AI Assistant for Eclipse

An AI-powered coding assistant plugin for the Eclipse IDE based on a fork of Wojciech Gradkowski's AssistAI.

Table of Contents

• Features
• System Requirements
• Installation
  • Plugin Installation
  • From Source
    • Prerequisites
    • Building
• Quick Start
• Initial Setup
  • Opening the AI Assistant View
  • First-Time Configuration
  • Supported AI Providers
  • Bookmarked API Settings
• Main Interface
  • Chat Conversation Area
  • User Input Area
    • Message History Navigation
    • Spell Checking
    • Context Menu
  • Button Bar
• Navigation & Shortcuts
  • Keyboard Shortcuts
  • Mouse Navigation
• Code Analysis Features
  • Editor Context Menu
  • Adding Context
• Interactive Code Blocks
  • Review Changes Dialog
  • Apply Patch Wizard
• Git Integration
  • Team Context Menu
  • Analysing File-Level Git Diffs
• Chat Area Context Menu
• Advanced Features
  • Reasoning Model Support
  • Usage Reports
  • LaTeX Mathematical Content
  • Conversation Management
• Customisation
  • Prompt Template System
  • JSON/TOML Overrides
  • UI Preferences
• Troubleshooting
  • Blank/Grey Browser Widget
  • Common Issues
• Security Considerations
• FAQ
• Contributing
  • Architecture
  • Adding Features
• Credits
• License

Features

• Multiple AI Provider Support: Works with OpenAI, OpenRouter, Ollama, and other OpenAI-compatible APIs
• Advanced Code Analysis: Comprehensive code review, explanation, debugging, and optimization capabilities
• Interactive Code Blocks: Copy, replace, and review code changes directly from AI responses
• Git Integration: Analyse diffs and generate commit messages from staged changes
• Customisable Prompts: Full template system with context-aware variables
• Reasoning Model Support: Special support for reasoning models with thinking blocks
• Rich Text Rendering: Code syntax highlighting and LaTeX math support
• Conversation Management: Import/export conversations, full undo/redo support

System Requirements

• Eclipse IDE: 2022-03 or later
• Java: 17 or later
• Operating System: Windows, macOS, or Linux
• Memory: Minimum 4GB RAM recommended for large conversations
• Network: Internet connection for cloud AI providers (not required for local providers)

Installation

Plugin Installation

1. Download the latest release:

   • Go to the Releases page
   • Download the latest eclipse.plugin.aiassistant_X.X.X.jar file from the most recent release

2. Install the plugin:

   • Copy the JAR file to your Eclipse dropins folder (usually eclipse/dropins/)
   • Restart Eclipse
   • The plugin will be automatically detected and installed

From Source

Prerequisites

1. Eclipse PDE (Plug-in Development Environment):

   • Go to Help → Eclipse Marketplace
   • Type "Eclipse PDE (Plug-in Development Environment)" in find textbox to search
   • Install "Eclipse PDE (Plug-in Development Environment) latest"

2. EGit - Git Integration for Eclipse:

   • Go to Help → Eclipse Marketplace
   • Type "EGit - Git Integration for Eclipse" in find textbox to search
   • Install "EGit - Git Integration for Eclipse X.X.X"

Building

1. Import the project from GitHub:

   • Go to: File → Import → Git → Projects from Git
   • Select "Clone URI" and click Next
   • Enter https://github.com/jukofyork/aiassistant in the URI field
   • (This should automatically fill the Host field as "github.com" and the Repository path field as "/jukofyork/aiassistant")
   • Leave the Protocol as "https" and the Port blank
   • Enter your GitHub username and a 'Personal Access Token' (NOT your GitHub password, see: https://github.com/settings/tokens)
   • Click finish

2. Build and install the plugin:

   • Go to: File → Export → Plug-in Development → Deployable plug-in and fragments
   • Select "eclipse.plugin.aiassistant (X.X.X qualifier)"
   • Set the destination as "Install into host repository:" and leave as default ".../org.eclipse.pde.core/install/" location
   • Click finish and restart the IDE when asked

Quick Start

1. Install the plugin following the Installation instructions
2. Open the AI Assistant view: Window → Show View → Other... → AI Assistant → AI Assistant
3. Click the "Settings" button and configure your AI provider
4. Right-click in any code editor and try "AI Assistant → Explain"
5. Start coding with AI assistance!

