Overview

The Code with AI feature in Softaculous is an AI-powered coding assistant integrated directly into your control panel. It helps you write, debug, refactor, and understand code in your projects without leaving your hosting environment.

Note: This feature is added in Softaculous 6.3.4

Key Features

• Multi-provider support – Connect OpenAI, Anthropic, Google, DeepSeek, Groq, Together AI, OpenRouter, Ollama, MiniMax, and custom OpenAI-compatible providers
• Free tier – OpenCode Zen (Free) is available with no API key required
• Build & Plan modes – Full-access coding or read-only analysis
• Dark / Light theme – Toggle with moon/sun icon in the header
• Command Palette – Press Ctrl+K to quickly access any command
• Slash Commands – 16+ slash commands for quick actions
• Inline Mode Switch – Type @build or @plan to switch modes inline
• Shell Mode – Type ! to run shell commands directly
• Session Timeline – Jump-to-message navigation
• Session Search – Search and filter through sessions
• Session Fork/Branch – Explore alternate conversation paths
• Undo/Redo – Undo or redo messages
• Regenerate Response – Re-generate the last AI response
• Edit Previous Messages – Edit any prior user message and re-submit
• Conversation Compact – Summarize long conversations to save context
• Model Favorites – Star your favorite models for quick access
• Model Variants – Adjust reasoning effort (Low/Medium/High)
• Context Window Tracking – Visual bar showing token usage percentage
• Browser Notifications – Browser notifications when tab is not focused
• Sound Notifications – Audio alerts for completion, errors, and questions
• Modified Files Panel – See all files changed by AI in the session
• Task Management – Built-in todo list with checkbox tracking
• Reasoning Display – Toggle thinking blocks (show/collapse/hide)
• Prompt History – Navigate through previous prompts with arrow keys
• Auto-Generated Titles – Sessions get AI-generated descriptive titles
• Customizable Keybindings – Remap any keyboard shortcut
• Project Management – Create and switch between multiple projects
• File Tree Browser – Browse and view project files
• Export as Markdown – Download or copy your conversation
• Permission System – Control tool access (Allow Once, Allow Always, Deny)
• Read & Write Files – AI can read and write files directly in your project directory
• Write Plugins – Generate complete WordPress plugins from scratch with a single prompt
• Make Websites – Create full websites, themes, and templates with AI assistance
• Edit Existing Files – Make targeted changes to existing files without rewriting everything
• Search Files – Find files by name or search content across your entire project
• Debug & Refactor – Identify bugs, fix errors, and refactor code for better structure

Accessing the Code with AI Feature

Log in to Softaculous

1. Log in to your hosting control panel (Webuzo, cPanel, DirectAdmin, etc.)
2. Navigate to the Softaculous section or icon
3. The Softaculous dashboard will open, In the Softaculous header/navigation, look for the Code with AI link
4. Click on it to open the AI Assistant interface

Theme (Dark / Light Mode)

Code with AI supports both light and dark themes. Click the moon/sun icon in the top-right corner of the header to toggle between themes.

• Moon icon (🌙) – Currently in light mode, click to switch to dark
• Sun icon (☀️) – Currently in dark mode, click to switch to light
• Your theme preference is saved in the browser and persists across sessions

Setting Up the AI Assistant

Connecting an AI Provider

Free Option:

• OpenCode Zen (Free) – Enabled by default. No API key required.

You can also connect to any other AI provider:

1. Click the Settings button in the AI Assistant header
2. Find the provider you want to use and click Connect
3. Enter your API key and click Connect
4. Use the Test button to verify your connection works

Custom OpenAI-Compatible Providers

Code with AI supports any OpenAI-compatible API endpoint as a custom provider. This means you can connect to self-hosted models, private inference servers, or any third-party API that implements the OpenAI chat completions format (/v1/chat/completions).

