One Project, Every Surface¶

One project. Every surface. All live.

truST keeps engineering, runtime, HMI, automation, and AI assistance tied to the same project instead of splitting them into disconnected toolchains.

Figure: The same project is edited, executed, inspected, operated, scripted, and AI-assisted through different surfaces. The surfaces do not become separate project models.

Pillars¶

One project, every surface, all live¶

The project files are the shared contract. VS Code, the runtime, Browser IDE, Browser HMI, CLI/CI, Agent API, and truST Mesh all point back to the same source, config, HMI, and bundle artifacts.

Use this mental model when a change crosses surfaces:

• edit and diagnose in VS Code
• run, inspect, debug, and reload through runtime surfaces
• expose operator views through Browser HMI
• automate repeatable work through CLI/CI and the Agent API
• connect runtimes and plant systems through truST Mesh

AI with truST tools: typed, audited, policy-guarded¶

truST's editor AI integration is not just chat over source files. The VS Code extension exposes typed language-model tools for ST diagnostics, navigation, file reads and edits, HMI authoring, telemetry, settings, and debug actions. Those tools are declared in the extension manifest, registered at activation, and covered by a contract test that checks the manifest, activation events, and registered tool names stay in sync.

Scope matters:

• Editor AI tools cover editor intelligence, file operations, HMI workflows, telemetry/settings, and debug actions.
• Build, validate, test, compile/reload, and deterministic harness orchestration are covered by the external Agent API.
• HMI writes remain guarded by descriptor policy, allowlists, and runtime authorization; AI tooling does not bypass those controls.

Evidence:

• VS Code tool registrations: editors/vscode/src/lm-tools.ts
• VS Code language-model tool manifest: editors/vscode/package.json
• Drift guard: lm-tools-contract.test.ts

Open artifacts¶

truST keeps the important project state in files that can be reviewed, diffed, generated, and scripted:

• ST source files
• runtime.toml
• io.toml
• hmi/ descriptor files
• generated bundle artifacts such as program.stbc
• PLCopen XML when exchanging with other ecosystems

Visual editors still feed the same project model. Ladder, SFC, Blockly, statecharts, and ST should be understood as authoring surfaces over open artifacts, not separate execution engines.

Measured behavior¶

The product story should stay tied to repeatable proof:

• benchmarks for runtime, T0, mesh, and dispatch behavior
• conformance for language and protocol expectations
• examples for runnable project workflows
• extension tests for VS Code behavior, including the AI tool contract
• public matrices that separate what is supported, partial, unavailable, or not applicable

Choosing A Surface¶

Start in VS Code when you are doing daily engineering work: edit ST, inspect diagnostics, use visual editors, debug, and watch runtime state in one place.

Use the Agent API when the work needs to be automated by a script, CI job, or external agent. Use the Browser IDE for runtime-hosted quick edits, walkthroughs, and demos. Use Browser HMI when the user is operating or inspecting an already-running system.

Surface Capability Matrix¶

Status key:

• full: this surface is a primary supported path for the capability
• partial: useful support exists, but the surface is not the full or primary path
• no: this surface does not currently provide the capability
• n/a: the capability does not fit the surface

Honest Limits¶

• VS Code is the primary integrated engineering surface. Browser IDE is useful for runtime-hosted editing and demos, but it is not a replacement for every VS Code extension workflow.
• Editor AI tools do not currently expose direct build/test/validate or deterministic harness methods. Use the Agent API for those loops.
• Editor AI tools can read and edit visual-editor artifact files, but there is not yet a dedicated visual-editor AI tool surface.
• Agent API v1 intentionally omits code actions and attached-session source-aware reload workflows. It also does not mirror full symbol navigation or rename/refactor behavior yet. The current contract documents those boundaries explicitly.
• Browser HMI is an operator and technician surface. It is not an authoring surface and should not be marketed as a SCADA replacement.
• Visual editors are strongest when their generated companion artifacts are validated through the same build/runtime path as ST.

Related¶

• truST Mesh
• Editors
• Program In VS Code
• Program In Browser IDE
• Agent Quickstart
• Agent API overview
• AI Assistance
• HMI And Web UI