SIGNAL FABRIC
A deterministic decision engine that turns market signals into daily calls.
Signal Fabric runs a Dagster-orchestrated production pipeline for US equities. It observes the market, produces BUY, SELL, or HOLD decisions, and emits notifications with traceable artifacts rather than manual guesswork. The planet is built around deterministic operations and observation-first discipline.
Overview
Signal Fabric is a decision-support platform and trading orchestrator for US equities. It acts as a deterministic engine that ingests market signals to produce explainable BUY, SELL, or HOLD recommendations.
Architecture
The system is built around a Python 3.11 data pipeline orchestrated by Dagster. It relies on strict data contracts, time-series storage via QuestDB, and a contract-first event bus to ensure all signal generation is idempotent and versioned.
Problem It Solves
Manual trading analysis is often emotional, non-reproducible, and opaque. Signal Fabric enforces a deterministic, observable "human-in-the-loop" approach: the system computes and recommends based on math, but a human executes the trade.
Current State
MVP operational. The Dagster-based orchestrator has been stabilized around deterministic signal generation, and the long-term evolvable architecture is defined.
MILESTONES
Dagster Pipeline Stabilized
Dagster-based trading orchestrator stabilized around deterministic BUY/SELL/HOLD signals.
Token Cascade Fix — Explicit Theme Block
Added explicit sky --theme-* block to signal-board after tokens.css Tier 2 defaults were removed ecosystem-wide; planet identity no longer relies on shared-library defaults.
Trading-Desk 502 Root Cause Identified
Diagnosed HTTP 502 on /trading-desk/ — docker-compose.yml profiles gate excluded service from network; Dagster webserver and annotator additions ruled out.
Trading-Desk Compose Profile Gate Removed
Removed profiles: [trading-desk] gate from docker-compose.yml; service now starts on plain docker compose up -d, resolving all /trading-desk/ 502 errors.