Aqium Documentation
Guides, architecture references, and administration docs for Aqium — the AI-powered platform that generates production-ready business software from structured specifications.
Welcome to Aqium Documentation
Aqium turns structured conversations about your business into running, production-ready software. It does this by building a living specification graph of your requirements, then driving a deterministic code-generation pipeline that stays in sync with that graph as your business evolves.
Start Here — Choose Your Path
How It Works
For business owners and product teams. Understand Aqium's discovery process, how your requirements become a specification, and how that specification drives software generation — without reading a line of code.
System Architecture
For technical readers. Deep-dive into the six-layer unified graph, the convergence model, the code-generation pipeline, and how all subsystems fit together.
Platform Administration
For developers and operators. Deployment, configuration, Railway per-customer instances, environment management, and operational runbooks.
What's New
Config Kits — ERPNext/Frappe-informed output path
Config Kits introduce a config-driven generation path alongside the existing code-generation path. Informed by Frappe/ERPNext's DocType model, Config Kits let you describe business objects and workflows declaratively; Aqium generates the correct config artifacts (JSON, YAML, DocType definitions) rather than imperative code when that is the right output for your target platform.
Six-Layer Unified Graph
The project graph has been extended from four layers to six, adding explicit Config and Deployment layers alongside Spec, Code, Data, and API. Every node in the graph carries cross-layer edges so drift, blast-radius, and coverage calculations span the full stack.
Deterministic Convergence Model
Convergence is now evaluated against five binary conditions rather than a heuristic score. A generation batch is considered converged only when all five pass: schema match, interface contract match, test-suite pass, lint clean, and CEGIS structural-diff match. This makes convergence a fact, not an opinion.
Brownfield Pipeline with CPG Extraction
Existing codebases can be imported via the brownfield pipeline. Aqium runs Code Property Graph (CPG) extraction on your repository, maps the resulting nodes into the unified graph, and identifies spec gaps between what the code does and what a generated specification would require.
Per-Customer Railway Deployment
Each customer instance is deployed as an isolated Railway project via the automated deployment pipeline. Provisioning, environment injection, and lifecycle management are handled by the platform — no manual infrastructure work required per tenant.
Documentation Sections
| Section | Description |
|---|---|
| Concepts | Discovery, obligations, specification structure, and the requirement lifecycle |
| Architecture | Unified graph, subsystem boundaries, data flow, and integration points |
| Code Generation | Generation pipeline, Config Kits, dependency ordering, pattern reuse, and CEGIS verification |
| Convergence | The five-condition convergence model and how the platform evaluates generation completeness |
| Flywheels | Pattern flywheel, standards flywheel, and how successful generations improve future output |
| Deployment | Railway per-customer deployment, environment configuration, and infrastructure operations |
| Guides | Step-by-step guides for greenfield projects, brownfield imports, and common workflows |
| Admin | Platform administration, tenant management, and operational runbooks |
| Reference | API reference, kit authoring reference, and schema reference |