How to leverage inline rendering of WaveDrom and Mermaid in DVT AI Assistant in Eclipse

Overview

This video showcases DVT AI Assistant's inline diagram rendering in several example workflows: design hardware specs with AI-generated FSM and timing diagrams, understand assertions through visual pass and fail scenarios, correlate code insights with simulation log messages to produce message sequence diagrams and timeline Gantt charts.

The video was made using DVT IDE version 26.1.13.

Details

DVT AI Assistant can render WaveDrom timing diagrams and Mermaid diagrams directly in the chat response. Let's take a look at some possible workflows.

Specifying a New Module with AI

Let's ask AI Assistant to plan the design of an SPI block, and explicitly tell it to include a wavedrom diagram for the intended protocol behavior as well as a finite state machine diagram for the actual implementation. This makes understanding and reviewing the AI-generated spec seamless.

Now let's refine the plan by suggesting a new stall state. That's done, so let's proceed to code generation. Finally, let's do a quick visual check against the deterministic DVT-generated state diagram.

Explaining Assertions with Inline Diagrams

Next let's jump to an assertion, and run DVT AI: Explain selected assertion blueprint. A new chat session is started with a carefully crafted recipe which instructs the AI Assistant to generate a detailed explanation of the assertion's behavior. The agent might gather extra information and will eventually produce the requested explanation. Notice how the AI can use WaveDrom to seamlessly present visual pass and fail scenarios for the assertion right inside the chat window.

By the way, blueprints are AI task recipes which you can run using DVT AI Assistant. An ever growing collection of blueprints - such as the one we've just seen - are shipped with the tool, but you can easily modify them or build your own.

Quickly visualize key project insights

In our next example workflow, let's explore a test run from the LowRISC OpenTitan project. Suppose we want to focus on stimulus and scoreboarding, so we lay out an exploration path with clear stages and request a Mermaid message sequence diagram as output.

Notice how AI Assistant makes extensive use of the tools provided by DVT MCP Server running inside the IDE — it gathers compiled files, symbols, identifier information and simulation log messages. It spawns subagents to reduce main context load and stay focused, and eventually generates the diagram inline.

Taking it a step further we ask the agent to map the information it discovered to the simulation log messages, and produce a Mermaid Gantt chart that illustrates key phases and events of the test run. This gives us an incredibly quick high-level overview of what's happening in the test, leveraging the power of large language models to summarize, correlate and aggregate data, deterministic ground truth from DVT IDE MCP tools and AI Assistant's smooth visual integration.