How to Add a Custom Provider

1. Open Settings in the Code with AI interface
2. Click + Add Connection 
3. A modal opens – select Custom Provider (OpenAI Compatible) from the dropdown
4. Fill in the following fields:
   • Provider Name – A custom name for your provider (e.g., “My Local LLM”)
   • Base URL – The API endpoint URL (e.g., https://api.myhost.com/v1)
   • API Key – Your API key (if required by the endpoint)
   • Models – Comma-separated list of models in id=name format (e.g., gpt-4o=GPT-4o, llama3=Llama 3)
5. Click Test Connection to verify connectivity
6. Click Connect to add the provider

Model Format

Models can be specified in two ways:

• With custom display name: model_id=Display Name (e.g., llama3-70b=Llama 3 70B)
• With model ID only: model_id (display name defaults to the ID)

Multiple models are separated by commas:

gpt-4o=GPT-4o, llama3=Llama 3, mistral=Mistral 7B

Use Cases

• Self-hosted LLMs – Connect to vLLM, text-generation-inference, or LM Studio servers
• Private cloud endpoints – Connect to your organization’s internal API gateway
• Regional providers – Connect to providers not listed in the built-in list
• Proxy gateways – Connect through an OpenAI-compatible proxy or load balancer

Requirements

• The endpoint must implement the OpenAI Chat Completions API format (POST /chat/completions)
• The endpoint must support Server-Sent Events (SSE) for streaming responses
• The Base URL should include the API version path (e.g., /v1) — Code with AI automatically appends /chat/completions
• API key authentication is sent as a Bearer token in the Authorization header

AI Provider Hook

The softaculous_ai_providers filter is already built into Softaculous. Hosting providers can use this filter to add custom AI providers to the providers list shown to the user:

insert_filter('softaculous_ai_providers', 'my_ai_providers', 1, 1);

function my_ai_providers($providers){

	$providers[] = array(
		'id' => 'custom:myhost-ai',
		'name' => 'My Host AI (Premium)',
		'logo' => 'custom',
		'auth_type' => 'api_key',
		'api_key' => 'sk-myhost-predefined-key-change-me',
		'base_url' => 'https://api.myhost.com/v1',
		'models' => array(
			'myhost-large' => 'MyHost Large',
			'myhost-fast' => 'MyHost Fast'
		),
		'default_model'	=> 'myhost-large'
	);
        return $providers;
}

Supported AI Providers:

Model Auto-Detection

Models are automatically detected and cached for 24 hours. The system fetches context window limits, output limits, reasoning capabilities, and tool-call support for each model automatically. You do not need to manually configure model metadata.

Selecting an AI Model

1. Once connected, choose an AI model from the custom dropdown in the header
2. The dropdown is organized by provider with group headers
3. The selected model will be used for all AI interactions in your current session

Model Favorites

Add your frequently used models to Favorites for quick access :

• Click the heart icon(❤️) next to any model in the dropdown to toggle it as a favorite
• Favorite models (marked with ❤️) appear at the top of the model dropdown
• Click the heart again to remove a model from favorites
• Favorites are saved per-user and persist across sessions

Model Variants (Reasoning Effort)

For supported models (Claude, GPT-4o, Gemini, DeepSeek, OpenCode Zen, etc), you can adjust the reasoning effort to control how deeply the model thinks before responding.

• Default – Standard reasoning (recommended)
• Low – Faster responses, less deep thinking
• Medium – Balanced reasoning
• High – Deepest reasoning, best for complex problems

How to change:

• Press Ctrl+Shift+V to cycle through variants quickly

Context Window Tracking

The status bar at the bottom of the screen shows a context window indicator with:

• Token count – Current tokens used
• Visual progress bar – Color-coded (green <75%, amber 75-90%, red >90%)
• Percentage – Context usage as a percentage of the model’s limit

Creating a Project

The AI Assistant works within projects, which are linked to directories on your server.