Initial Setup

Opening the AI Assistant View

1. Go to Window → Show View → Other...
2. Expand "AI Assistant" category
3. Select "AI Assistant" and click Open

First-Time Configuration

1. Click the "Settings" button in the AI Assistant view
2. Configure your AI provider settings:
   • Model Name: The specific model to use (e.g., llama3.1:8b, gpt-4, claude-3-sonnet)
   • API URL: The base URL for your AI provider
   • API Key: Your authentication key (leave blank for local providers like Ollama)
   • JSON Overrides: Additional API parameters in JSON or TOML format

Supported AI Providers

The plugin supports multiple AI providers with pre-configured examples:

Provider	Example URL	Notes
OpenAI	https://api.openai.com/v1	Supports all OpenAI's models including reasoning models
OpenRouter	https://openrouter.ai/api/v1	Access to Anthropic's Claude and other models
Ollama	http://localhost:11434/v1	Local model hosting
llama.cpp	http://localhost:8080/v1	Local model hosting
TabbyAPI	http://localhost:5000/v1	Local model hosting

Bookmarked API Settings

Use the bookmarked settings table to quickly switch between different AI providers and models:

• Bookmark: Save current settings for easy switching
• Populate: Automatically discover available models from your API provider
• Sort: Organise bookmarks alphabetically
• Clear: Remove all bookmarked settings

Main Interface

Chat Conversation Area

The main chat area displays conversations between you and the AI assistant. Messages are color-coded:

• User messages: Darker background
• Assistant responses: Lighter background (darker for collapsible thinking blocks)
• Code blocks: Black background with syntax highlighting based on the detected language
• Notifications: Red background for errors/warnings and blue background for status updates

User Input Area

The user input area provides a rich text editing experience with navigation and spell-checking capabilities:

Message History Navigation

The yellow arrow buttons on the right side of the input area allow you to navigate through your message history:

This feature helps you quickly access and reuse previous prompts without retyping them, and the history is saved between sessions.

Spell Checking

The input area includes built-in spell checking that highlights misspelled words with red wavy underlines.

To correct a misspelled word:

1. Left-click on the misspelled word to position your cursor
2. Select the correct spelling from the suggested alternatives

Context Menu

The Right-click context menu provides standard text editing operations including Undo, Redo, Cut, Copy, Paste, and Select All:

Button Bar

• Stop: Cancel current AI request
• Clear: Clear conversation history
• Undo: Remove last interaction
• Redo: Restore undone interaction
• Import: Load conversation from JSON file
• Export: Save conversation as JSON or Markdown
• Settings: Open preferences dialog

Navigation & Shortcuts

Keyboard Shortcuts

Shortcut	Action
Enter	Send message and request AI response
Shift+Enter	Insert newline
Ctrl+Enter	Send message without requesting immediate response
Ctrl+Z	Undo text changes
Ctrl+Shift+Z	Redo text changes

Mouse Navigation

• Ctrl+Scroll: Navigate to top/bottom of conversation
• Shift+Scroll: Navigate between messages
• Shift+Hold: Highlight current message with a blue border whilst scrolling

Code Analysis Features

Editor Context Menu

Right-click in any code editor to access AI-powered analysis tools:

Code Understanding

• Discuss: Open-ended discussion about code
• Explain: Detailed explanation of code functionality

Code Quality & Review

• Code Review: Comprehensive code analysis and suggestions
• Best Practices: Check adherence to coding standards
• Robustify: Improve error handling and edge cases
• Debug: Identify potential bugs and issues

