Everyone wishes they could go back and buy Bitcoin at $1. Polyseer brings the future to you, so you never have to wonder "what if?" again.
NOT FINANCIAL ADVICE | Polyseer provides analysis for entertainment and research purposes only. Always DYOR.
git clone https://github.com/yorkeccak/polyseer.git
cd polyseer
npm install
# Create .env.local with:
# OPENAI_API_KEY=sk-... # Get from platform.openai.com
# + Valyu OAuth credentials (see below)
npm run devOpen localhost:3000, sign in with Valyu, paste any Polymarket or Kalshi URL, and get your analysis.
Or, we have a hosted version here
Prediction markets tell you what might happen. Polyseer tells you why.
Drop in any Polymarket or Kalshi URL and get a structured analysis that breaks down the actual factors driving an outcome. Instead of gut feelings or surface-level takes, you get systematic research across academic papers, news, market data, and expert analysis.
The system uses multiple AI agents to research both sides of a question, then aggregates the evidence using Bayesian probability math. Think of it as having a research team that can read thousands of sources in minutes and give you the key insights.
Core features:
- Systematic research across academic, web, and market data sources
- Evidence classification and quality scoring
- Mathematical probability aggregation (not just vibes)
- Both sides research to avoid confirmation bias
- Real-time data, not stale information
Built for developers, researchers, and anyone who wants rigorous analysis instead of speculation.
Polyseer is built on a multi-agent AI architecture that orchestrates specialized agents to conduct deep analysis. Here's how it works:
graph TD
A[User Input: Market URL] --> B[Platform Detector]
B --> C{Polymarket or Kalshi?}
C -->|Polymarket| D[Polymarket API Client]
C -->|Kalshi| E[Kalshi API Client]
D --> F[Unified Market Data]
E --> F
B --> C[Orchestrator]
C --> D[Planner Agent]
D --> E[Research Agents]
E --> F[Valyu Search Network]
F --> G[Evidence Collection]
G --> H[Critic Agent]
H --> I[Analyst Agent]
I --> J[Reporter Agent]
J --> K[Final Verdict]
style A fill:#e1f5fe
style K fill:#c8e6c9
style F fill:#fff3e0
style C fill:#f3e5f5
sequenceDiagram
participant User
participant Orch as Orchestrator
participant Plan as Planner
participant Res as Researcher
participant Valyu as Valyu Network
participant Critic
participant Analyst
participant Reporter
User->>Orch: Polymarket URL
Orch->>Plan: Generate research strategy
Plan->>Orch: Subclaims + search seeds
par Research Cycle 1
Orch->>Res: Research PRO evidence
Res->>Valyu: Deep + Web searches
Valyu-->>Res: Academic papers, news, data
and
Orch->>Res: Research CON evidence
Res->>Valyu: Targeted counter-searches
Valyu-->>Res: Contradicting evidence
end
Orch->>Critic: Analyze evidence gaps
Critic->>Orch: Follow-up search recommendations
par Research Cycle 2 (if gaps found)
Orch->>Res: Targeted follow-up searches
Res->>Valyu: Fill identified gaps
Valyu-->>Res: Missing evidence
end
Orch->>Analyst: Bayesian probability aggregation
Analyst->>Orch: pNeutral, pAware, evidence weights
Orch->>Reporter: Generate final report
Reporter->>User: Analyst-grade verdict
Polyseer uses Sign in with Valyu for authentication and search API access. Valyu is the information backbone of Polyseer, providing access to:
- Academic Papers: Real-time research publications
- Web Intelligence: Fresh news and analysis
- Market Data: Financial and trading information
- Proprietary Datasets: Exclusive Valyu intelligence
All API costs are charged to your Valyu organization credits via OAuth proxy. New accounts get $10 in free credits.
