Skip to content

Quickstart

The fastest way to run the complete development stack is Docker Compose.

Prerequisites

  • Docker with Compose support
  • At least several gigabytes of available memory and disk space
  • A modern browser

Java, Maven, Python, Node.js, AWS CLI, SAM CLI, and Go are only required for source development or generated-project verification.

Start the Stack

From the repository root:

docker compose up --build

Wait for PostgreSQL and the backend to become healthy, then open:

  • Modeling frontend: http://127.0.0.1:8082
  • Backend health: http://127.0.0.1:8080/api/health
  • API browser: http://127.0.0.1:8080/swagger-ui.html
  • Landing site: http://127.0.0.1:8083
  • Container logs: http://127.0.0.1:9999

First Project

  1. Register a user in the frontend.
  2. Create a project.
  3. Open the CIM workspace.
  4. Create a model or import mde/samples/cim.xmi.
  5. Save and validate the CIM.
  6. Select Generate PIM, review the result, then select Generate PSM.
  7. From PSM, select Generate Artifacts.
  8. Open the artifact explorer and download the generated ZIP.

The included climate-relief sample exercises a broad portion of the CIM language and is the best starting point for exploring the full pipeline.

Enable the AI Assistant

The assistant is configurable and should remain disabled unless provider credentials, operating cost, and organizational policy are intentional.

MODRISS_AI_ENABLED=true
MODRISS_AI_PROVIDER=openai
OPENAI_COMPATIBLE_BASE_URL=your_arvan_openai_compatible_base_url
OPENAI_COMPATIBLE_API_KEY=your_api_key
MODRISS_AI_MODEL=Gemma-4-31B-IT
MODRISS_AI_MODE=unified
MODRISS_AI_OPENAI_PROTOCOL=json_schema
MODRISS_AI_NATIVE_TOOLS_PREFERRED=false
MODRISS_AI_FORCED_TOOL_CHOICE_RELIABLE=false
MODRISS_AI_REQUEST_TIMEOUT=240s
MODRISS_AI_TURN_TIMEOUT=12m
MODRISS_AI_SOURCE_TURN_TIMEOUT=25m
MODRISS_AI_MAX_SYSTEM_CHARS=32000
MODRISS_AI_MAX_COMPLETION_TOKENS=16000
MODRISS_AI_MAX_PROVIDER_CALLS_PER_TURN=64
MODRISS_AI_MAX_PROVIDER_CALLS_SOURCE_TURN=96
MODRISS_AI_PROVIDER_RETRY_ATTEMPTS=0
MODRISS_AI_PREFER_LLM_SOURCE_EXTRACTION=true
MODRISS_AI_LLM_REVIEW_ENABLED=true
docker compose up --build

The validated deployment uses Arvan and Gemma-4-31B-IT. unified automatically chooses the internal modeling workflow; users and API clients do not select it. These limits are the tested Gemma profile: a standard turn reserves up to two calls for adaptive routing and exposes the remaining bounded budget to obligation planning, type selection, blueprinting, private slices, corrections and independent review. The conceptual schema caps blueprints at 96 objects/types and requirement ledgers at 64 obligations; required Ecore closure and call reserves can lower the effective capacity.

Stop or Reset

Stop containers while preserving data:

docker compose down

Remove local PostgreSQL and emulator volumes (only when you intentionally want a clean state):

docker compose down -v

The second command permanently removes local runtime data.