Source logic for Idea #149
Go to file
agent-7e3bbc424e07835b c93be013f3 build(agent): new-agents-2#7e3bbc iteration 2026-04-20 15:11:27 +02:00
deltax_forge_cross build(agent): new-agents-2#7e3bbc iteration 2026-04-20 15:11:27 +02:00
tests build(agent): new-agents-2#7e3bbc iteration 2026-04-19 23:04:30 +02:00
.gitignore build(agent): new-agents-2#7e3bbc iteration 2026-04-19 23:04:30 +02:00
AGENTS.md build(agent): new-agents-2#7e3bbc iteration 2026-04-19 23:04:30 +02:00
README.md build(agent): new-agents#a6e6ec iteration 2026-04-20 15:07:26 +02:00
pyproject.toml build(agent): new-agents-2#7e3bbc iteration 2026-04-19 23:04:30 +02:00
test.sh build(agent): new-agents-2#7e3bbc iteration 2026-04-19 23:04:30 +02:00

README.md

DeltaX-Forge: Cross-Venue Delta-Driven Execution Slicer for Equities

DeltaX-Forge is an open, real-time cross-venue execution planner that coordinates hedging actions across multiple venues while preserving auditability and privacy guarantees.

  • Canonical data model and DSL: StrategyDelta, Asset, MarketSignal, PlanDelta with Authority and AuditLog blocks.
  • Dual-tier coordination: lightweight local venue solvers at each venue and a central curator enforcing cross-venue coherence.
  • Adapters: starter adapters for data feeds (price signals) and venue execution (hedge routing).
  • Deterministic replay: a log-based replay engine to deterministically rebuild past runs for compliance and post-trade analysis.
  • Privacy by design: secure aggregation and policy-driven data minimization.
  • Governance ledger: cryptographic signatures and tamper-evident logging for regulatory reviews.

This repository implements a production-ready MVP for cross-venue hedging with a canonical DSL and two-tier solver scaffolding. See tests for usage surface and the existing adapters.

How to run tests locally

  • Install dependencies and run tests:
    • bash test.sh
  • Build package and run tests:
    • python3 -m build

Extending DeltaX-Forge

  • New functionality should be added with small, well-scoped changes and accompanied by unit tests.
  • The MVP currently includes a toy LocalVenueSolver, a simple CentralCurator, two starter adapters, deterministic replay, and governance/signing primitives.

For contributors

  • See AGENTS.md for architectural rules and contribution guidelines.
  • The repository is designed to be self-contained and extensible to multi-venue and multi-asset scenarios.

READY_TO_PUBLISH marker is added at the repository root when the project is ready for publishing.