UNPKG

universal-ai-brain

Version:

🧠 UNIVERSAL AI BRAIN 3.3 - The world's most advanced cognitive architecture with 24 specialized systems, MongoDB 8.1 $rankFusion hybrid search, latest Voyage 3.5 embeddings, and framework-agnostic design. Works with Mastra, Vercel AI, LangChain, OpenAI A

229 lines (208 loc) • 8.12 kB
import { Agent } from '@mastra/core/agent'; import { openai } from '@ai-sdk/openai'; import { createTool } from '@mastra/core/tools'; import { z } from 'zod'; import { MongoClient } from 'mongodb'; // Load environment variables import dotenv from 'dotenv'; dotenv.config(); // MongoDB connection for real data operations const mongoClient = new MongoClient(process.env.MONGODB_URI!); // Tool for writing real test data to MongoDB const writeTestDataTool = createTool({ id: 'write-test-data', description: 'Writes real test data to MongoDB for cognitive system validation', inputSchema: z.object({ collection: z.string().describe('MongoDB collection name'), data: z.any().describe('Data to write to MongoDB'), testType: z.string().describe('Type of cognitive test being performed') }), execute: async ({ context }) => { try { await mongoClient.connect(); const db = mongoClient.db(process.env.TEST_DATABASE_NAME); const collection = db.collection(`${process.env.TEST_COLLECTION_PREFIX}${context.collection}`); const document = { ...context.data, testType: context.testType, timestamp: new Date(), testId: `test_${Date.now()}_${Math.random().toString(36).substr(2, 9)}` }; const result = await collection.insertOne(document); return { success: true, insertedId: result.insertedId.toString(), document: document, message: `Successfully wrote ${context.testType} test data to ${context.collection}` }; } catch (error) { return { success: false, error: error.message, message: `Failed to write test data: ${error.message}` }; } } }); // Tool for reading and analyzing test data from MongoDB const analyzeTestDataTool = createTool({ id: 'analyze-test-data', description: 'Reads and analyzes test data from MongoDB to validate cognitive systems', inputSchema: z.object({ collection: z.string().describe('MongoDB collection name'), testType: z.string().describe('Type of cognitive test to analyze'), query: z.any().optional().describe('Optional MongoDB query filter') }), execute: async ({ context }) => { try { await mongoClient.connect(); const db = mongoClient.db(process.env.TEST_DATABASE_NAME); const collection = db.collection(`${process.env.TEST_COLLECTION_PREFIX}${context.collection}`); const query = { testType: context.testType, ...context.query }; const documents = await collection.find(query).sort({ timestamp: -1 }).limit(10).toArray(); return { success: true, documentsFound: documents.length, documents: documents, analysis: { latestTest: documents[0], totalTests: documents.length, timeRange: documents.length > 0 ? { latest: documents[0].timestamp, oldest: documents[documents.length - 1]?.timestamp } : null } }; } catch (error) { return { success: false, error: error.message, message: `Failed to analyze test data: ${error.message}` }; } } }); // Tool for hybrid search testing (MongoDB $rankFusion) const hybridSearchTool = createTool({ id: 'hybrid-search-test', description: 'Tests MongoDB hybrid search with $rankFusion using real data', inputSchema: z.object({ collection: z.string().describe('MongoDB collection name'), textQuery: z.string().describe('Text search query'), vectorQuery: z.array(z.number()).optional().describe('Vector search query (embeddings)') }), execute: async ({ context }) => { try { await mongoClient.connect(); const db = mongoClient.db(process.env.TEST_DATABASE_NAME); const collection = db.collection(`${process.env.TEST_COLLECTION_PREFIX}${context.collection}`); // MongoDB $rankFusion hybrid search pipeline const pipeline = [ { $rankFusion: { input: { pipelines: [ // Text search pipeline [ { $search: { index: "default", text: { query: context.textQuery, path: ["content", "description", "title"] } } } ], // Vector search pipeline (if vector provided) ...(context.vectorQuery ? [[ { $vectorSearch: { index: "vector_index", path: "embedding", queryVector: context.vectorQuery, numCandidates: 100, limit: 20 } } ]] : []) ] } } }, { $limit: 10 }, { $project: { _id: 1, content: 1, score: { $meta: "searchScore" } } } ]; const results = await collection.aggregate(pipeline).toArray(); return { success: true, query: context.textQuery, resultsCount: results.length, results: results, hybridSearchWorking: results.length > 0, message: `Hybrid search found ${results.length} results for "${context.textQuery}"` }; } catch (error) { return { success: false, error: error.message, message: `Hybrid search failed: ${error.message}` }; } } }); export const cognitiveTestAgent = new Agent({ name: 'Cognitive Systems Test Agent', description: 'Agent for testing all 24 cognitive systems with real MongoDB data', instructions: ` You are a specialized testing agent for Universal AI Brain 3.0's cognitive systems. Your mission is to test each of the 24 cognitive systems using REAL DATA: TESTING PROTOCOL: 1. ALWAYS write real test data to MongoDB first using writeTestDataTool 2. IMMEDIATELY fetch and analyze the data using analyzeTestDataTool 3. Test the specific cognitive system with the real data 4. Validate the system's performance and behavior 5. Document results with concrete evidence NEVER use mock data - only real data written to and read from MongoDB Atlas. The 24 cognitive systems to test are: 1. Working Memory - Active information processing 2. Episodic Memory - Personal experiences and events 3. Semantic Memory - Facts and knowledge 4. Memory Decay - Forgetting mechanisms 5. Analogical Mapping - Finding similarities 6. Causal Reasoning - Cause and effect relationships 7. Attention Management - Focus and filtering 8. Confidence Tracking - Uncertainty quantification 9. Emotional Intelligence - Emotion recognition and response 10. Social Intelligence - Social dynamics understanding 11. Cultural Knowledge - Cultural awareness and adaptation 12. Goal Hierarchy - Goal decomposition and management 13. Temporal Planning - Time-based planning 14. Skill Capability Management - Skill assessment and development 15. Human Feedback Integration - Learning from feedback 16. Self-Improvement - Continuous learning 17. Safety Guardrails - Ethical and safety constraints 18. Multi-Modal Processing - Handling different data types 19. Tool Interface - External tool integration 20. Workflow Orchestration - Process management 21. Vector Search - Semantic similarity search 22. Hybrid Search - Combined text and vector search 23. Context Injection - Dynamic context enhancement 24. Real-time Monitoring - System performance tracking For each test, provide: - Real data written to MongoDB - Analysis of the retrieved data - Cognitive system performance metrics - Validation results - Recommendations for improvements `, model: openai('gpt-4o'), tools: { writeTestDataTool, analyzeTestDataTool, hybridSearchTool } });