AI Response Generator Plugin for osTicket (Enhanced Fork)

Note: This is a fork of mhajder/ai-response-generator with significant enhancements including vision support, streaming responses, performance optimizations, and improved AI context.

An AI-powered response generator plugin for osTicket that helps agents generate intelligent, context-aware replies using OpenAI, Anthropic Claude, or any OpenAI-compatible API.

✨ Enhanced Features (This Fork)

🎯 Core Features

• Multi-AI Provider Support: OpenAI (GPT-4, GPT-4o), Anthropic (Claude 3/4), and OpenAI-compatible APIs
• Multi-Instance Support: Configure multiple AI providers simultaneously for different teams/workflows
• Smart Context Building: Includes recent ticket conversation with message types (Customer, Agent, Internal Note)
• Response Templates: Customizable output format with placeholders ({user_name}, {ticket_number}, etc.)

🚀 New in This Fork

🔄 OpenAI Responses API (v0.4.0)

• Migrated to OpenAI Responses API: Modern API with better streaming support
• File attachments support: Send PDF, TXT, CSV and other documents to the AI
• Improved streaming: Native SSE event parsing for real-time responses
• Configurable defaults: Optimized default values (timeout: 10s, context: 3 messages)
• Placeholder hints: Config fields show default values as placeholder text

👁️ Vision Support (v0.3.0)

• AI Image Analysis: Send image attachments to vision-capable models (GPT-4o, Claude 3+)
• Analyze screenshots, error messages, product photos, documents
• Configurable limits: max images, file size, inline/attachment filtering
• Automatic format conversion for OpenAI and Anthropic
• Cost protection: Disabled by default, conservative limits

⚡ Streaming Responses (v0.2.0)

• Real-time typewriter effect for AI responses
• Works with both OpenAI and Anthropic APIs
• Server-Sent Events (SSE) implementation
• Configurable per instance (enable/disable)

🔧 Performance Optimizations (v0.2.3)

• Efficient database queries using osTicket's QuerySet methods
• Only loads required thread entries (not all messages)
• Better message type labeling for improved AI context
• Fixed message ordering (newest first → most relevant context)

🐛 Bug Fixes

• v0.3.1: Fixed getPoster() returning string instead of object
• v0.2.1: Fixed pjax navigation ticket context bug
• v0.2.2: Fixed thread message order (was sending oldest instead of newest)

📋 Requirements

• osTicket: Version 1.18 or later
• PHP: Version 8.2 - 8.4 (8.4 recommended)
• AI Provider:
  • OpenAI API account (for GPT models)
  • Anthropic API account (for Claude models)
  • Or any OpenAI-compatible API endpoint

🚀 Installation

1. Download/Clone this repository
2. Copy the plugin folder to include/plugins/ in your osTicket installation:

   cd /path/to/osticket/include/plugins/
   git clone https://github.com/istoutjesdijk/ai-response-generator.git

3. Navigate to osTicket Admin Panel → Manage → Plugins
4. Click Add New Plugin and select AI Response Generator
5. Configure the instance (see Configuration section below)

⚙️ Configuration

Basic Setup

1. API URL: Your AI provider endpoint

   • OpenAI: https://api.openai.com/v1/responses
   • Anthropic: https://api.anthropic.com/v1/messages
   • Custom: Your OpenAI-compatible Responses API endpoint

2. API Key: Your authentication key

3. Model Name: Model identifier

   • OpenAI: gpt-4o, gpt-4-turbo, gpt-3.5-turbo
   • Anthropic: claude-3-opus-20240229, claude-3-sonnet-20240229, claude-3-haiku-20240307

Advanced Options

• Max Tokens: Response length limit (default: 512)
• Temperature: Creativity level 0.0-2.0 (default: 1.0)
• Timeout: API request timeout in seconds (default: 10)
• Max Thread Entries: Number of recent messages to include (default: 3)
• System Prompt: Custom instructions for AI behavior
• Response Template: Format output with placeholders
• Show Instructions Popup: Allow agents to add special instructions per response
• Enable Streaming: Real-time typewriter effect (default: disabled)
• Include Internal Notes: Include internal notes in AI context (default: enabled)

