VORTEX-GEN 3.0 "Centaur"

VORTEX is a production‑grade, agentic workflow engine for generative AI. It combines a Rust control plane, Python inference workers, and a vertical rack UI that compiles to a DAG for execution. The system is local‑first and cluster‑capable, with strong governance and auditability built in.

License: Apache 2.0.

What This Project Is

• An MCP‑style tool box for humans and AI agents.
• A template‑driven workflow system (templates → runs → artifacts).
• A governance‑first platform with human‑in‑the‑loop approvals.

Why It’s Different

• Rack‑first UX: vertical rack + signal bus (no infinite canvas).
• Zero‑copy execution: Apache Arrow shared memory.
• Server‑enforced policy: approvals and authZ are not UI‑only.
• Cluster‑ready: CPU control plane + GPU worker fabric.

Architecture (High‑Level)

UI (Svelte 5)  <->  Core Engine (Rust)  <->  Worker Fabric (Python)
   |                      |                        |
   | HTTP/WS              | Protobuf + UDS/TLS      | PyTorch + Arrow
   v                      v                        v
Vertical Rack         Graph Compiler           GPU/CPU Execution
                      + Scheduler              + SHM (local to worker)

Documentation (Source of Truth)

All requirements live in the SRS:

• gemini/docs/specs/00_master_srs.md
• gemini/docs/specs/01_core_engine_srs.md
• gemini/docs/specs/02_frontend_ui_srs.md
• gemini/docs/specs/03_compute_fabric_srs.md
• gemini/docs/specs/04_registry_srs.md
• gemini/docs/specs/05_ui_screens_spec.md
• gemini/docs/specs/06_environment_config_spec.md
• gemini/docs/specs/06_enterprise_infrastructure_srs.md
• gemini/docs/specs/07_data_flow_spec.md
• gemini/docs/specs/08_workspace_tools_spec.md

Project overview and rules:

• gemini/agent.md
• gemini/rules.md

Status

SRS complete, implementation in progress.

License

Apache License 2.0. See LICENSE.