Pre loader

SciChart GPT is out of Beta! Version 1.0.0 Released

SciChart GPT is out of Beta! Version 1.0.0 Released

We’re excited to announce a major update to the SciChart AI Assistant, designed to deliver more accurate answers, better code samples and a smoother user experience.

The SciChart-GPT serves retrieval augmented generation (RAG) results from the entire database of SciChart content: over 25,000 pages of documentation, demos/examples, blog posts, FAQs and forum knowledgebase across all products/platforms.

Ask any question about SciChart WPF, SciChart.js, SciChart iOS / Android, how to build something, if a feature is supported or technical questions, and the AI Assistant will respond with accurate, up to date results.

Read what’s new below!

What’s new?

Dramatically improved relevance, accuracy and fewer hallucinations

Our upgraded retrieval and ranking pipeline reduces hallucinations and delivers far more accurate, context-aware code examples across all platforms.

A lot of work has gone into the back-end database to ensure that we serve relevant, up to date content with more complete documentation data. Front-end query processing improvements ensure more accurate interpretation of technical questions, with fewer code / API hallucinations.

Smarter AI model selection

Guest users now use a lightweight 10B class model for fast responses, while signed-in users are automatically routed to a 300B+ class model for deeper reasoning, more accurate technical guidance and richer code samples.+

SciChart GPT now features a Weak model (10bn parameters) vs. Strong model (300bn parameters) for guest and signed-in users respectively
Weak model (10bn parameters) vs. Strong model (300bn parameters) for guest and signed-in users respectively

For the best results and most accurate answers: sign in when using SciChart GPT. It’s free!

Intelligent query reformulation

Multi-turn questions are automatically rewritten into a single, precise query – leading to dramatically improved answer quality.

For example, if you ask:

  • “How do I make a bubble chart?” -> the assistant may serve a multiple answers (WPF, iOS, Android, JS) or choose a random platform as it does not know the platform/product
  • If you then ask “… I’m using JS” -> the assistant will reformulate the query to “How to create a bubble chart in SciChart.js”
  • Adding “…with scichart-react” -> will reformulate the query to “How to create a bubble chart in JavaScript using scichart-react”
SciChart-GPT now features query reformulation, improving multi-turn AI Chat answers by tracking long term intent
SciChart-GPT now features query reformulation, improving multi-turn AI Chat answers by tracking long term intent

This query reformulation adjusts if you switch context mid-stream, e.g. Mentioning

  • “… and how do you do that in WPF” -> will then query for “How to create a bubble chart in SciChart WPF”

Try it out! chat.scichart.com

Better platform intent detection

In the Beta, many users reported that the platform intent would lose track of what product you were working on. For example, if you ask for WPF answers, the AI Assistant could start giving you SciChart.js or even iOS/Android answers and code samples after a few successive turns.

Now, when you mention a platform (or if a platform/language is detected), the assistant shifts to documentation and examples for that platform only, and persists the platform intent until you redirect it.

If you switch context mid-stream to ask about another product, the assistant will adjust the context from that point onward.

New floating action menu with live sales chat

On the website, we’ve integrated a floating-action menu alongside the AI Assistant with the option to contact sales via live-chat. Now you can bug our sales team while GPT-ing for answers!

UI/UX improvements

Smarter formatting, cleaner output and a better mobile-friendly layout.

By Andrew Burnett-Thompson | Dec 05, 2025
CEO / Founder of SciChart. Masters (MEng) and PhD in Electronics & Signal Processing.Follow me on LinkedIn for more SciChart content, or twitter at @drandrewbt.

Leave a Reply