning Maintains multiple promising paths simultaneously, exploring them in parallel with periodic pruning.` }, { type: 'code', source: `// Beam Search Implementation class BeamSearchPath { constructor(thoughts = [], score = 0) { this.thoughts = thoughts; this.score = score; this.status = 'active'; } expand(newThought, scoreIncrement) { return new BeamSearchPath( [...this.thoughts, newThought], this.score + scoreIncrement ); } } class BeamSearch { constructor(problem, beamWidth = 3, maxIterations = 5) { this.problem = problem; this.beamWidth = beamWidth; this.maxIterations = maxIterations; this.beam = [new BeamSearchPath([problem], 0)]; } generateCandidates(path) { // Generate next steps from current path const candidates = []; for (let i = 0; i < 3; i++) { const thought = \`Step \${path.thoughts.length}: Option \${i + 1}\`; const score = Math.random(); // Placeholder scoring candidates.push(path.expand(thought, score)); } return candidates; } iterate() { const allCandidates = []; // Generate candidates for each path in beam for (const path of this.beam) { const candidates = this.generateCandidates(path); allCandidates.push(...candidates); } // Sort by score and keep top beamWidth allCandidates.sort((a, b) => b.score - a.score); this.beam = allCandidates.slice(0, this.beamWidth); // Log current beam console.log(\`Beam after iteration:\`); this.beam.forEach((path, i) => { console.log(\` Path \${i + 1} (score: \${path.score.toFixed(2)}):\`, path.thoughts[path.thoughts.length - 1]); }); } search() { for (let i = 0; i < this.maxIterations; i++) { console.log(\`\\nIteration \${i + 1}:\`); this.iterate(); } return this.beam[0]; // Return best path } } // Example usage const beamSearch = new BeamSearch( "Design a distributed cache system", 3, // beam width 4 // iterations ); const bestPath = beamSearch.search(); console.log("\\nBest path found:", bestPath.thoughts);`, language: 'javascript' } ] }; export const MCTS_PRESET = { name: 'Monte Carlo Tree Search', description: 'Balances exploration and exploitation through simulations', cells: [ { type: 'markdown', source: `# Monte Carlo Tree Search (MCTS) MCTS combines tree search with random sampling for decision-making under uncertainty. ## Four Phases: 1. **Selection**: Navigate to promising leaf using UCB 2. **Expansion**: Add new child node 3. **Simulation**: Random rollout to terminal state 4. **Backpropagation**: Update values up the tree` }, { type: 'code', source: `// MCTS Implementation for Reasoning class MCTSNode { constructor(state, parent = null) { this.state = state; this.parent = parent; this.children = []; this.visits = 0; this.value = 0; this.untried_actions = this.getPossibleActions(); } getPossibleActions() { // Return possible next thoughts/actions return ['Action A', 'Action B', 'Action C']; } ucb1(c = 1.414) { if (this.visits === 0) return Infinity; const exploitation = this.value / this.visits; const exploration = c * Math.sqrt(Math.log(this.parent.visits) / this.visits); return exploitation + exploration; } selectChild() { return this.children.reduce((best, child) => child.ucb1() > best.ucb1() ? child : best ); } expand() { const action = this.untried_actions.pop(); const childState = this.state + ' -> ' + action; const child = new MCTSNode(childState, this); this.children.push(child); return child; } simulate() { // Random rollout from this node let score = Math.random(); // Placeholder evaluation // Simulate random decisions for (let depth = 0; depth < 3; depth++) { score *= Math.random(); } return score; } backpropagate(value) { this.visits++; this.value += value; if (this.parent) { this.parent.backpropagate(value); } } } class MCTS { constructor(rootState, simulations = 100) { this.root = new MCTSNode(rootState); this.simulations = simulations; } runSimulation() { let node = this.root; // Selection while (node.children.length > 0 && node.untried_actions.length === 0) { node = node.selectChild(); } // Expansion if (node.untried_actions.length > 0) { node = node.expand(); } // Simulation const value = node.simulate(); // Backpropagation node.backpropagate(value); } search() { for (let i = 0; i < this.simulations; i++) { this.runSimulation(); if ((i + 1) % 20 === 0) { console.log(\`Completed \${i + 1} simulations\`); } } // Select best action from root const bestChild = this.root.children.reduce((best, child) => child.visits > best.visits ? child : best ); return bestChild.state; } } // Example usage const mcts = new MCTS("Optimize resource allocation", 50); const bestAction = mcts.search(); console.log("Best action:", bestAction); // Display statistics console.log("\\nRoot children statistics:"); mcts.root.children.forEach(child => { console.log(\` "\${child.state}": visits=\${child.visits}, avg_value=\${(child.value/child.visits).toFixed(3)}\`); });`, language: 'javascript' } ] }; export const GRAPH_OF_THOUGHT_PRESET = { name: 'Graph of Thought', description: 'Non-hierarchical connections between thoughts', cells: [ { type: 'markdown', source: `# Graph of Thought Reasoning Graph of Thought allows arbitrary connections between reasoning nodes, enabling complex relationships and feedback loops.` }, { type: 'code', source: `// Graph of Thought Implementation class GraphNode { constructor(id, content, nodeType = 'thought') { this.id = id; this.content = content; this.nodeType = nodeType; // thought, evidence, conclusion this.edges = new Map(); // id -> edge this.strength = 1.0; } } class GraphEdge { constructor(source, target, edgeType = 'relates', weight = 1.0) { this.source = source; this.target = target; this.edgeType = edgeType; // relates, supports, contradicts, leads-to this.weight = weight; } } class GraphOfThought { constructor(problem) { this.nodes = new Map(); this.edges = []; this.nodeIdCounter = 0; // Add root node this.addNode(problem, 'thought'); } addNode(content, nodeType = 'thought') { const id = \`node_\${this.nodeIdCounter++}\`; const node = new GraphNode(id, content, nodeType); this.nodes.set(id, node); return id; } addEdge(sourceId, targetId, edgeType = 'relates', weight = 1.0) { const edge = new GraphEdge(sourceId, targetId, edgeType, weight); this.edges.push(edge); const sourceNode = this.nodes.get(sourceId); if (sourceNode) { sourceNode.edges.set(targetId, edge); } return edge; } findPaths(startId, endId, maxDepth = 5) { const paths = []; const visited = new Set(); const dfs = (nodeId, currentPath, depth) => { if (depth > maxDepth) return; if (nodeId === endId) { paths.push([...currentPath]); return; } visited.add(nodeId); const node = this.nodes.get(nodeId); if (node) { for (const [targetId, edge] of node.edges) { if (!visited.has(targetId)) { currentPath.push(edge); dfs(targetId, currentPath, depth + 1); currentPath.pop(); } } } visited.delete(nodeId); }; dfs(startId, [], 0); return paths; } analyze() { // Calculate node centrality const centrality = new Map(); for (const [id, node] of this.nodes) { const incomingEdges = this.edges.filter(e => e.target === id).length; const outgoingEdges = node.edges.size; centrality.set(id, incomingEdges + outgoingEdges); } // Find most central nodes const sortedNodes = Array.from(centrality.entries()) .sort((a, b) => b[1] - a[1]); return { centralNodes: sortedNodes.slice(0, 3), totalNodes: this.nodes.size, totalEdges: this.edges.length }; } } // Example usage const got = new GraphOfThought("Climate change impacts"); // Add nodes const agricultureId = got.addNode("Impact on agriculture", "evidence"); const economyId = got.addNode("Economic consequences", "thought"); const policyId = got.addNode("Policy interventions needed", "conclusion"); const tempId = got.addNode("Rising temperatures", "evidence"); // Add edges got.addEdge("node_0", agricultureId, "leads-to", 0.9); got.addEdge("node_0", economyId, "leads-to", 0.8); got.addEdge(agricultureId, economyId, "supports", 0.7); got.addEdge(economyId, policyId, "leads-to", 0.9); got.addEdge(tempId, agricultureId, "supports", 0.95); got.addEdge("node_0", tempId, "relates", 1.0); // Analyze graph const analysis = got.analyze(); console.log("Graph Analysis:", analysis); // Find paths const paths = got.findPaths("node_0", policyId); console.log(\`\\nFound \${paths.length} paths from root to policy node\`);`, language: 'javascript' }, { type: 'markdown', source: `## Graph Visualization \`\`\`mermaid graph LR A[Climate Change] --> B[Rising Temperatures] A --> C[Agriculture Impact] A --> D[Economic Impact] B --> C C --> D D --> E[Policy Needed] style A fill:#FFE4B5 style E fill:#90EE90 \`\`\`` } ] }; export const ORCHESTRATION_SUGGEST_PRESET = { name: 'Orchestration Suggest', description: 'Multi-agent coordination for complex tasks', cells: [ { type: 'markdown', source: `# Orchestration Suggestion This notebook demonstrates how to decompose complex tasks and suggest tool combinations for efficient problem-solving.