Code Transformation

• Optimize: Performance improvement suggestions
• Refactor: Restructure code for better maintainability
• Write Comments: Generate documentation and comments
• Code Generation: Create boilerplate code
• Code Completion: Fill in missing implementations

Error Resolution

When Eclipse detects compiler errors or warnings, additional options appear:

• Fix Errors: Address compilation errors
• Fix Warnings: Resolve compiler warnings
• Fix Errors and Warnings: Address both simultaneously

Adding Context

• Add To Message: Include code/file in your message
• Add As Context: Provide background information for AI analysis
• Add Staged Diff: Include Git changes in your request

Interactive Code Blocks

AI responses containing code include interactive buttons:

• Copy Code: Copy to clipboard
• Replace Selection: Replace selected code in editor
• Review Changes: Open side-by-side comparison

• Apply Patch: Use Eclipse patch wizard (for diff blocks)

Review Changes Dialog

The Review Changes feature opens Eclipse's compare editor for selective code application:

Apply Patch Wizard

For code blocks marked as diff, use Eclipse's built-in patch application:

If any hunks fail to match, you'll get a warning notification that can be pasted back to the AI to help diagnose the problem:

You can either partially apply the successful hunks or ask for a complete redo depending on your choice in the patch wizard.

Git Integration

Team Context Menu

Access Git-specific features through the Team menu:

• Add Staged Diff: Analyse current project-level staged changes to help review modifications
• Git Commit Comment: Automatically generate descriptive commit messages based on staged changes

Analysing File-Level Git Diffs

The AI can also analyse file-level Git differences using the right-click context menu:

Chat Area Context Menu

Right-click in the chat area for additional options:

• Copy: Copy selected text to clipboard
• Replace Selection: Replace editor selection with chat text
• Review Changes: Compare chat text with editor selection
• Paste To Message: Add clipboard content as message
• Paste As Context: Add clipboard content as context

The first three options are only enabled when a section of code is selected. These trigger the same action as the buttons, but for the selected section of code only:

Advanced Features

Reasoning Model Support

Special support for reasoning models with collapsible thinking blocks:

When using reasoning models (such as o3, deepseek-reasoner, claude-sonnet-4, and similar models), the AI Assistant automatically detects and displays the model's internal reasoning process in collapsible "Thinking..." blocks. These blocks contain the step-by-step reasoning that the model performs before providing its final answer.

Key Features:

• Collapsible Display: Thinking blocks are shown as collapsible sections that can be expanded or collapsed
• Automatic Detection: The plugin automatically identifies reasoning content via simple <think> tags or from reasoning/reasoning_content sections in the returned response
• Clean Conversation History: Thinking blocks are automatically removed from messages before they're sent back to the AI model, ensuring the conversation history remains clean and focused on the actual dialogue

NOTE: Some providers such as OpenAI and Google do not return their reasoning tokens.

Usage Reports

After each AI response, a detailed usage report is displayed in the notification area:

The usage report provides comprehensive information about the API request:

Model Information:

• Model: The exact model name used for the request
• Finish: Reason the response ended (stop, length, content_filter, etc.)

Token Usage:

• Prompt: Input tokens consumed with percentage of model's context window used
• Cached: Cached tokens from previous requests (shown in brackets when available)
• Response: Output tokens generated with percentage of model's output limit used
• Reasoning: Additional reasoning tokens (shown in brackets for reasoning models)

Cost Analysis:

• Charge: Total cost for the request with breakdown showing prompt cost + response cost
• Uses cents (￠) for amounts under $1.00 and dollars (＄) for larger amounts
• Cost data may not be available for all models/providers

Availability Notes:

• Token percentages require model context limits to be known
• Reasoning tokens only appear for reasoning-capable models
• Cached tokens depend on provider support for prompt caching
• Cost information requires pricing data in the model database

LaTeX Mathematical Content

Render mathematical formulas and equations:

