Persona MCP

# Context Detection - Smart Expert Matching ## How Context Detection Works ### Overview The system analyzes conversation context to suggest the most appropriate expert(s) for your needs. This document explains the detection logic and how to get the best expert matches. --- ## Keyword-Based Detection ### Engineering Keywords ``` FULLSTACK DEVELOPER (101) Triggers: fullstack, full-stack, end-to-end, MERN, MEAN, web app Context: Building complete applications REACT EXPERT (102) Triggers: React, Redux, Next.js, hooks, components, JSX, virtual DOM Context: Frontend development with React ecosystem DEVOPS ENGINEER (108/109) Triggers: CI/CD, pipeline, Jenkins, GitHub Actions, deployment, Docker, Kubernetes Context: Infrastructure and deployment automation SECURITY EXPERT (126/127) Triggers: security, OWASP, vulnerability, penetration, authentication, encryption Context: Application and infrastructure security SYSTEMS ARCHITECT (128/129) Triggers: architecture, scalability, microservices, design patterns, distributed Context: System design and technical planning ``` ### Business Keywords ``` STRATEGY CONSULTANT (303) Triggers: strategy, competitive analysis, market entry, positioning, strategic Context: Business strategy and planning CFO ADVISOR (312) Triggers: financial, budget, ROI, revenue, P&L, funding, valuation, cash flow Context: Financial decisions and analysis PRODUCT MANAGER (306) Triggers: roadmap, prioritization, feature, user story, backlog, sprint Context: Product development and management MARKETING DIRECTOR (314) Triggers: marketing, campaign, brand, SEO, content, leads, conversion Context: Marketing strategy and execution SALES COACH (313) Triggers: sales, pipeline, quota, closing, negotiation, objection handling Context: Sales strategy and techniques ``` ### AI/ML Keywords ``` LLM ENGINEER (410) Triggers: LLM, GPT, Claude, prompt engineering, fine-tuning, RAG Context: Large language model development DATA SCIENTIST (401) Triggers: machine learning, model, prediction, classification, regression Context: ML model development and analysis AI ETHICS (404) Triggers: AI ethics, bias, fairness, responsible AI, governance Context: Ethical AI development and deployment AI AGENT DEVELOPER (411) Triggers: agent, autonomous, LangChain, tool use, multi-agent Context: AI agent development ``` --- ## Intent-Based Detection ### Problem-Solving Intents ``` DEBUGGING/TROUBLESHOOTING "I have an error..." → Relevant domain expert "Something is broken..." → Systems + domain expert "Not working as expected..." → QA + domain expert PLANNING/DESIGN "How should I architect..." → Systems Architect (128) "What's the best approach..." → Strategy Consultant (303) "Help me plan..." → Product Manager (306) REVIEW/OPTIMIZATION "Can you review my code..." → Security Expert (126) + domain "How can I improve..." → Relevant domain expert "Optimize this..." → Performance-focused expert LEARNING/EXPLANATION "Explain how..." → Educator (600s) or domain expert "What is..." → Relevant domain expert "Why does..." → Domain expert + theory focus ``` ### Action-Based Intents ``` BUILD/CREATE "Build a..." → Engineering experts "Create a..." → Design or Engineering "Develop..." → Engineering experts ANALYZE/EVALUATE "Analyze this..." → Data Scientist (401) or domain "Evaluate..." → CFO (312) or domain "Compare..." → Strategy (303) or domain DECIDE/CHOOSE "Should I..." → Strategy (303) or CFO (312) "Which option..." → Domain expert "Help me decide..." → Multiple experts recommended COMMUNICATE/PRESENT "Write a..." → Creative Writer (206) or domain "Present to..." → Leadership Coach (801) "Pitch..." → Sales Coach (313) ``` --- ## Domain Detection Patterns ### Technology Domain ``` FRONTEND Patterns: UI, interface, React, Vue, CSS, responsive, browser Expert: React Expert (102), Vue Specialist (103), UX Designer (201) BACKEND Patterns: API, server, database, endpoint, REST, GraphQL Expert: Backend expertise, API Architect (117) INFRASTRUCTURE Patterns: deploy, server, cloud, AWS, Azure, GCP, scaling Expert: DevOps (108), Systems Architect (128) DATA Patterns: database, SQL, analytics, warehouse, pipeline Expert: Data Engineer (106), Database Expert (112) ``` ### Business Domain ``` FINANCE Patterns: budget, revenue, cost, investment, ROI, profit Expert: CFO Advisor (312), Accountant (318) MARKETING Patterns: brand, campaign, leads, conversion, SEO, content Expert: Marketing Director (314), Brand Strategist (321) OPERATIONS Patterns: process, efficiency, workflow, optimization, lean Expert: Operations Manager (316) PEOPLE Patterns: hiring, culture, team, performance, feedback Expert: HR Consultant (315), Leadership Coach (801) ``` --- ## Multi-Expert Scenarios ### When Multiple Experts Are Recommended ``` SCENARIO: "Help me launch a new product" DETECTED NEEDS: ├── Product strategy → Product Manager (306) ├── Technical implementation → Systems Architect (128) ├── Market positioning → Marketing Director (314) ├── Financial planning → CFO Advisor (312) └── Legal review → Legal Advisor (901) RECOMMENDATION: Primary: Product Manager (306) Chain: PM → Architect → Marketing → CFO ``` ``` SCENARIO: "Review my startup pitch deck" DETECTED NEEDS: ├── Content strategy → Content Strategist (218) ├── Financial model → CFO Advisor (312) ├── Investor perspective → Venture Capitalist (317) └── Presentation → Business Storyteller (333) RECOMMENDATION: Primary: Venture Capitalist (317) Support: CFO Advisor (312) ``` ``` SCENARIO: "My API is slow and unreliable" DETECTED NEEDS: ├── Performance analysis → Systems Architect (128) ├── Database optimization → Database Expert (112) ├── Caching strategy → Redis Expert (121) └── Monitoring → DevOps Engineer (108) RECOMMENDATION: Primary: Systems Architect (128) Chain: Architect → Database → DevOps ``` --- ## Confidence Scoring ### How Matches Are Scored ``` SCORING FACTORS: KEYWORD MATCH (40%) - Exact keyword matches - Related term matches - Domain terminology INTENT MATCH (30%) - Action type alignment - Problem type alignment - Goal alignment CONTEXT FIT (20%) - Industry relevance - Complexity level - Scope appropriateness HISTORY (10%) - Previous successful matches - User preferences - Conversation continuity SCORE INTERPRETATION: 90-100%: Perfect match, high confidence 70-89%: Good match, likely appropriate 50-69%: Possible match, consider alternatives <50%: Low confidence, ask for clarification ``` --- ## Manual Override ### How to Specify Experts Directly ``` DIRECT CALL BY NAME: "@security-expert" → Security Expert (126) "Call the CFO" → CFO Advisor (312) "I need the strategy consultant" → Strategy Consultant (303) DIRECT CALL BY ID: "@128" → Systems Architect "Expert 312" → CFO Advisor "Use persona 410" → LLM Engineer CATEGORY CALL: "Engineering expert" → Shows 100s options "Business expert" → Shows 300s options "Any AI expert" → Shows 400s options CHAIN REQUEST: "Chain: architect → security → devops" "First product manager, then marketing" "Sequential: 303, 312, 901" ``` --- ## Improving Detection Accuracy ### Tips for Better Matches ``` BE SPECIFIC: ✗ "Help with my code" ✓ "Help optimize my React component's rendering performance" INCLUDE CONTEXT: ✗ "Review this" ✓ "Review this API design for a high-traffic e-commerce platform" STATE YOUR GOAL: ✗ "I have a question about marketing" ✓ "I need to create a go-to-market strategy for a B2B SaaS product" MENTION CONSTRAINTS: ✗ "Build me an app" ✓ "Build a mobile app with limited budget and 3-month timeline" ``` ### Feedback Loop ``` CORRECT MISMATCHES: "Actually, I need more of a [X] perspective" "Can you switch to the [Y] expert?" "This is more of a [Z] question" The system learns from: - Explicit corrections - Follow-up expert changes - Conversation outcomes ```