` }, { type: 'code', source: `// Task Orchestration Framework class TaskOrchestrator { constructor(mainTask) { this.mainTask = mainTask; this.subtasks = []; this.toolSuggestions = new Map(); this.dependencies = new Map(); } decomposeTask() { // Analyze task and break into subtasks const keywords = this.extractKeywords(this.mainTask); if (keywords.includes('analyze')) { this.subtasks.push({ id: 'analyze', description: 'Analyze current state', priority: 'high' }); } if (keywords.includes('optimize')) { this.subtasks.push({ id: 'optimize', description: 'Optimize solution', priority: 'medium' }); } if (keywords.includes('implement')) { this.subtasks.push({ id: 'implement', description: 'Implement solution', priority: 'high' }); } if (keywords.includes('test')) { this.subtasks.push({ id: 'test', description: 'Test and validate', priority: 'high' }); } return this.subtasks; } extractKeywords(text) { const keywords = []; const patterns = ['analyze', 'optimize', 'implement', 'test', 'design', 'build']; for (const pattern of patterns) { if (text.toLowerCase().includes(pattern)) { keywords.push(pattern); } } return keywords; } suggestTools() { const toolMap = { 'analyze': ['tree_of_thought', 'graph_of_thought'], 'optimize': ['beam_search', 'mcts'], 'implement': ['sequential_thinking', 'code_execution'], 'test': ['debugging_approach', 'statistical_reasoning'] }; for (const subtask of this.subtasks) { const tools = toolMap[subtask.id] || ['sequential_thinking']; this.toolSuggestions.set(subtask.id, tools); } return this.toolSuggestions; } createDependencies() { // Define dependencies between subtasks const depGraph = { 'implement': ['analyze'], 'test': ['implement'], 'optimize': ['test'] }; for (const [task, deps] of Object.entries(depGraph)) { if (this.subtasks.find(s => s.id === task)) { this.dependencies.set(task, deps.filter(d => this.subtasks.find(s => s.id === d) )); } } return this.dependencies; } generatePlan() { this.decomposeTask(); this.suggestTools(); this.createDependencies(); // Topological sort for execution order const executed = new Set(); const plan = []; const canExecute = (taskId) => { const deps = this.dependencies.get(taskId) || []; return deps.every(d => executed.has(d)); }; while (plan.length < this.subtasks.length) { for (const subtask of this.subtasks) { if (!executed.has(subtask.id) && canExecute(subtask.id)) { plan.push({ ...subtask, tools: this.toolSuggestions.get(subtask.id) }); executed.add(subtask.id); } } } return plan; } } // Example usage const orchestrator = new TaskOrchestrator( "Analyze and optimize the database schema, then implement and test the changes" ); const plan = orchestrator.generatePlan(); console.log("Execution Plan:"); console.log("=============="); plan.forEach((step, index) => { console.log(\`\\nStep \${index + 1}: \${step.description}\`); console.log(\` Priority: \${step.priority}\`); console.log(\` Suggested tools: \${step.tools.join(', ')}\`); const deps = orchestrator.dependencies.get(step.id) || []; if (deps.length > 0) { console.log(\` Dependencies: \${deps.join(', ')}\`); } });`, language: 'javascript' } ] }; export const OODA_LOOP_PRESET = { name: 'OODA Loop Sprint', description: 'Observe-Orient-Decide-Act rapid decision cycles', cells: [ { type: 'markdown', source: `# OODA Loop Sprint Session ## Overview The OODA Loop (Observe-Orient-Decide-Act) is a rapid decision-making framework with automated timing and hypothesis tracking. ## Current Loop - **Phase**: Observe - **Loop #**: 1 - **Target Loop Time**: < 15 minutes ## Phase Checklist - [ ] Collect raw data from environment - [ ] Note anomalies and changes - [ ] Identify emerging patterns - [ ] Document observations without interpretation` }, { type: 'code', source: `// Initialize OODA Loop Session const session = { loopNumber: 1, currentPhase: 'observe', startTime: Date.now(), hypotheses: [], observations: [], decisions: [], actions: [] }; // Phase transition helper function advancePhase() { const phases = ['observe', 'orient', 'decide', 'act']; const currentIndex = phases.indexOf(session.currentPhase); session.currentPhase = phases[(currentIndex + 1) % 4]; if (session.currentPhase === 'observe') { session.loopNumber++; console.log(\`Starting Loop #\${session.loopNumber}\`); } return session.currentPhase; } // Evidence collection function collectEvidence(phase, evidence) { const