Create a New Project

1. Click Projects in the AI Assistant header
2. Click + New Project
3. Choose one of the following options:
   • Select an existing WordPress installation from the list
   • Or enter a custom directory path manually
4. The project will be created and linked to the selected directory

Switch Between Projects

• Use the project dropdown in the header (next to the sidebar toggle button) to switch between projects
• Each project has its own chat sessions, conversation history, and file tree

Using the AI Assistant

Starting a Chat Session

1. Select your project from the dropdown
2. Choose an AI model from the header dropdown
3. Type your request in the chat input at the bottom
4. Press Enter to send (or Shift+Enter for a new line)
5. The AI will respond and can perform actions like editing files or running commands

Creating New Sessions

• Click + New Session to start a fresh conversation
• Previous sessions are listed in the sidebar
• Click on any previous session to resume it
• Sessions are automatically grouped by date (Today, Yesterday, older dates)
• Press Alt+1 through Alt+9 to quickly switch between the first 9 sessions

Session Management

Session Timeline

The Timeline panel shows all user messages in the current conversation, allowing you to quickly jump to any point in the conversation.

• Click the chevron icon ( >) on the left edge of the screen to open the timeline
• Or use the slash command /timeline
• Click any message in the timeline to scroll directly to it

Session Search

Use the search box at the top of the sidebar to filter sessions by title or content:

• Click the search icon at the top of the sidebar
• Type your search query
• Sessions matching your query will be highlighted

Session Rename

• Double-click on any session title in the sidebar to rename it
• Type the new name and press Enter to confirm
• Or use the slash command /rename to rename the current session

Auto-Generated Session Titles

When you start a new conversation, the AI automatically generates a short, descriptive title based on the content of your first message and the AI’s response. This makes it easy to identify sessions at a glance.

Delete Sessions

• Click the × button next to any session in the sidebar to delete it
• Deleted sessions cannot be recovered

Command Palette

Press Ctrl+K to open the Command Palette – a quick-access search interface for all available commands.

• Type to search for commands like “New Session”, “Plan Mode”, “Compact”, “Undo”, “Fork”, “Settings”, etc.
• Use arrow keys to navigate the results
• Press Enter to execute the selected command
• Press Escape to close the palette

Slash Commands

Type / in the chat input to see available commands. A dropdown will appear showing matching commands as you type.

Command	Description
/build	Switch to Build mode (full access)
/plan	Switch to Plan mode (read-only)
/clear	Clear current conversation
/export	Download conversation as Markdown
/copy	Copy transcript to clipboard
/compact	Summarize conversation to reduce context
/undo	Undo last message pair
/redo	Redo last undone message
/fork	Fork conversation from last message
/rename	Rename this session
/todo	Add a task item (e.g. /todo Fix login bug)
/sound	Toggle sound notifications on/off
/notify	Toggle browser notifications
/keybindings	Customize keyboard shortcuts
/timeline	Toggle message timeline navigation
/help	Show keyboard shortcuts and commands

Inline Commands (@build / @plan)

You can switch modes inline while typing a message:

• Type @build in your message to switch to Build mode
• Type @plan in your message to switch to Plan mode
• A dropdown will appear showing matching mode options
• Select an option to switch modes and continue your message

Shell Mode

Run shell commands directly from the chat input without the AI:

• Type ! followed by a command (e.g., !ls -la, !cat index.php)
• The command executes immediately with terminal-style output
• Output is displayed in a styled terminal block with $ prefix
• Whitelisted commands include: ls, cat, php, node, npm, composer, and more

Build Mode vs. Plan Mode

The AI Assistant has two operating modes:

Build Mode

• The AI has full access to read, write, edit files, and run commands
• Use this when you want the AI to make changes to your project
• Default mode for active development

Plan Mode

• The AI operates in read-only mode
• It can explore and analyze your code but cannot modify anything
• Use this for:
  • Code reviews
  • Explanations
  • Planning changes before implementing them

Switching Modes

• Use the mode dropdown in the header
• Or use slash commands: /build or /plan
• Or use inline commands: @build or @plan
• Or use keyboard shortcuts: Alt+Shift+B (Build) or Alt+Shift+P (Plan)

What the AI Can Do

File Operations

• Read files – View file contents
• Write files – Create new files
• Edit files – Make targeted changes to existing files
• Search files – Find files by name or content (glob/grep)

System Operations

• Run shell commands – Execute commands in your project directory
• Create directory listings – View project structure

Code Assistance

• Write new code
• Troubleshoot website issues
• Debug existing code
• Refactor and optimize code
• Explain code functionality
• Generate code snippets

Reasoning / Thinking Display

When using any AI model, the AI’s thinking process is displayed in a special reasoning block.

