Language Server (LSP)

The OPC UA Modeler ships a fully standard Language Server Protocol server. VS Code users get it automatically via the VS Code Extension. For all other editors — Neovim, Zed, Emacs, or any LSP-compliant tool — this page covers manual setup.

Starting the server

# stdio mode  — for editors that spawn the server as a child process
opcua-modeler lsp --stdio

# socket mode — for multi-client / WebSocket setups (e.g. browser-based editors)
opcua-modeler lsp --socket 7000

---

Neovim (nvim-lspconfig)

Add the following to your Neovim configuration (Lua):

local lspconfig = require("lspconfig")
local configs = require("lspconfig.configs")

-- Register the custom server if not already known
if not configs.opcua_modeler then
  configs.opcua_modeler = {
    default_config = {
      cmd = { "opcua-modeler", "lsp", "--stdio" },
      filetypes = {
        "yaml",           -- general YAML; restrict via root_dir if needed
      },
      root_dir = lspconfig.util.root_pattern("*.model.yaml", "*.nodeset.yaml", ".git"),
      single_file_support = true,
      settings = {},
    },
  }
end

lspconfig.opcua_modeler.setup({})

Tip: If you only want the OPC UA LSP to activate on .model.yaml / .nodeset.yaml files, add a custom filetypes detection or use vim.filetype.add to map those extensions to a dedicated filetype (e.g. opcua_yaml) before calling setup.

---

Zed

Add the following to your Zed settings.json (or ~/.config/zed/settings.json):

{
  "lsp": {
    "opcua-modeler": {
      "binary": {
        "path": "opcua-modeler",
        "arguments": ["lsp", "--stdio"]
      }
    }
  },
  "languages": {
    "YAML": {
      "language_servers": ["opcua-modeler", "..."]
    }
  }
}

---

Emacs

With eglot (built-in from Emacs 29+)

(with-eval-after-load 'eglot
  (add-to-list 'eglot-server-programs
               '(yaml-mode . ("opcua-modeler" "lsp" "--stdio"))))

;; Auto-start eglot when opening model files
(add-hook 'yaml-mode-hook #'eglot-ensure)

With lsp-mode

(with-eval-after-load 'lsp-mode
  (lsp-register-client
   (make-lsp-client
    :new-connection (lsp-stdio-connection
                     '("opcua-modeler" "lsp" "--stdio"))
    :activation-fn (lsp-activate-on "yaml")
    :server-id 'opcua-modeler)))

---

Socket / multi-client mode

For setups where multiple editors or a browser-based editor connect to a single shared server, use the --socket flag:

# Start a persistent LSP server on port 7000
opcua-modeler lsp --socket 7000

Clients connect via WebSocket or TCP to localhost:7000. The server maintains isolated per-connection state (document buffers, debounce timers, capability flags) while sharing the expensive OPC UA address space cache across all connections.

---

Diagnostics reference

The server emits diagnostics using structured codes in the form DSL-E-*, DSL-W-*, and DSL-I-*.

Prefix	Severity	Description
DSL-E-*	Error	Must be fixed — invalid or illegal model construct
DSL-W-*	Warning	Potentially incorrect — shown by default, suppressed with --quiet
DSL-I-*	Info	Style / conformance hints — hidden by default, shown with --style

Semantic checks performed

The following semantic diagnostics are produced by Layer 2:

Check	Example message
Invalid typeDefinition reference	Unknown type 'MySensor' — not found in address space
Invalid subtypeOf reference	'MyBase' is not an ObjectType
Broken organizedBy / componentOf browse path	Path 'Root/Objects/Tank' does not exist in node tree
Invalid promotedToMandatory entry	'Status' is not an optional member of 'SensorType'
Duplicate initializers: variable: entries	Duplicate initializer path 'Parameters/MaxValue'
Duplicate $anchor IDs	Anchor 'pump1' is already defined on line 42
Self-referencing placement	Node cannot reference itself or its own subtree as placement parent
YAML syntax error	YAML parse error at line 12: unexpected ':'

Suppressing specific codes

Use the --silence flag on the CLI, or configure suppressions in your project:

# Suppress a specific warning code
opcua-modeler generate --silence DSL-W-001

# Suppress all warnings matching a prefix
opcua-modeler generate --silence DSL-W

---

Address space cache

The server loads the OPC UA base address space (and all imported companion specs) once and caches it across all open connections. This means:

• The first open of a model file may take a second while the address space loads
• Subsequent completions and diagnostics are near-instant

The cache is invalidated when the list of namespaces: imports in your file changes.