entry = { phase, evidence, timestamp: new Date().toISOString(), loopNumber: session.loopNumber }; switch(phase) { case 'observe': session.observations.push(entry); break; case 'decide': session.decisions.push(entry); break; case 'act': session.actions.push(entry); break; } return entry; } // Example observation collectEvidence('observe', [ 'System response time increased by 200ms', 'Memory usage at 78%', 'Error rate spike at 14:30' ]); console.log('Current session:', session);`, language: 'javascript' }, { type: 'markdown', source: `## Hypothesis Tracking Track hypotheses across loops to measure learning rate:` }, { type: 'code', source: `// Hypothesis management class Hypothesis { constructor(statement, confidence = 0.5) { this.id = \`hyp-\${Date.now()}\`; this.statement = statement; this.confidence = confidence; this.status = 'proposed'; this.evidence = []; this.loopProposed = session.loopNumber; } addEvidence(evidence, supportive = true) { this.evidence.push({ evidence, supportive, loop: session.loopNumber }); // Adjust confidence based on evidence if (supportive) { this.confidence = Math.min(1, this.confidence + 0.1); } else { this.confidence = Math.max(0, this.confidence - 0.15); } // Update status if (this.confidence > 0.8) { this.status = 'validated'; } else if (this.confidence < 0.2) { this.status = 'invalidated'; } } } // Create and track hypotheses const h1 = new Hypothesis('Database queries are causing the latency', 0.6); const h2 = new Hypothesis('Memory leak in background job', 0.4); session.hypotheses.push(h1, h2); // Test hypothesis with evidence h1.addEvidence('Query logs show 5s+ execution times', true); h1.addEvidence('CPU usage normal during spikes', false); console.log('Hypothesis status:', h1);`, language: 'javascript' }, { type: 'markdown', source: `## Learning Rate Calculation Learning rate = (validated hypotheses / total) × loop efficiency` }, { type: 'code', source: `// Calculate learning metrics function calculateLearningRate() { const validated = session.hypotheses.filter(h => h.status === 'validated').length; const total = session.hypotheses.length; if (total === 0) return 0; const validationRate = validated / total; const loopTime = Date.now() - session.startTime; const targetTime = 15 * 60 * 1000; // 15 minutes const efficiency = Math.min(1, targetTime / loopTime); return { learningRate: validationRate * efficiency, validationRate, efficiency, loopTimeMinutes: Math.round(loopTime / 1000 / 60) }; } const metrics = calculateLearningRate(); console.log('Learning Metrics:', metrics);`, language: 'javascript' } ] }; export const ULYSSES_PROTOCOL_PRESET = { name: 'Ulysses Protocol', description: 'Time-boxed execution with iteration limits and confidence tracking', cells: [ { type: 'markdown', source: `# Ulysses Protocol Session ## Overview Enforces disciplined, time-boxed execution with strict iteration limits to prevent scope creep. ## Constraints - **Time Budget**: 4 hours - **Max Iterations**: 3 - **Min Confidence**: 80% - **Max Scope Drift**: 1 change ## Current Status - **Phase**: Reconnaissance - **Gate Status**: 🔓 Open - **Confidence**: 0% - **Time Remaining**: 4:00:00` }, { type: 'code', source: `// Initialize Ulysses Protocol Session const protocol = { sessionId: \`ulysses-\${Date.now()}\`, startTime: Date.now(), currentPhase: 'reconnaissance', gates: [ { id: 'recon', phase: 'reconnaissance', status: 'open', entryCriteria: [], exitCriteria: [] }, { id: 'plan', phase: 'planning', status: 'locked', entryCriteria: [], exitCriteria: [] }, { id: 'impl', phase: 'implementation', status: 'locked', entryCriteria: [], exitCriteria: [] }, { id: 'validate', phase: 'validation', status: 'locked', entryCriteria: [], exitCriteria: [] }, { id: 'ship_abort', phase: 'ship_or_abort', status: 'locked', entryCriteria: [], exitCriteria: [] } ], constraints: { timeboxMs: 4 * 60 * 60 * 1000, // 4 hours maxIterations: 3, minConfidence: 0.8, maxScopeDrift: 1 }, metrics: { iterations: 0, confidence: 0, scopeDrift: 0, escalations: 0 } }; // Gate management function checkGateCriteria(gateId, evidence) { const gate = protocol.gates.find(g => g.id === gateId); if (!gate) return { passed: false, reason: 'Gate not found' }; // Simple criteria check const requiredEvidence = 3; if (evidence.length < requiredEvidence) { return { passed: false, reason: \`Need \${requiredEvidence} evidence items, have \${evidence.length}\` }; } return { passed: true }; } // Attempt gate passage const evidence = [ 'Requirements documented', 'Stakeholders identified', 'Success criteria defined' ]; const result = checkGateCriteria('recon', evidence); console.log('Gate check result:', result);`, language: 'javascript' }, { type: 'markdown', source: `## Constraint Monitoring & Auto-Escalation` }, { type: 'code', source: `// Constraint violation detection function checkConstraints() { const violations = []; const elapsed = Date.now() - protocol.startTime; // Time constraint if (elapsed > protocol.constraints.timeboxMs) { violations.push({ type: 'time', message: \`Time limit exceeded: \${Math.round(elapsed/1000/60)} min\` }); } // Iteration constraint (only in implementation) if (protocol.currentPhase === 'implementation' && protocol.metrics.iterations > protocol.constraints.maxIterations) { violations.push({ type: 'iteration', message: \`Iteration limit exceeded: \${protocol.metrics.iterations}/\${protocol.constraints.maxIterations}\` }); } // Confidence constraint if (protocol.metrics.confidence < protocol.constraints.minConfidence) { violations.push({ type: 'confidence', message: \`Confidence below threshold: \${(protocol.metrics.confidence*100).toFixed(0)}%\` }); } // Auto-escalation if (violations.length > 0) { protocol.metrics.escalations++; const escalation = { timestamp: new Date().toISOString(), violations, action: violations.some(v => v.type === 'time' || v.type === 'iteration') ? 'abort' : 'reduce_scope' }; console.log('⚠️ AUTO-ESCALATION TRIGGERED:', escalation); return escalation; } return null; } // Simulate some progress protocol.metrics.iterations = 4; protocol.metrics.confidence = 0.65; const escalation = checkConstraints();`, language: 'javascript' }, { type: 'markdown', source: `## Confidence Tracking Track confidence throughout phases to ensure quality decisions:` }, { type: 'code', source: `// Confidence calculation class ConfidenceTracker { constructor() { this.history = []; this.factors = { evidenceQuality: 0, testCoverage: 0, riskMitigation: 0, stakeholderAlignment: 0 }; } updateFactor(factor, value) { this.factors[factor] = Math.max(0, Math.min(1, value)); this.calculate(); } calculate() { // Weighted average of factors const weights = { evidenceQuality: 0.3, testCoverage: 0.3, riskMitigation: 0.2, stakeholderAlignment: 0.2 }; let confidence = 0; for (const [factor, weight] of Object.entries(weights)) { confidence += this.factors[factor] * weight; } // Apply iteration penalty if (protocol.metrics.iterations > protocol.constraints.maxIterations) { const penalty = 0.1 * (protocol.metrics.iterations - protocol.constraints.maxIterations); confidence = Math.max(0, confidence - penalty); } protocol.metrics.confidence = confidence; this.history.push({ timestamp: new Date().toISOString(), confidence, phase: protocol.currentPhase }); return confidence; } getRecommendation() { const conf = protocol.metrics.confidence; if (conf >= protocol.constraints.minConfidence) { return '✅ Confidence threshold met - ready to ship'; } else if (conf < 0.5) { return '❌ Low confidence - consider abort or major pivot'; } else { return '⚠️ Moderate confidence - address key risks before proceeding'; } } } const tracker = new ConfidenceTracker(); tracker.updateFactor('evidenceQuality', 0.8); tracker.updateFactor('testCoverage', 0.6); tracker.updateFactor('riskMitigation', 0.7); tracker.updateFactor('stakeholderAlignment', 0.9); console.log('Current confidence:', (protocol.metrics.confidence * 100).toFixed(1) + '%'); console.log('Recommendation:', tracker.getRecommendation());`, language: 'javascript' }, { type: 'markdown', source: `## Ship/Abort Decision Matrix | Metric | Status | Recommendation | |--------|--------|----------------| | Confidence | -- | Based on threshold | | Iterations | -- | Based on limit | | Time | ⏰ | Check remaining | | Scope | -- | Based on drift | Update the table dynamically based on your protocol.metrics values.` } ] }; // Export all presets export const NOTEBOOK_PRESETS = { tree_of_thought: TREE_OF_THOUGHT_PRESET, beam_search: BEAM_SEARCH_PRESET, mcts: MCTS_PRESET, graph_of_thought: GRAPH_OF_THOUGHT_PRESET, orchestration_suggest: ORCHESTRATION_SUGGEST_PRESET, ooda_loop: OODA_LOOP_PRESET, ulysses_protocol: ULYSSES_PROTOCOL_PRESET }; export function getPresetForPattern(pattern) { return NOTEBOOK_PRESETS[pattern]; }