Topic

A Big Language Model-Driven Approach to Building Intelligent Programming Assistants

In today's rapidly evolving world of Artificial Intelligence, big language models are reshaping the paradigm of software development. From the initial code-completion tools, to Prompt-based intelligent assistants, to agents with autonomous planning and execution capabilities, we are experiencing a revolutionary shift in the development paradigm. In this talk, we will discuss how to build a new generation of intelligent programming assistants based on the big language model, and share the complete practical experience from concept to implementation. We will first analyze the pain points of current R&D efficiency, show the limitations of traditional development tools, and the new opportunities brought by the big language model. Then, we will introduce the technical architecture of intelligent programming assistant in detail, including the core design concept of Agent, the engineering capability construction method, and the deep integration practice with IDE. Through the demonstration of real cases, we will show how intelligent assistants can provide effective support in code understanding, generation, optimization and other scenarios. As frontline practitioners, we will also share our experience and lessons learned in the process of engineering landing, including the thinking of architecture selection, prompt optimization, system integration and other key decision points. Finally, we will look forward to the development direction of intelligent programming assistants and discuss how to build a truly intelligent R&D partner. Outline: 1, pain points & problems of intelligent programming assistant -Analysis of R&D efficiency pain points -Limitations of traditional tools -Opportunities brought by the big language model 2、Technical Architecture and Core Capabilities -Agent architecture design -Engineering capability construction -Multi-round dialog and memory management -Code knowledge mapping and context-awareness. -Prompt design and tuning -Code comprehension and generation -IDE system integration 3、Practice effect and case sharing -Typical application scenarios -Demonstration of real cases