Pump Party Docs

Appearance

Pump Party Docs ​

This is the front door for the Clash of Clones product spec, clone strategy model, data ingestion, and SQLite backend rewrite.

Summary ​

The core strategy model is:

text
Data Stream Block -> Asset Block -> Trading Prompt Block

The workflow canvas also shows a Trading Clone Block as the global runtime terminal:

text
Data Stream Block -> Asset Block -> Trading Prompt Block -> Trading Clone Block

Each Trading Prompt Block defines one strategy. At runtime, the backend expands that strategy across the enabled candidate assets connected to it:

text
one clone -> multiple enabled assets -> one trading prompt run per asset

There is no separate v1 insight-generation layer. The trading prompt consumes compact structured market memory directly from SQLite, and large prompt/context/response payloads should be archived to R2 once the artifact adapter is wired.

Product Loop ​

The broader product is documented in Clash Of Clones Product Spec.

text
Creation -> Audition -> Promotion Queue -> Arena -> Retirement
   pay fee    24h test       wait for slot    compete    archived career

Users create clones, pay an Audition Fee, configure public identity and private strategy, and run a 24-hour synthetic audition. Passing clones enter the Clash Arena directly when slots are open, or wait in the Promotion Queue. Arena clones compete in 6-hour synthetic races while users buy 0.01 ETH tickets for the currently accumulating race.

Race pools split into:

Pool sliceShare
Protocol rake5%
Creator pool5%
Ticket-holder distribution pool90%

The paper-trading runtime owns the synthetic trading results that auditions and races consume. Other product surfaces can build around those results: payments, moderation, ticket sales, settlement, creator royalties, promotion, retirement, and Hall of Fame views.

Current Backend Status ​

The new services/market-backend backend now has:

  • Global SQLite market ingestion tables for Hyperliquid, Polymarket, and DefiLlama.
  • Clone-owned pipeline graph storage.
  • Workflow graph layout metadata for the visual Trading Clone Block.
  • Pipeline preview API with compiled read plans and effective context packets.
  • Asset latest-data API for reviewing the structured memory a connected asset receives.
  • Compact per-asset decision run/action tables.
  • A decision worker loop with cadence checks and a pluggable decision engine.
  • A local no-op hold engine for safe scheduler/storage testing.
  • A model-backed dry-run engine path.
  • Dry-run execution safety gates.
  • Execution audit tables for dry-run orders and ledger events.

Still pending:

  • R2 artifact adapter for full context/prompt/response/reason payloads.
  • Paper fill simulation and synthetic account/position mutation.
  • Live order adapter and personal-wallet execution boundary.
  • Full frontend controls polish against the new graph builder.

Data Sources ​

V1 market context comes from three source families:

SourceStored context
HyperliquidMid prices, OHLC candles, liquidity profile, funding pressure, volatility profile, positioning pressure
PolymarketMarket discovery, immediate market, later market, recently closed markets
DefiLlamaCapital base, capital flows, economic throughput

Hyperliquid is the trade universe source. Polymarket and DefiLlama add related market and macro context. DefiLlama is resolved as macro context: enabled assets receive global DeFi rows, plus matching chain or stablecoin rows when the DefiLlama subject can be linked to the asset symbol.

See Data Sources for the source-specific ingestion pages and diagrams.

Selectable Asset Categories ​

Users select Hyperliquid assets by category/subcategory, then can manually include or exclude specific symbols.

text
High level:
  All
  Perps
  Spot
  Crypto
  Tradfi
  Trending

Crypto:
  All
  AI
  Defi
  Gaming
  Layer 1
  Layer 2
  Meme

Tradfi:
  All
  Stocks
  Indices
  Commodities
  FX
  Pre-IPO

Spot, if exposed:
  All
  USDC
  USDH
  USDT

Generic HIP-3 category exposure is out of scope.

Control Granularities ​

ModeUser-facing control
BasicGenerated graph with simple defaults
CustomIndividual assets plus source/profile presets
Power UserEditable data stream, asset, and trading prompt blocks with advanced per-channel and prompt-owned execution controls

Retrieval is deterministic. Users do not prompt an AI to fetch data.

Data Pipeline ​

The v1 runtime path is:

text
1. Global source ingestion
2. SQLite market memory
3. Clone strategy graph compilation
4. Per-strategy asset expansion
5. Deterministic context retrieval
6. Trading prompt assembly
7. Decision worker creates one due run per asset
8. Decision engine proposes an action for that asset
9. Safety gates validate or reject the action
10. Valid executable actions create dry-run orders today
11. SQLite decision/action/order/ledger storage
12. R2 archive for large audit payloads

Local worker commands:

bash
cd services/market-backend
npm run decisions:run-once -- --clone-id=42 --force
npm run decisions:worker

Review Docs ​

Reference Contracts ​