Skip to content

Latest commit

 

History

History
140 lines (109 loc) · 5.19 KB

File metadata and controls

140 lines (109 loc) · 5.19 KB

ArchIToken · Repository Manifest

Status: active working repository Updated: 2026-05-16 License: Apache-2.0 OR MIT Development entry: AGENTS.md for ChatGPT / Codex


1. Positioning

ArchIToken = AEC AI-Native + Harness Engineering + OpenBIM CDE Workflow OS

ArchIToken is not a clone or direct replacement for Revit, Tekla, PKPM, Glodon, ZWCAD, Siemens Building X or similar mature single-point products. It is the open engineering intelligence layer above those tools.


2. Root Files

File Role Current status
AGENTS.md ChatGPT / Codex development-agent instruction entry active
README.md Repository entrypoint and architecture reading order active
LICENSE Apache-2.0 license text active
LICENSE-MIT MIT license text active
SECURITY.md Vulnerability reporting and security model active
CONTRIBUTING.md Contributor and PR rules active
CODE_OF_CONDUCT.md Community conduct policy active
CHANGELOG.md Human-readable change log active
versions.toml Version and upstream reference registry active
.env.production.example Production environment template active
.env.phase7.example Phase 7 runtime environment template active
docker-compose.production.yml Production-oriented compose entry active
docker-compose.phase7.yml Phase 7 local runtime compose entry active
docker-compose.phase8-scale.yml Phase 8 scale-test compose entry active

CLAUDE.md, tracked .claude/ files and docs/ZED-CLAUDE-CODE-SETUP.md are retired. Claude / Anthropic references may remain only as optional model-provider adapters, historical changelog entries or reference material.


3. Source Of Truth

Read in this order:

  1. 02-architecture/ARCHITOKEN-SOURCE-OF-TRUTH.md
  2. 02-architecture/CONSTITUTION.md
  3. 02-architecture/POSITIONING_AND_COMPETITIVE_STRATEGY.md
  4. 02-architecture/PROFESSIONAL_STANDARDS_COMPLIANCE.md
  5. AGENTS.md
  6. 02-architecture/MODULES.md
  7. 02-architecture/MODULE-REGISTRY.md
  8. 02-architecture/ARCHITECTURE.md
  9. 01-product/PRD.md
  10. README.md

4. Repository Layout

Path Role
.github/ GitHub issue, PR and workflow templates
01-product/ Product requirements and market scope
02-architecture/ Constitution, architecture, module registry and positioning truth
03-frontend/ Next.js / React / TypeScript primary production frontend workbench
03-frontend-vite/ Experimental Phase 7 Open AEC sidecar workbench; not the production primary frontend
04-backend/ Rust backend and Python agent orchestrator
05-infra/ Docker, Kubernetes, runtime, observability and scale infrastructure manifests
06-workers/ CAD, BIM, Office, PDF, GIS, media and AI worker adapters
07-deployment/ Deployment runbooks
08-sdk/ SDK generation and API client contracts
config/ Shared seed configuration such as the technology radar
docs/ Supporting technical notes, patches and historical records
infra/ Phase 8 runtime, observability and scale baseline manifests
tools/ Local tooling and load-test utilities

5. Active Module Model

ArchIToken uses 16 peer modules through registry-based contracts:

personal_center
marketing_service
planning_management
concept_design
standard_library
detailed_design
quantity_costing
material_logistics
production_manufacturing
construction_management
digital_twin
digital_archive
finance_management
human_resources
ai_center
settings_center

Modules must share the same platform shell, Open CDE file workbench, lifecycle, approval, audit, AI assistant and business-object workflow. /app/modules/digital_twin is not an isolated digital twin cockpit; standalone /app/digital-twin is retired and must not be reintroduced as a separate product entry.


6. Engineering Constraints

  • Registry replaces hardcoded Enum for modules, tools, models, renderers, workflows and rules.
  • Every module, term, rule and AI output must bind to professional roles, regulators, standards/codes, evidence and review state across AEC, production, logistics, customs, tax, finance, accounting, HR, organization, AI, data, cybersecurity and software.
  • AI calls must route through internal ModelRouter / InferenceRouter / ToolRouter / WorkflowRouter boundaries.
  • Generator and Evaluator must remain separated through Planner -> Generator -> Evaluator -> RuleChecker -> SchemaValidator -> Approver.
  • Default frontend theme is wechat_light, a WeChat-style white/gray/green product shell.
  • WebGPU is the primary high-performance rendering path; Three.js is fallback and ecosystem layer.
  • External providers, including OpenAI and Anthropic, are runtime adapter choices, not project identities.
  • Security-critical actions must be auditable: files, lifecycle, approvals, model calls, tool calls and worker derivatives.

7. Validation

Common checks:

cd 03-frontend
bun run lint
bun run typecheck
bun run test
bun run build

cd ../04-backend
cargo check
cargo test

cd agent-orchestrator
pytest

Run only the checks relevant to the files touched by a change when working locally, but CI should remain strict.