Thread Content
Hello, developers and architects! I have recently developed an automated intelligent chemical engineering software platform based on the Think-Act-Observe cycle, and I would like to invite professionals in the community to review the code and offer suggestions for improvement. Technical highlights: Modular decoupled design – The DocumentParser service (Apache Tika/Camelot), Agent Engine (LLM-based decision-making), and Automation Adapters (DWSIM/Excel/Aspen) are all independently scalable. Think-Act-Observe cycle – A complete closed loop of task understanding → plan generation → execution → result observation. Memory system design – A three-tier architecture comprising session memory, knowledge memory, and execution memory to continuously improve accuracy. Project architecture: (see README). Everyone is welcome to contribute: fork the code and submit PRs; improve the documentation quality; propose requests for new adapters (such as AutoCAD, HYSYS, etc.). We look forward to working together with chemical engineering software developers in the community to build this open-source tool ecosystem! 🚀 Project address: https://github.com/Stanlyjin54/UnveilChem_AiFiller. Current capabilities include: automatic identification of table parameters → automatic filling of Excel from PDFs, parsing of flowcharts → export of structured data, and OCR text recognition (with support for chemical industry terminology). Expected outcome: processing the same document takes only 10 minutes instead of 2 hours, with a accuracy rate of 99%+