Longevity Priority Navigator

Live Demo: https://neuroidss.github.io/Longevity-Priority-Navigator/

This project is a prototype for an AI-powered system designed to navigate and prioritize research in the science of longevity. Its mission is to move beyond simple data retrieval and create a tool for knowledge synthesis and strategic insight.

---

Core Philosophy: From Noise to Signal

The guiding principle of this project is to perfectly execute the most valuable function for a longevity researcher: cutting through the noise to find the signal. Instead of just presenting data, this tool uses AI agents to generate high-priority, justifiable research directions based on the latest scientific context. The entire user experience is built to serve this singular goal.

Philosophical Underpinnings

This tool is built on the belief that genius is the ability to discern what is important and dedicate one's time to it. In the complex field of longevity, identifying the most critical questions is the biggest challenge. The field faces a "chicken-and-egg" problem: we need effective interventions to find good biomarkers, but we need good biomarkers to validate new interventions.

This app aims to be a tool to break that cycle. It is inspired by a desire to augment our own intelligence and overcome cognitive limitations, guided by ideas from thinkers like Neil deGrasse Tyson and Aldous Huxley.

"I often wonder about the questions we don't yet know how to ask... because the questions we know how to ask, I'm not so interested in them." - Neil deGrasse Tyson

"The victim of mind-manipulation does not know that he is a victim. To him the walls of his prison are invisible, and he believes himself to be free." - Aldous Huxley

Our goal is to create a tool that helps break free from existing paradigms and see the "invisible walls" of our current knowledge, enabling us to ask new, more powerful questions.

How It Works: A User's Journey

1. Define Focus: The user enters a broad research area (e.g., "senolytics," "epigenetic clocks").
2. Select Lens & Dispatch: They choose an analytical perspective (e.g., "High-Risk/High-Reward", "Clinical Translation") and dispatch one of two specialized AI agents.
3. Receive Briefing: In seconds, the user receives an AI-generated strategic briefing. This isn't a list of links; it's a synthesis that includes:
   • A single, profound Key Strategic Question.
   • A handful of prioritized research opportunities, complete with justifications, potential impact, and confidence scores.
   • An analysis of contradictions and synergies found in the source material.
4. Explore Visually: An interactive Knowledge Graph is automatically generated, visually mapping the concepts and relationships discussed in the briefing. Hovering over a research opportunity highlights its context in the graph.
5. Converse for Depth: The user can then "chat" with an AI assistant that uses the entire workspace (briefing, graph, sources) as its context to answer follow-up questions, elaborate on connections, and help refine new hypotheses.

Key Features

• AI-Driven Prioritization: The core of the app. Agents don't just summarize; they analyze, critique, and propose concrete, high-impact research directions.
• Dual-Agent System:
  • Knowledge Navigator: Analyzes the current state of a field.
  • Trend Analyzer: Analyzes the evolution of a field over time, identifying emerging and fading concepts.
• Insight-First UI: The AI's proposed research opportunities and key questions are the primary output, not an afterthought. The graph and sources provide interactive, explorable context.
• Interactive Knowledge Graph: A custom-built SVG physics simulation visualizes the conceptual landscape, making complex relationships intuitive.
• Context-Aware Chat: The chat assistant maintains the context of the entire analysis, allowing for deep, meaningful exploration of the generated insights.
• Persistent Workspaces: All generated analyses and chats are saved in the browser, allowing researchers to pick up where they left off.

The Agentic Engine: How AI Delivers Insight

The "brain" of the application is a set of carefully engineered system prompts that define two distinct AI agent roles, powered by the Google Gemini API and local Ollama models.

• System Prompts (services/agentPrompts.ts): This is where the agent's "personality," goals, and constraints are defined. We instruct the model to act as a world-class research strategist, to identify the "chicken-and-egg" problem, and to structure its output in a specific, detailed JSON format.
• Dynamic Agents (AgentType):
  • KnowledgeNavigator: Scans the current landscape to identify the state-of-the-art, key players, and immediate opportunities.
  • TrendAnalyzer: Takes a historical view, identifying which concepts are gaining momentum and which are fading, providing a strategic "where-to-next" perspective.
• Two-Stage Search & Validation Engine: The system employs a sophisticated two-stage process to ensure the analytical foundation is built on high-quality scientific evidence.
  1. Stage 1: Broad Discovery & Filtering: The system performs parallel searches across multiple providers (PubMed, Google Patents, web search, preprint archives). For noisy sources like live preprint feeds or general web search, a specialized AI filter is used to identify relevant documents before they proceed.
  2. Stage 2: AI-Powered Validation & Enrichment: The combined, filtered results are then passed to a second AI agent. This "validation agent" acts as an expert scientific librarian. It enriches metadata, assesses the reliability of each source, and generates a concise summary of its core claims. This crucial step ensures that the final analysis is grounded in verifiable scientific content, not just popular interpretations.

