Stories

Stories are the agent’s field notes: structured, time-stamped observations about on-chain behavior.

What a story is

A story is an LLM-generated narrative triggered by a pattern detector. It includes a headline, supporting evidence, addresses, and a score so you can quickly see what happened and why it matters.

How stories are generated

The pipeline looks like this:

1. Blockchain data is ingested into ClickHouse
2. Pattern-detection scripts find structural events
3. The LLM synthesizes the event into a story
4. The story is written back as a structured record

Common story types

• Rabbit Loop — circular flow and recycling patterns
• Rabbit Cluster — coordinated wallet group structure
• Rabbit Mirror — wallets moving in suspiciously similar ways
• Rabbit Staging — readiness patterns before an action
• Rug Liquidity Pull — liquidity removal signals
• Whale Stealth Flow — large holder movements with obfuscation
• Surge — sudden price/volume spike with validation
• Governance Delegation Surge — rapid voting-power shifts
• Ecosystem Shift — chain-level capital rotation
• Bridge Paths — cross-chain movement traces

Why it matters

Stories tell you that something happened. They are the starting point for investigation, not the final answer.

Story shape

{
  "id": "uuid",
  "chain": "ETH",
  "story_type": "SURGE",
  "score": 100,
  "title": "...",
  "summary": "...",
  "meta_json": {},
  "ts_created": "2026-04-21T00:00:00Z"
}

Next

• Read Theses
• See the Story Feed guide
• Open the API Reference