Cortex (AI Workspace)

This section outlines the strategic milestones and development timeline for the consolidated Cortex AI processing workspace.

Completed Milestones

AI Stack Consolidation: Unified the AI ecosystem by co-locating the API Ingress Gateway, Model Context Protocol (MCP) data plane, and agent runtimes under cortex/. Removed legacy configurations from /tools and /platform.
Layered Domain Reorganization: Removed the redundant core/ wrapper directory, restructuring the workspace into flat root-level domains: gateway/, memory/, mcp/, and agents/.
Gateway Port Alignment: Remapped container ports and updated internal agent settings to route traffic through the Go-based gateway on port 8080.
Database Startup Resiliency: Implemented container healthchecks inside the compose manifest to ensure downstream adapters wait until databases are fully initialized before attempting socket connections.

Database Layout: Implement a self-hosted MongoDB persistent storage layer split across chat_history, entities, and relations collections.
API & Core Libraries: Code Mongoose schema validation models and Fastify API endpoints to handle transactional memory storage and $vectorSearch similarity queries.
Visual Memory Dashboard: Build an interactive React + Vite + Tailwind CSS dashboard within cortex/memory/services/web to render the Context Graph using React Flow.
Docker-Only Execution: Package the memory sub-stack so that database, API, and Web dashboard containers build and launch strictly via Docker Compose.

E2E Agent Validation: Validate volume mount integrity, OAuth token persistence, and tool connectivity from within containerized agent terminal shells.
Host CLI Decommission: Prune legacy configuration folders (.agents/, .github/copilot/) and decommission host-global agent installations once container environments are certified.