Home

Published from the source document. Propose an edit through a pull request; direct wiki edits are replaced by the next publication.

APEX documentation

This index describes the runnable Rust implementation and its current, unversioned input schemas, which ship with the application release. Runtime capabilities and generated schemas are authoritative for the installed build; the migration audit records the scope of historical v2 comparisons.

Use and integrate APEX

Guide	Read it for
Container start	One-command server startup, credentials and retained state
Operating guide	Build, run, verify and understand runtime limits
Agent workflows	Data intake, planner conversations and extension responsibilities
Agent integration	MCP, HTTP, authentication, pagination and large imports
Server configuration	Enabled packages, per-call selection and scoped state
Control platform	Tenants, versioned scenarios, background runs, approval, PostgreSQL, MCP/HTTP and the desktop display client
MCP App insights	Revision-bound dashboards, charts, source evidence and customization view/renderer hooks
Executable data model	Units, routes, orders, conditional activities, locks, objectives and KPIs
Material preparation	Existing-supply allocation, flexible routes and material modes
Search and objectives	Q policies, XH, XT, replay and search limits
Combined improvement	One shared budget for XH, XT and optional XE
Direct schedule evolution	Priority chromosomes, genetic operators and extension contracts
Declarative scheduling	Typed constraints, mandatory dispatch policies and explanations
Synthetic customization	Domain knowledge, native hooks and validation requirements

Schemas and demo

Start with the data model and planning workflow for the two import paths, optional production expansion, lot creation and planning options. Material preparation explains when supply is allocated. The schema roles and tool releases describe the three input contracts and release boundary; the schema directory README maps generated files to Rust types and gives regeneration commands.

The factory showcase presents the MES/Excel workflow. Its customization package contains an optional source adapter that reads the live MES and saved workbook into canonical input JSON, plus independent technical policy examples. Import, scenario submission and planning are explicit steps. Small synthetic inputs used to verify individual engine capabilities remain test fixtures.

Work on the repository

Current architecture maps modules, the shared evaluation flow, service state and customization boundaries. Start there before changing implementation; its change map links responsibilities to tests.

Read the contribution guide for setup, review and contribution rights. Repository maintenance describes CI, dependency updates, GitHub protections and the release process. Use the support guide for questions and the security policy for private vulnerability reports.

Evidence and release

• Migration audit consolidates v2 comparison evidence, implemented coverage and remaining boundaries.
• The 24 September 2026 benchmark provides a frozen reproduction ZIP, public scientific instances and the original results. Later algorithm development is outside this baseline.
• Licensing documents explain the canonical public license, pricing and eligibility.
• Publication preparation lists release checks.
• Wiki publication explains how reviewed documentation is published automatically.

The data-model reference covers the current scheduling problem and its input contract. The removed v2 source and execution harnesses cannot be run from this checkout; see migration audit.