• Show – All reasoning blocks are expanded and visible
• Collapse – Reasoning blocks are collapsed (click to expand individually)
• Hide – All reasoning blocks are hidden

Click the Reasoning button in the header to cycle through these states. The button only appears when reasoning blocks are present in the conversation.

Modified Files Panel

Click the Modified Files button in the header to see a list of all files that the AI has written, edited, or deleted during the current session.

• Green badge (added) – New files created by the AI
• Blue badge (modified) – Existing files edited by the AI
• Red badge (deleted) – Files deleted by the AI

Task Management (Todo List)

Click the Tasks button in the header to open the task management panel. The AI can create and track tasks as it works on your project.

• Add tasks manually with /todo <task text>
• Click the checkbox to mark tasks as done/undone
• Tasks are persisted in the browser and saved per-session

Browser & Sound Notifications

Browser Notifications

Get browser notifications when the AI finishes responding, even when the tab is not focused.

• Enable with /notify slash command or Alt+Shift+D shortcut
• Click the notification to return to the Code with AI tab
• Notifications appear for: AI response ready, errors, and AI questions

Sound Notifications

Audio alerts provide feedback without watching the screen:

• Completion chime – Ascending tones when the AI finishes
• Error chime – Descending tones on errors
• Question chime – Three rising notes when the AI asks a question
• Toggle with /sound slash command or Alt+Shift+O shortcut
• Sound notifications are enabled by default

Keyboard Shortcuts

Code with AI supports a comprehensive set of keyboard shortcuts for efficient navigation:

Shortcut	Action
Ctrl+K	Open command palette
Ctrl+Shift+V	Cycle model variant (reasoning effort)
Alt+1 – Alt+9	Switch to session 1–9
Alt+Shift+N	New session
Alt+Shift+P	Switch to Plan mode
Alt+Shift+B	Switch to Build mode
Alt+Shift+S	Open Settings
Alt+Shift+L	Toggle sidebar
Alt+Shift+F	Toggle file tree panel
Alt+Shift+T	Toggle tasks panel
Alt+Shift+M	Toggle modified files panel
Alt+Shift+C	Compact conversation
Alt+Shift+E	Export as Markdown
Alt+Shift+X	Clear conversation
Alt+Shift+H	Open Help & FAQ
Alt+Shift+O	Toggle sound notifications
Alt+Shift+D	Toggle browser notifications
Alt+Shift+K	Customize keybindings
Alt+Shift+Y	Copy transcript
Escape	Stop streaming / Close modals
Enter	Send message
Shift+Enter	New line in input
↑ (at start of input)	Previous prompt in history
↓ (at end of input)	Next prompt in history
! (at start of input)	Shell mode prefix

Customizable Keybindings

All keyboard shortcuts can be customized to your preference:

1. Press /keybindings or Alt+Shift+K to open the keybindings settings
2. Each shortcut is displayed with its current key combination
3. Click Edit next to any shortcut
4. Press your desired key combination
5. The new binding is saved automatically
6. Click Reset to Defaults to restore all shortcuts

Conversation Fork / Branch

Fork a conversation to explore an alternate path without losing your current chat:

1. Type /fork or press Ctrl+K and select Fork Conversation
2. A modal shows all user messages in the conversation
3. Click Fork entire session to copy the full conversation, or click a specific message to fork from that point
4. A new session is created with all messages up to the fork point
5. The forked session gets its own title and can be edited independently

Session Compact & Summarize

When your conversation gets long, compact it to reduce token usage:

• Type /compact or press Alt+Shift+C
• The AI summarizes earlier messages while keeping the most recent 3 messages intact
• This reduces context usage and keeps the conversation within the model’s context window

Undo / Redo Messages

Made a mistake or want to rephrase? Use undo and redo:

• Undo (/undo) – Removes the last user message and all responses after it
• Redo (/redo) – Restores the last undone message pair
• You can undo multiple times to go back through the conversation

Regenerate Response

If you’re not satisfied with the AI’s response, you can regenerate it:

• Hover over the last assistant message
• Click the regenerate button (circular arrow icon) that appears
• The AI will generate a new response to your last message
• Useful for getting alternative approaches or fixes

Edit Previous Messages

You can edit any previous user message and re-submit it:

1. Hover over any user message in the conversation
2. Click the pencil icon (🖉) that appears
3. Edit the text inline
4. Click Send to submit the edited message

The edited message is sent as a new prompt. Existing messages are preserved below it.

Export Conversation

Export your conversation for documentation or sharing:

• Export as Markdown (/export) – Downloads a complete Markdown file with all messages, including reasoning blocks, tool calls, and code blocks
• Copy to Clipboard (/copy) – Copies the full transcript to your clipboard
• Both options include the full conversation with proper formatting

Prompt History Navigation

Quickly access and re-use your previous prompts:

• Press ↑ (Up arrow) at the start of the input to cycle through previous prompts
• Press ↓ (Down arrow) to move forward in history
• Or use Ctrl+↑ / Ctrl+↓ to cycle history at any cursor position
• Prompt history is saved per-session and persists across browser sessions

Best Practices

When to Use Build Mode

• Writing new code
• Fixing bugs
• Refactoring existing code
• Running commands to set up your project
• Making configuration changes

When to Use Plan Mode

• Reviewing code before making changes
• Understanding complex code
• Planning architectural changes
• Getting explanations of how code works
• Analyzing potential issues

Writing Effective Prompts

1. Be specific about what you want the AI to do
2. Provide context about your project structure
3. Mention file names when referring to specific files
4. Break complex tasks into smaller steps
5. Review AI changes before accepting them

Session Organization Tips

• Use descriptive session names (rename with /rename)
• Fork conversations when exploring different approaches
• Compact long sessions to stay within context limits
• Use Plan mode for exploration and Build mode for implementation
• Create separate projects for different codebases

Security Considerations

• Review all file changes before applying them
• Don’t share sensitive API keys or passwords in chat
• Use Plan mode when exploring unfamiliar code
• The AI will never reveal API keys, credentials, or secrets
• Tool isolation ensures the AI can only access allowed directories

Troubleshooting

AI Provider Connection Issues

Problem: Cannot connect to AI provider.

Solutions:

1. Verify your API key is correct
2. Check if your API key has available credits/quota
3. Use the “Test” button in Settings to verify connectivity
4. Use OpenCode Zen (Free) if you don’t have an API key

AI Not Responding

Problem: Messages are not getting responses.

Solutions:

1. Check your internet connection
2. Verify the selected AI model is available
3. Check if you’ve reached API rate limits
4. Try switching to a different AI provider

File Access Issues

Problem: AI cannot read or write files.

Solutions:

1. Verify the project directory path is correct
2. Check file permissions on your server
3. Ensure the directory exists and is accessible
4. Try creating a new project with the correct path

Session Management

Problem: Cannot find previous conversations.

Solutions:

1. Check the sidebar for previous sessions
2. Use the search box to filter sessions
3. Ensure you’re in the correct project
4. Sessions are project-specific – switch projects if needed

Context Window Full

Problem: The AI stops responding or shows errors about context length.

Solutions:

1. Use /compact to summarize the conversation
2. Start a new session for a fresh context
3. Use /fork to continue from a specific point

Sound / Notification Not Working

Problem: No audio or browser notifications.

Solutions:

1. Check /sound status – ensure sound notifications are ON
2. Check /notify status – ensure browser notifications are ON
3. Allow browser notifications when prompted
4. Check browser settings for notification permissions