Capacitr API Docs

Accepts a news URL or freetext query. Scrapes the content (URL path via Jina AI), runs LLM extraction to identify tickers, entities, keywords, and categories, then fans out to Polymarket, HyperLiquid, and Deribit to find matching markets. Results are enriched with Quotient mispricing signals and sorted by edge opportunity. Search results are logged to the market_searches Supabase table.

POST /api/analyze-link
Content-Type: application/json

{
  "url": "https://reuters.com/technology/nvidia-earnings-beat-2025"
}

{
  "summary": "NVIDIA reported record Q1 earnings...",
  "keywords": ["NVIDIA", "earnings", "semiconductors", "AI chips"],
  "entities": ["NVIDIA", "Jensen Huang"],
  "tickers": ["NVDA"],
  "categories": ["tech", "finance"],
  "predictions": [
    {
      "id": "will-nvidia-hit-1000",
      "question": "Will NVIDIA stock hit $1000 by Q3 2025?",
      "yesPrice": 0.62,
      "noPrice": 0.38,
      "volume": 485000,
      "mispricingScore": 0.14,
      "bluf": "Market underpricing YES given recent earnings beat"
    }
  ],
  "perps": [
    {
      "symbol": "NVDA-USD",
      "price": 872.50,
      "funding": 0.0003,
      "openInterest": 12400000
    }
  ],
  "options": []
}

Proxies requests to the Quotient API for mispricing intelligence on Polymarket markets. Supports three actions: fetching mispriced markets, looking up specific markets by slug, and querying general market intelligence.

GET /api/quotient?action=mispriced&limit=5&min_spread=0.08

GET /api/quotient?action=lookup&slugs=will-bitcoin-hit-100k,eth-merge-by-q3

Streams a response from the Capacitr market research agent using the Vercel AI SDK (Server-Sent Events). The agent can call tools: searchMarkets (Polymarket text search), getMispricedMarkets (Quotient top opportunities), and getMarketIntelligence (Quotient intelligence for a query). Use the useChat hook from ai/react client-side.

POST /api/chat
Content-Type: application/json

{
  "messages": [
    { "role": "user", "content": "What prediction markets exist around the US election?" }
  ]
}

Returns a streaming text/event-stream response in Vercel AI SDK Data Stream format. Compatible with the useChat hook.

Fetches top posts from Reddit subreddits and Hacker News based on the user's interest categories (crypto, politics, ai, tech, finance, sports, pop-culture, music, food, environment). Each post is analyzed by LLM to extract tickers, entities, and generate a prediction card with Polymarket + HyperLiquid + Deribit markets. Results are enriched with Quotient mispricing signals.

GET /api/feed?privyId=did:privy:abc123

{
  "items": [
    {
      "id": "abc123",
      "title": "Bitcoin ETF inflows hit record $2B in single day",
      "source": "reddit",
      "subreddit": "cryptocurrency",
      "url": "https://reddit.com/r/cryptocurrency/...",
      "summary": "Bitcoin ETFs recorded record inflows...",
      "entities": ["Bitcoin", "BlackRock"],
      "tickers": ["BTC"],
      "predictions": [...],
      "perps": [...],
      "options": [],
      "categories": ["crypto", "finance"]
    }
  ]
}

{
  "user": {
    "id": "uuid",
    "privy_id": "did:privy:abc123",
    "privy_provider": "twitter",
    "display_name": "Atown",
    "avatar_url": "https://...",
    "twitter_handle": "capacitr_xyz",
    "farcaster_fid": null,
    "created_at": "2025-01-01T00:00:00Z",
    "interests": ["crypto", "ai", "finance"]
  }
}

{ "interests": ["crypto", "ai", "finance"] }

POST /api/interests
Content-Type: application/json

{ "privyId": "did:privy:abc123", "interests": ["crypto", "ai"] }

{
  "profile": {
    "tier": "aggressive",
    "scores": { "riskTolerance": 8, "timeHorizon": 7, "experience": 9 },
    "completed_at": "2025-03-15T10:22:00Z"
  }
}

Called automatically on login. Upserts the Privy user record into Supabase — creates if new, updates profile fields if existing. Called from the PrivyProviderWrapper on onSuccess.

{ "user": { "id": "uuid", "privy_id": "did:privy:abc123", "display_name": "Atown", ... } }

Returns aggregated analytics from the market_searches table: total search volume, unique URLs, top entities, top tickers, top categories, and recent searches. Admin-only — requires password query param.

{
  "totalSearches": 142,
  "uniqueUrls": 89,
  "topEntities": [["Bitcoin", 34], ["NVIDIA", 21], ...],
  "topTickers": [["BTC", 41], ["NVDA", 19], ...],
  "topCategories": [["crypto", 67], ["tech", 44], ...],
  "recentSearches": [...]
}

⚠️ Destructive. Deletes all rows from user_interests, feed_items, and users tables in sequence. For development/testing only — never run in production.

{ "success": true, "message": "All users wiped" }