AI Assistant

The AI Assistant is a chat-based assistant integrated into DaVinci Configurator. It answers your questions in natural language using Vector product documentation, so you can find information without searching through several documents manually. It can also read your ECU configuration and related information to provide answers that are specific to your DaVinci Project.

Data Sources

The AI Assistant draws its answers from documents that are part of your BSW Package or other sources you have access to.

Example Sources:

  • Publicly available documentation in Vector Embedded Hub (for example AUTOSAR documentation)

  • Technical References

  • User Manuals

The AI Assistant can only answer questions that are covered by the selected data source.

Use the AI Assistant

Prerequisites

The delivered AI Assistant is a preview of the 2027 version.
From July 2027, it is available only as a Subscription, not as part of a Perpetual license.

To use the AI Assistant, you need the following:

  • A valid DaVinci Configurator license that allows AI features to be used.

  • A Vector Embedded Hub account and a token.

  • An OpenAI-compatible endpoint (for example Qwen or GPT).

LLM Requirements

Native tool/function calling over an OpenAI-compatible /v1/chat/completions endpoint is required, with automatic tool choice enabled on the serving side. A context window size of at least 32,000 is required; 64,000 or higher is recommended.

Steps

  1. Open View  AI Assistant.

  2. Type your question in plain language on the opened AI Assistant window.

    The AI Assistant displays the answer together with a link to the sources it drew the answer from.

Always check the referenced sources to verify the answer.

+ As with any AI assistant, answers can occasionally be incomplete or inaccurate. Treat the AI Assistant as a fast way to locate documented information, not as an authoritative source in itself.

Rules for Effective Questions

The following rules will help you get better answers through specific questions.

Rules

  • Name the module, parameter, function, or exact error message you are asking about.

  • Ask one thing at a time.

  • Use follow-up questions to refine or narrow down an answer.

  • Open a new tab or session whenever your next question is unrelated to the previous one. This keeps the context clean.

Example Questions

  • How do I configure event memory in the Dem module?

  • What does <error message> during generation mean and how do I resolve it?

  • Which parameters control <feature>?