claude-flow
Version:
Ruflo - Enterprise AI agent orchestration for Claude Code. Deploy 60+ specialized agents in coordinated swarms with self-learning, fault-tolerant consensus, vector memory, and MCP integration
148 lines (139 loc) • 6.06 kB
text/typescript
/**
* Pipeline message schemas — ADR-126 Phase 5
*
* The four neural-trader agents (`market-analyst`, `trading-strategist`,
* `risk-analyst`, `backtest-engineer`) coordinate as a typed SendMessage
* pipeline, NOT as parallel role descriptions:
*
* market-analyst — (RegimeVerdict) ─→ trading-strategist
* trading-strategist — (SignalProposal[]) ─→ risk-analyst ◄── BLOCKING
* risk-analyst — (RiskDecision) ─→ trading-strategist
* └─→ execute-or-halt
*
* `trading-strategist` MUST NOT call the live broker (`--broker <name>`)
* without a `RiskDecision` with `decision: 'approved'` from `risk-analyst`
* in the current SendMessage trace. This is the structural risk-gate per
* ADR-126 — see `trading-strategist.md` for the in-agent enforcement.
*
* `backtest-engineer` is an orthogonal lane — it produces signed-artifact
* promotion candidates (ADR-126 Phase 4) and does NOT participate in the
* live execution pipeline. It runs in parallel during research, but the
* hot path of live trading never depends on it.
*
* These types are documentation + a future programmatic dispatcher
* contract. SendMessage payloads today are JSON literals; tomorrow a
* typed dispatcher can import + validate against this schema.
*
* Refs:
* - ADR-126 Phase 5 — SendMessage pipeline plan
* - CLAUDE.md — SendMessage protocol + named-agent rules
* - SKILL.md files for each agent — the actual call sites
*/
/* ---------------------------------------------------------------------- */
/* market-analyst → trading-strategist */
/* ---------------------------------------------------------------------- */
/**
* Regime verdict — produced by `market-analyst`, consumed by
* `trading-strategist`. Identifies the current market regime so the
* strategist can pick an appropriate strategy family before generating
* signals.
*/
export interface RegimeVerdict {
/** Message schema discriminator. */
type: 'regime-verdict/v1';
/** Originating agent name (must be `market-analyst` for valid messages). */
from: 'market-analyst';
/** ISO timestamp when the verdict was produced. */
timestamp: string;
/** The classified regime — one of the six market-analyst categories. */
regime:
| 'bull-trending'
| 'bear-trending'
| 'ranging'
| 'high-volatility'
| 'low-volatility'
| 'transitioning';
/** The ticker(s) the verdict applies to. */
symbols: string[];
/** Confidence score in [0,1]. */
confidence: number;
/** Optional supporting indicators (ADX, RSI, MACD, VIX, ...). */
indicators?: Record<string, number>;
}
/* ---------------------------------------------------------------------- */
/* trading-strategist → risk-analyst */
/* ---------------------------------------------------------------------- */
/**
* Signal proposal — produced by `trading-strategist`, consumed by
* `risk-analyst`. Represents a proposed live-trade action that requires
* risk approval before execution.
*/
export interface SignalProposal {
/** Message schema discriminator. */
type: 'signal-proposal/v1';
/** Originating agent name. */
from: 'trading-strategist';
/** Unique signal id — used to correlate the RiskDecision response. */
signalId: string;
/** ISO timestamp when the proposal was created. */
timestamp: string;
/** Symbol to trade. */
symbol: string;
/** Direction. */
side: 'long' | 'short' | 'close';
/** Strategy that produced the proposal. */
strategyId: string;
/** Position size as % of portfolio (e.g. 0.02 = 2%). */
sizePct: number;
/** Z-score / confidence score driving the entry. */
confidence: number;
/** Reference to the upstream regime that motivated the trade. */
regime?: RegimeVerdict['regime'];
/** Optional metadata — model attribution, indicator snapshot, etc. */
metadata?: Record<string, unknown>;
}
/* ---------------------------------------------------------------------- */
/* risk-analyst → trading-strategist */
/* ---------------------------------------------------------------------- */
/**
* Risk decision — produced by `risk-analyst`, consumed by
* `trading-strategist`. The BLOCKING GATE: `trading-strategist` MUST NOT
* execute a live broker call without a corresponding `RiskDecision` with
* `decision: 'approved'` for the proposal's `signalId`.
*
* Rejection reasons map to the risk-analyst's circuit-breaker / sizing
* checks (VaR, CVaR, concentration, correlation, drawdown).
*/
export interface RiskDecision {
/** Message schema discriminator. */
type: 'risk-decision/v1';
/** Originating agent name. */
from: 'risk-analyst';
/** The signal id this decision responds to (correlates with SignalProposal.signalId). */
signalId: string;
/** ISO timestamp when the decision was made. */
timestamp: string;
/** Decision — approved means the strategist may call the broker. */
decision: 'approved' | 'rejected';
/** Optional size adjustment — risk-analyst may shrink the proposal. */
adjustedSizePct?: number;
/** Human-readable reasons (empty array on approval). */
reasons: string[];
/** Risk metrics computed for the decision (VaR/CVaR snapshots, etc.). */
metrics?: {
var95?: number;
cvar95?: number;
portfolioCorrelation?: number;
concentrationPct?: number;
drawdownPct?: number;
};
}
/* ---------------------------------------------------------------------- */
/* Union for typed dispatchers */
/* ---------------------------------------------------------------------- */
/**
* Discriminated union of every valid SendMessage payload in the
* neural-trader pipeline. A future programmatic dispatcher can switch
* on `type` to validate + route incoming messages.
*/
export type PipelineMessage = RegimeVerdict | SignalProposal | RiskDecision;