graph LR
A[Research Query] --> B[Valyu Deep Search]
B --> C[Academic Sources]
B --> D[Web Sources]
B --> E[Market Data]
B --> F[Proprietary Intel]
C --> G[Evidence Classification]
D --> G
E --> G
F --> G
G --> H[Type A: Primary Sources]
G --> I[Type B: High-Quality Secondary]
G --> J[Type C: Standard Secondary]
G --> K[Type D: Weak/Speculative]
style B fill:#fff3e0
style H fill:#c8e6c9
style I fill:#dcedc8
style J fill:#f0f4c3
style K fill:#ffcdd2
Each piece of evidence is rigorously classified:
| Type | Description | Cap | Examples |
|---|---|---|---|
| A | Primary Sources | 2.0 | Official documents, press releases, regulatory filings |
| B | High-Quality Secondary | 1.6 | Reuters, Bloomberg, WSJ, expert analysis |
| C | Standard Secondary | 0.8 | Reputable news with citations, industry publications |
| D | Weak/Speculative | 0.3 | Social media, unverified claims, rumors |
Polyseer uses sophisticated mathematical models to combine evidence:
graph TD
A[Prior Probability p0] --> B[Evidence Weights]
B --> C[Log Likelihood Ratios]
C --> D[Correlation Adjustments]
D --> E[Cluster Analysis]
E --> F[Final Probabilities]
F --> G[pNeutral: Objective Assessment]
F --> H[pAware: Market-Informed]
style A fill:#e3f2fd
style F fill:#c8e6c9
style G fill:#dcedc8
style H fill:#f0f4c3
Key Formulas:
- Log Likelihood Ratio:
LLR = log(P(evidence|YES) / P(evidence|NO)) - Probability Update:
p_new = p_old × exp(LLR) - Correlation Adjustment: Accounts for evidence clustering and dependencies
Each piece of evidence receives an influence score based on:
- Verifiability: Can the claim be independently verified?
- Consistency: Internal logical coherence
- Independence: Number of independent corroborations
- Recency: How fresh is the information?
- Next.js 15.5 - React framework with Turbopack
- Tailwind CSS 4 - Utility-first styling
- Framer Motion - Smooth animations
- Radix UI - Accessible components
- React 19 - Latest React features
- AI SDK - LLM orchestration
- GPT-5 - Advanced reasoning model
- Valyu OAuth - Authentication & search API access
- Polymarket API - Market data fetching
- Kalshi API - Market data fetching
- Supabase - Database and session management
- Zustand - Simple state management
- TanStack Query - Server state synchronization
- Supabase SSR - Server-side authentication
- TypeScript - Type safety throughout
- Zod - Runtime type validation
- ESLint - Code quality
- Vercel - Deployment platform
- Node.js 18+
- npm/pnpm/yarn
- OpenAI API key - For GPT-5 access
- Valyu OAuth credentials - Get from platform.valyu.ai
- Supabase account - For database and session management
git clone https://github.com/your-org/polyseer.git
cd polyseernpm install
# or
pnpm installCreate .env.local with the following variables:
# ===========================================
# App Configuration
# ===========================================
NEXT_PUBLIC_APP_MODE=development # Set to 'production' for production deployment
NEXT_PUBLIC_APP_URL=http://localhost:3000
# ===========================================
# Valyu OAuth Configuration (Required)
# ===========================================
# Get these from Valyu Platform Dashboard: https://platform.valyu.ai
# Settings -> OAuth Apps -> Create new OAuth App
NEXT_PUBLIC_VALYU_SUPABASE_URL=https://xxx.supabase.co # Valyu Platform's Supabase URL
NEXT_PUBLIC_VALYU_CLIENT_ID=your-oauth-client-id
VALYU_CLIENT_SECRET=your-oauth-client-secret # Server-only! Do not expose to client
VALYU_APP_URL=https://platform.valyu.ai # Valyu Platform URL for userinfo endpoint
# ===========================================
# App's Own Supabase (Required)
# ===========================================
# Your app's Supabase project for user data, chat history, etc.
NEXT_PUBLIC_SUPABASE_URL=https://your-app.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=your-supabase-anon-key
SUPABASE_SERVICE_ROLE_KEY=your-service-role-key # Server-only!
