regyssilveira/radia-plugin

RadIA is a plugin for the Embarcadero Delphi programming environment that adds an AI chat assistant directly inside the IDE. Delphi is a development tool for building Windows applications, and its IDE is the workspace where developers write and compile their code. This plugin attaches a chat panel to the sidebar of that workspace so you can ask AI questions without leaving the editor. The README is primarily in Brazilian Portuguese. The chat panel is rendered using a built-in Edge browser component, which lets it display formatted messages with code highlighting and Markdown support. It connects to whichever AI provider you configure: supported options include Google Gemini, OpenAI, Anthropic Claude, DeepSeek, Groq, and local models run through a tool called Ollama. You bring your own API key for each service. Conversation history is saved automatically to a local file so you can pick up where you left off after closing and reopening the IDE. Beyond basic chat, the plugin integrates with the Delphi code editor in several ways. You can select any block of code, right-click it, and get options to explain what it does, improve or reorganize it, generate unit tests for it, or look for bugs. When the plugin suggests a rewritten version, it shows a side-by-side comparison highlighting removed lines in red and new lines in green, with a single button to apply the change directly to your file. The plugin also connects to the compiler output panel, so you can right-click a build error and ask the AI to explain and suggest a fix. Additional features include automatic generation of structured documentation comments above method definitions, and a converter that can produce Delphi data classes from a JSON example or an SQL table definition. Slash commands in the chat box give quick access to common actions. The plugin requires Delphi version 10.4 or later and Windows 10 or 11. Installation can be done through a PowerShell script or manually through the IDE's package manager. Adding support for a new AI provider involves registering a new class in a metadata-driven registry rather than modifying the core plugin code.