Gigasoft Tool Aims to Eliminate AI 'Hallucinations' in Charting Code
Developer tool company Gigasoft has released ProEssentials v10, which includes a Python tool designed to solve errors in AI-generated charting code. The tool, `pe_query.py`, validates AI-written code against the compiled DLL binary to eliminate 'hallucinated' or non-existent property names. The company claims this is the only charting AI tool to offer such validation.
- The `pe_query.py` tool provides AI assistants with access to a knowledge base extracted directly from the compiled library, including 1,104 properties, 167 enumerations, 116 code examples, and 32 knowledge files covering architectural patterns. - The validation method is deterministic, meaning it directly checks every generated property path against the ground truth of the DLL binary, rather than using probabilistic methods like documentation searching to reduce errors. - This AI tool was released as part of ProEssentials v10.0.0.20, which also expanded the library's use of GPU compute shaders, reportedly rendering 100 million data points in approximately 15 milliseconds using a zero-copy data method. - The tool is designed to work with a range of large language models, including Claude, Gemini, GitHub Copilot, and ChatGPT, as well as locally run models. - A 2026 competitive analysis by Gigasoft notes that other charting tools like SciChart are using RAG-based documentation chatbots for AI assistance, while Syncfusion offers broader server extensions across its entire suite. - API hallucination is a frequent issue with AI code generation, where models invent plausible-sounding but non-existent functions or properties, particularly for complex, specialized libraries not heavily represented in training data. - The ProEssentials library maintains broad platform support beyond modern .NET, with interfaces for C++ MFC, Delphi VCL, and ActiveX, catering to development environments with legacy codebases. - The initial maintenance release for ProEssentials v10, version 10.0.0.8, was made available in January 2025, with the AI-focused version 10.0.0.20 released around February 2026.