Conversation Management

• Import/Export conversations as JSON
• Export conversations as Markdown

Customisation

Prompt Template System

Customise all prompt templates through the preferences:

Template Variables

Templates use the StringTemplate syntax with these context variables:

Variable	Description
<taskname>	Name of the current prompt task
<usertext>	User input from the chat interface
<project>	Current project name
<filename>	Active file name
<language>	Programming language of active file
<tag>	Markdown language tag for syntax highlighting
<warnings>	Compiler warnings for active file
<errors>	Compiler errors for active file
<document>	Full text of active document
<clipboard>	Current clipboard contents
<selection>	Currently selected text (or full document if none)
<lines>	Description of selected line numbers
<documentation>	Appropriate documentation format for language
<file_diff>	Git diff for current file
<project_diff>	Git diff for entire project

Role Switching

Use the special <<switch-roles>> tag to create multi-turn conversations:

Analyze this code:
```java
<selection>
```
<<switch-roles>>

This creates alternating user and assistant messages for complex interactions.

JSON/TOML Overrides

Configure advanced API parameters using either JSON or TOML syntax:

JSON Examples:

"temperature": 0.7, "max_tokens": 2000, "top_p": 0.9

TOML Examples:

temperature = 0.7, max_tokens = 2000, reasoning_effort = "high"

UI Preferences

Customize interface settings:

• Font Sizes: Adjust chat and notification text size
• Timeouts: Configure connection and request timeouts
• Streaming: Enable/disable real-time response streaming
• Message Types: Choose between System and Developer messages

Troubleshooting

Blank/Grey Browser Widget

If the chat area appears blank or grey, set these environment variables before starting Eclipse:

Linux/macOS:

export WEBKIT_DISABLE_COMPOSITING_MODE=1
export WEBKIT_DISABLE_DMABUF_RENDERER=1
./eclipse

Windows:

SET WEBKIT_DISABLE_COMPOSITING_MODE=1
SET WEBKIT_DISABLE_DMABUF_RENDERER=1
start eclipse.exe

Common Issues

• Connection Errors: Verify API URL and key in settings
• Model Not Found: Ensure model name matches provider's available models
• Slow Responses: Increase request timeout in preferences
• Large Conversations: Use Clear button periodically for better performance
• Memory Issues: Restart Eclipse if conversations become very large
• SSL/TLS Errors: Check firewall and proxy settings

Security Considerations

Important: This plugin sends your code to external AI services. Consider the following:

• Sensitive Code: Be cautious when analyzing proprietary or sensitive code
• Local Providers: Use Ollama or llama.cpp for complete privacy
• API Keys: Store API keys securely and rotate them regularly
• Network: Use HTTPS endpoints and consider VPN for additional security

FAQ

Q: Can I use this plugin offline?

A: Yes, with local providers like Ollama or llama.cpp.

Q: How do I switch between different AI models quickly?

A: Use the bookmarked API settings feature to save and switch between configurations.

Q: Can I customise the prompts?

A: Yes, all prompts are fully customisable through the Prompt Templates preferences page.

Contributing

Architecture

The plugin follows Eclipse's standard architecture:

• Main View: MainView and MainPresenter handle UI and business logic
• Browser Integration: Custom browser functions for code interaction
• Network Layer: OpenAI-compatible API client with streaming support
• Prompt System: Template-based prompt generation with context injection

Adding Features

1. New Prompt Types: Add to Prompts enum and create template file
2. Browser Functions: Extend DisableableBrowserFunction for new interactions
3. Context Variables: Modify Context class to add new template variables

Credits

• Wojciech Gradkowski for the original AssistAI project that this plugin is based on
• The Highlight.js team for code block syntax highlighting
• The MathJax team for the mathematical notation rendering
• The LiteLLM team for the model_prices_and_context_window.json file this project uses to calculate the API charges and context windows for the usage reports

License

The MIT License is inherited from Wojciech Gradkowski's original AssistAI project - see LICENSE for details.