Template Variables

Both System Prompt and Response Template support osTicket variables:

Variable	Description
%{ticket.number}	Ticket number
%{ticket.subject}	Ticket subject
%{ticket.user.name}	Customer name
%{ticket.user.email}	Customer email
%{ticket.dept}	Department name
%{ticket.status}	Ticket status
%{ticket.priority}	Priority level
%{ticket.create_date}	Creation date
%{staff.name}	Agent name
%{date}	Current date (Y-m-d)
%{time}	Current time (H:i)
%{datetime}	Current date & time
%{day}	Day of week
%{ai_text}	AI response (template only)

Vision Support Configuration

⚠️ Warning: Vision increases API costs significantly!

• Enable Vision Support: Master switch (default: disabled)
• Max Images: Limit per request (default: 5, max: 10 for OpenAI, 100 for Anthropic)
• Max Image Size: Size limit in MB (default: 5 MB)
• Include Inline Images: Include embedded images like signatures (default: disabled)

Supported formats: JPEG, PNG, GIF, WebP

📖 Usage

For Agents

1. Open a ticket in the agent panel
2. Click the "AI Response" button (in toolbar or "More" dropdown)
3. (Optional) Add special instructions in the popup (e.g., "Offer refund")
4. Wait for the AI to generate a response
5. Review and edit the generated response as needed
6. Send the reply to the customer

Keyboard shortcuts in instructions popup:

• Ctrl+Enter (or Cmd+Enter): Submit and generate
• Escape: Close popup

With Vision Support

When vision is enabled and a customer includes screenshots or images:

• The AI automatically analyzes the images
• You'll get context-aware responses based on visual content
• Useful for error screenshots, product photos, diagrams, etc.

Multiple Instances

Configure multiple plugin instances for:

• Different AI providers (GPT-4 for complex tickets, Claude for cost savings)
• Different departments (Sales uses Claude, Support uses GPT-4o)
• A/B testing different models or prompts

🔒 Security

• Only staff with reply permission can use AI response generation
• API keys stored securely using osTicket's PasswordField (encrypted)
• All requests go through backend (no client-side API calls)
• Vision support processes images server-side only
• CSRF token validation on all AJAX requests
• Extra instructions input limited to 500 characters

🐛 Troubleshooting

Button doesn't appear

• Check plugin instance is enabled in Admin Panel
• Verify user has staff reply permissions
• Check browser console for JavaScript errors

API errors

• Verify API URL format (must include full endpoint path)
• Check API key is valid and has sufficient credits
• Increase timeout for slow API responses
• Ensure model name matches provider format

Vision not working

• Confirm vision support is enabled in config
• Use vision-capable model (GPT-4o, Claude 3+)
• Check image format (JPEG, PNG, GIF, WebP only)
• Verify image size is under configured limit

🆚 Differences from Original

Feature	Original	This Fork
API	Chat Completions	✅ Responses API
Anthropic Claude	❌	✅ Full support
Streaming	❌	✅ Real-time SSE
Vision/Images	❌	✅ GPT-4o & Claude 3+
File Attachments	❌	✅ PDF, TXT, CSV
Ticket Subject	❌	✅ Included in context
Thread Context	❌	✅ Recent messages
Internal Notes	❌	✅ Optional in context
Template Variables	Basic	✅ Full osTicket vars
Instructions Popup	❌	✅ Per-response guidance

🙏 Credits

• Original Plugin: Mateusz Hajder - mhajder/ai-response-generator
• Enhanced Fork: Additional features and improvements by the osTicket community

📄 License

MIT License - See LICENSE file for details

This fork maintains the original MIT license while adding significant enhancements. All original copyright notices are preserved.

🤝 Contributing

Contributions are welcome! Please:

1. Fork this repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

📧 Support

• Issues: GitHub Issues
• Original Plugin: mhajder/ai-response-generator
• osTicket: osTicket Documentation

---

⚠️ Cost Warning: Using AI APIs, especially with vision support, incurs costs. Monitor your usage and set appropriate limits. This plugin is provided as-is with no warranty.