# ===========================================
# OpenAI Configuration (Required)
# ===========================================
OPENAI_API_KEY=your-openai-api-key
# ===========================================
# Optional Services
# ===========================================
# Weaviate Memory (optional)
MEMORY_ENABLED=false
WEAVIATE_HOST=your-weaviate-host
WEAVIATE_API_KEY=your-weaviate-api-key
# Kalshi Integration (optional)
KALSHI_API_KEY=your-kalshi-api-key
# Groq (optional - for faster inference)
GROQ_API_KEY=your-groq-api-keySet up your Supabase database with the following table:
-- Users table
CREATE TABLE users (
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
email TEXT UNIQUE NOT NULL,
full_name TEXT,
avatar_url TEXT,
created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
updated_at TIMESTAMP WITH TIME ZONE DEFAULT NOW()
);
-- Analysis sessions
CREATE TABLE analysis_sessions (
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
user_id UUID REFERENCES users(id),
market_url TEXT NOT NULL,
market_question TEXT,
status TEXT DEFAULT 'pending',
started_at TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
completed_at TIMESTAMP WITH TIME ZONE,
duration_seconds INTEGER,
valyu_cost DECIMAL(10,6),
analysis_steps JSONB,
forecast_card JSONB,
markdown_report TEXT,
current_step TEXT,
progress_events JSONB,
p0 DECIMAL(5,4),
p_neutral DECIMAL(5,4),
p_aware DECIMAL(5,4),
drivers TEXT[],
error_message TEXT,
created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
updated_at TIMESTAMP WITH TIME ZONE DEFAULT NOW()
);npm run devOpen http://localhost:3000, sign in with Valyu, and start analyzing.
Purpose: Break down complex questions into research pathways Input: Market question Output: Subclaims, search seeds, key variables, decision criteria
interface Plan {
subclaims: string[]; // Causal pathways to outcome
keyVariables: string[]; // Leading indicators to monitor
searchSeeds: string[]; // Targeted search queries
decisionCriteria: string[]; // Evidence evaluation criteria
}Purpose: Gather evidence from multiple sources Tools: Valyu Deep Search, Valyu Web Search Process:
- Initial bilateral research (PRO/CON)
- Evidence classification (A/B/C/D)
- Follow-up targeted searches
Purpose: Identify gaps and provide quality feedback Analysis:
- Missing evidence areas
- Duplication detection
- Data quality concerns
- Correlation adjustments
- Follow-up search recommendations
Purpose: Mathematical probability aggregation Methods:
- Bayesian updating
- Evidence clustering
- Correlation adjustments
- Log-likelihood calculations
Purpose: Generate human-readable analysis Output: Markdown report with:
- Executive summary
- Evidence synthesis
- Risk factors
- Confidence assessment
- End-to-end encryption for sensitive data
- Secure session management via Supabase
- Input sanitization for all user data
- No personal data stored in search queries
- OAuth 2.1 with PKCE for Valyu authentication
- CORS policies for secure cross-origin requests
- Request validation using Zod schemas
- Audit logging for all API calls
We welcome contributions! Here's how to get started:
- Fork the repository
- Create a feature branch:
git checkout -b feature/amazing-feature - Make your changes
- Add tests:
npm run test - Submit a pull request
- TypeScript: Strict mode enabled
- ESLint: Follow the configuration
- Prettier: Auto-formatting on save
- Conventional Commits: Use semantic commit messages
- Turbopack for fast development builds
- Edge runtime for serverless function optimization
- Code splitting for minimal bundle sizes
- Smart caching for repeated queries
- Real-time metrics via Vercel Analytics
- Error tracking with detailed logging
- Performance monitoring for all agents
- Cost tracking for API usage
NOT FINANCIAL ADVICE: Polyseer provides analysis for entertainment and research purposes only. All predictions are probabilistic and should not be used as the sole basis for financial decisions.
- Privacy Policy: We respect your privacy
- Terms of Use: Fair use and guidelines
- Liability: Limited liability for predictions
- Jurisdiction: Governed by applicable laws
This project is licensed under the MIT License - see the LICENSE file for details.
- Valyu Network: Authentication & real-time search API
- OpenAI GPT-5: Advanced reasoning capabilities
- Polymarket: Prediction market data
- Kalshi: Prediction market data
- Supabase: Backend infrastructure
Ready to see the future? Start analyzing markets at polyseer.xyz
Remember: The future belongs to those who can see it coming. Don't miss out again.
See the Future. Don't Miss Out.