GPT4_lll downloads statistics

Features

Agent Chat View with Block Rendering

A completely redesigned chat interface with block-based content rendering. Supports six content block types: Markdown, Code, Thinking Process, Tool Call, Tool Use, and Tool Result. Provides smooth streaming display with smart auto-scrolling and per-turn bubble layout for both user and assistant messages.

Function Calling (Tool Use) Engine

Built-in Function Calling orchestrator enabling AI models to invoke tools within the IDE. Supports multi-round tool call loops (up to 20 rounds), automatic parameter validation, execution with user approval, and result feedback. Compatible with OpenAI, Anthropic and Markdown protocol formats, with built-in degradation fallback mechanism for reliability.

MCP Tool System

Model Context Protocol (MCP) tool system providing ten built-in tools for AI Agent autonomous operation:

• read_file — Read file content with optional line range • write_file — Write files with four modes: overwrite, patch (str_replace), append and insert_after_line, using atomic writes for data safety • grep — Keyword search with context lines, file size limits, directory blacklist and binary file filtering • tree — Project directory tree with configurable ignore patterns • shell_exec — Execute non-interactive local commands with 4-layer security (schema → validator → policy → runtime), structured output including stdout/stderr/exit code/risk level, configurable via .gpt4lll/shell-policy.json • open_files — Inspect currently open editor tabs and identify the active file for context-aware operations • diagnostics — Get compile, lint, and type diagnostics from the IDE's built-in code analysis engine • ask_user — Human-in-the-loop interaction: ask the user for confirmation, clarification, or choice during agent execution • invoke_skill — Dynamically load predefined skill guidance (code review, testing, refactoring, etc.) by task scenario

All tools are sandboxed to the project workspace for security.

Unified LLM Client

Centralized LLM client abstraction layer handling HTTP and SSE communication with multiple providers. Supports streaming chat, synchronous chat, and raw JSON mode for function calling. Provider-specific differences (URL, API key format, message structure, SSE parsing) are handled transparently via adapter pattern.

Conversation Memory Management

Intelligent conversation memory system with multiple strategies: sliding window (token-based truncation), summarizing (LLM auto-summarization when context exceeds threshold), adaptive (similarity-based), and composite (multi-tier). Includes real token usage tracking from API responses.

Runtime Status Indicator

Visual status feedback via animated border: idle (no border), running (animated green sweep), completed (solid green), error (red). A Stop button appears during execution for task cancellation. Status automatically resets after 3 seconds.

Free AIGC Access

The AIGC capability is now freely available to all users. Enjoy intelligent generation services at no extra cost. However, due to high demand, you may need to queue each time you use this feature.

Multi-Platform Support

The plugin supports multiple AI platforms with unified interface and seamless provider switching.

Supported List

• OpenAI (GPT-4, GPT-4o, etc.) • Grok • Tongyi Qianwen (Qwen3-Coder-Plus, Qwen3-Coder-Flash, etc.) • Baidu ERNIE Bot • DeepSeek (V3, R1) • 🔥 Xiaomi MiMo (MiMo-7B-RL, etc.) — Newly integrated! Take advantage of the Creator Incentive Program token price drop! • OpenAI-standard custom API

Code Generation Based on TODO

Introduced the ability to generate code based on TODO comments within a project. This feature enables rapid development by transforming actionable TODOs into functional code segments.

Natural Language Recognition

Improved the logic for natural language recognition, enhancing accuracy and performance. The plugin supports over 100 languages and automatically detects the user's system language to provide responses in their preferred language.

Automatic Code Type Recognition

Enhanced the code type recognition feature to automatically identify code types.

Generate Code On Demand

Implemented the capability to generate code on demand, providing code snippets when needed.

Automatic Code Comments Generation

Added a feature that allows you to auto-generate code comments using AI. By simply highlighting the code, the plugin will call on the AI model to generate comprehensive and contextually appropriate code comments.

Intelligent Comment Translation

Real-time translation for comments in external library files with streaming translation experience. Supports seamless toggling between original and translated content, with regeneration support. Significantly improves code reading and study efficiency.

History Recording Feature

Added a history recording feature to prevent data loss. Every conversation will now be recorded, allowing easy access to previous interactions.

Code Assessment

Code Assessment feature allowing team leaders to evaluate team members' code. Facilitates code reviews, improves code quality, and fosters continuous learning within the team.

Unit Test Generation

One-click unit test generation feature that automatically creates executable unit test files from selected code blocks.

SQL Statement Evaluation and Optimization

Comprehensive SQL statement evaluation and optimization feature that provides syntax correctness checks, performance optimization suggestions, security assessments, index usage recommendations, query efficiency analysis, and best practice suggestions.

AI-Powered Commit Message Generation

Automatically generates standardized commit messages based on Git changes using AI technology.

Git Work Report Generation

Generate detailed work reports by filtering Git commits by author and time range for team collaboration and project management.

Batch File Assessment…

Description from the JetBrains Marketplace listing.