Technical Architecture

The application is a modern, client-side web app built with React and TypeScript.

• Frontend: Built with React and Vite. The UI is crafted with Tailwind CSS for a clean, responsive, and professional design.
• AI Service Layer (services/geminiService.ts): This is the core communication hub. It uses the @google/genai SDK and local fetch to interact with both Gemini and Ollama APIs. It handles responses, parses structured JSON, and manages the entire agentic workflow.
• State Management: Application state is managed through a set of cohesive custom React hooks (useWorkspaceManager, useAppSettings, useDebugLog) and persisted in localStorage for seamless session continuity. This keeps component logic clean and state management predictable.
• Visualization (components/KnowledgeGraphView.tsx): The knowledge graph is rendered using SVG and a custom, lightweight physics simulation engine. This provides a dynamic, interactive visualization without heavy library dependencies, offering full control over node behavior and styling.

Strategic Roadmap: From Research to Action

This project has a clear path to evolve from a powerful research tool into a comprehensive platform for both scientific discovery and personal/investment action.

Phase 1: Immediate Application: Leverage the Existing "Innovation Agent"

The InnovationAgent is specifically designed to bridge the gap between science and commerce.

• Action: Use the "Analyze Innovation" function on topics that have already shown promising human trial results (e.g., "SGLT2 inhibitors and aging," "GLP-1 agonists and all-cause mortality").
• Expected Output: The AI will generate a MarketInnovationAnalysis, providing:
  • Initial ideas for therapeutics or diagnostics.
  • A list of active companies and research groups (initial investment targets).
  • An understanding of the risks and timeline for potential products.
• Benefit: This provides an immediate, AI-vetted list of potential investment areas, directly answering the question of "where to spend the money."

Phase 2: Feature Extension: The "Applied Longevity" Agent

The next step is to create a new, specialized agent focused explicitly on consumer-facing products and publicly traded companies.

• Action: Develop a new agent type, AppliedLongevityAgent, with a system prompt to identify:
  1. Supplements & Nutraceuticals: Compounds with evidence for impacting longevity pathways.
  2. Publicly Traded Companies: Companies (with ticker symbols) related to the research opportunities.
  3. Investment Thesis: A concise, AI-formulated investment thesis for promising areas.
• Benefit: This directly creates the "shopping list" of both supplements and stocks, making the insights tangible for a broader audience.

Phase 3: UI/UX Development: The Personalized "Longevity Portfolio"

This phase translates the AI's output into a user-centric dashboard.

• Action: Design and build a new "Portfolio" tab in the UI.
• Features:
  • Health & Wellness Wing: A "Longevity Shopping Cart" listing supplements with evidence levels and links.
  • Financial Wing: An "Investment Portfolio" listing publicly traded companies with the AI's investment thesis.
• Benefit: This creates a powerful, dual-purpose interface for users to "invest" in their personal health and financial future, based on the same underlying scientific analysis.

Phase 4: Full Integration: Connecting to Live Data Streams

This final phase makes the dashboard dynamic and truly actionable.

• Action: Integrate with third-party APIs:
  • Financial APIs (e.g., Finnhub, Alpha Vantage): For real-time stock prices and data.
  • E-commerce/Affiliate APIs: To show pricing and availability for supplements.
• Benefit: This transforms the application from a static analysis tool into a live, personalized decision-support system for longevity.

Additional Future Directions

• Agent Collaboration: Introduce a "Critic" agent that reviews the output of the primary agent, challenging its assumptions and improving the robustness of the final recommendations.
• Deeper Integration: Directly integrate with APIs from PubMed, bioRxiv, and patent offices for more structured and reliable data sourcing.
• Long-Term Memory: Implement vector-based memory for the chat assistant, allowing it to recall information across different research sessions.
• User Accounts: Allow users to create accounts to save, organize, and compare multiple research workspaces.

How to Run Locally

Prerequisites: Node.js

1. Clone the repository.
2. Install dependencies:

   npm install

3. Set up your environment (Optional, but recommended): To use the Google Gemini API without entering the key in the browser, create a .env file in the root directory and add your API key:

   VITE_API_KEY=your_google_ai_api_key_here
   

   If this is not set, you will be prompted to enter your API key in the application's settings panel.
4. Run the development server:

   npm run dev

This will start the development server, and you can view the application in your browser at the local URL provided in your terminal (usually http://localhost:5173 or similar).