Filaforge HuggingFace Chat

A powerful Filament plugin that integrates HuggingFace AI chat capabilities directly into your admin panel.

Features

• HuggingFace AI Integration: Chat with thousands of AI models from HuggingFace
• Conversation Management: Save, organize, and continue chat conversations
• Model Selection: Choose from a wide variety of AI models
• Customizable Settings: Configure API tokens, models, and chat parameters
• Real-time Chat: Live chat experience with streaming responses
• Conversation History: Keep track of all your AI conversations
• Export Conversations: Save and share chat transcripts
• Role-based Access: Configurable user permissions and access control
• Multi-model Support: Switch between different HuggingFace models
• Context Awareness: Maintain conversation context across sessions

Installation

1. Install via Composer

composer require filaforge/huggingface-chat

2. Publish & Migrate

# Publish provider groups (config, views, migrations)
php artisan vendor:publish --provider="Filaforge\\HuggingfaceChat\\Providers\\HfChatServiceProvider"

# Run migrations
php artisan migrate

3. Register Plugin

Add the plugin to your Filament panel provider:

use Filament\Panel;

public function panel(Panel $panel): Panel
{
    return $panel
        // ... other configuration
        ->plugin(\Filaforge\HuggingfaceChat\Providers\HfChatPanelPlugin::make());
}

Setup

Configuration

The plugin will automatically:

• Publish configuration files to config/hf-chat.php
• Publish view files to resources/views/vendor/hf-chat/
• Publish migration files to database/migrations/
• Register necessary routes and middleware

HuggingFace API Configuration

Configure your HuggingFace API in the published config file:

// config/hf-chat.php
return [
    'api_token' => env('HF_API_TOKEN'),
    'base_url' => env('HF_BASE_URL', 'https://api-inference.huggingface.co'),
    'default_model' => env('HF_MODEL_ID', 'meta-llama/Meta-Llama-3-8B-Instruct'),
    'max_length' => env('HF_MAX_LENGTH', 512),
    'temperature' => env('HF_TEMPERATURE', 0.7),
    'stream' => env('HF_STREAM', false),
    'timeout' => env('HF_TIMEOUT', 60),
    'use_openai_format' => env('HF_USE_OPENAI', true),
];

Environment Variables

Add these to your .env file:

HF_API_TOKEN=your_huggingface_api_token_here
HF_BASE_URL=https://api-inference.huggingface.co
HF_MODEL_ID=meta-llama/Meta-Llama-3-8B-Instruct
HF_MAX_LENGTH=512
HF_TEMPERATURE=0.7
HF_STREAM=false
HF_TIMEOUT=60
HF_USE_OPENAI=true

Getting Your HuggingFace API Token

1. Visit HuggingFace
2. Create an account or sign in
3. Go to your profile settings
4. Navigate to "Access Tokens"
5. Generate a new token
6. Copy the token to your .env file

Usage

Accessing HuggingFace Chat

1. Navigate to your Filament admin panel
2. Look for the "HF Chat" menu item
3. Start chatting with AI models

Starting a Conversation

1. Select Model: Choose from available HuggingFace models
2. Type Your Message: Enter your question or prompt
3. Send Message: Submit your message to the AI
4. View Response: See the AI's response
5. Continue Chat: Keep the conversation going

Managing Conversations

1. New Chat: Start a fresh conversation
2. Save Chat: Automatically save important conversations
3. Load Chat: Resume previous conversations
4. Export Chat: Download conversation transcripts
5. Delete Chat: Remove unwanted conversations

Advanced Features

• Model Selection: Switch between different HuggingFace models
• Parameter Tuning: Adjust temperature, max length, and other settings
• Context Management: Maintain conversation context across sessions
• Streaming Responses: Real-time AI responses (when supported)

Troubleshooting

Common Issues

• API token errors: Verify your HuggingFace API token is correct
• Rate limiting: Check your HuggingFace API rate limits and usage
• Model not available: Ensure the selected model is available and loaded
• Connection timeouts: Check network connectivity and timeout settings

Debug Steps

1. Check the plugin configuration:

php artisan config:show hf-chat

2. Verify routes are registered:

php artisan route:list | grep hf-chat

3. Test API connectivity:

php artisan tinker
# Test your API token manually

4. Check environment variables:

php artisan tinker
echo env('HF_API_TOKEN');

5. Clear caches:

php artisan optimize:clear

6. Check logs for errors:

tail -f storage/logs/laravel.log

API Error Codes

• 401 Unauthorized: Invalid or expired API token
• 429 Too Many Requests: Rate limit exceeded
• 503 Service Unavailable: Model is currently loading or unavailable
• Timeout: Request took too long to complete

Security Considerations

Access Control

• Role-based permissions: Restrict access to authorized users only
• API token security: Never expose API tokens in client-side code
• User isolation: Ensure users can only access their own conversations
• Audit logging: Track all chat activities and API usage

Best Practices

• Use environment variables for API tokens
• Implement proper user authentication
• Monitor API usage and costs
• Regularly rotate API tokens
• Set appropriate rate limits

Uninstall

1. Remove Plugin Registration

Remove the plugin from your panel provider:

// remove ->plugin(\Filaforge\HuggingfaceChat\Providers\HfChatPanelPlugin::make())

2. Roll Back Migrations (Optional)

php artisan migrate:rollback
# or roll back specific published files if needed

3. Remove Published Assets (Optional)

rm -f config/hf-chat.php
rm -rf resources/views/vendor/hf-chat

4. Remove Package and Clear Caches

composer remove filaforge/huggingface-chat
php artisan optimize:clear

5. Clean Up Environment Variables

Remove these from your .env file:

HF_API_TOKEN=your_huggingface_api_token_here
HF_BASE_URL=https://api-inference.huggingface.co
HF_MODEL_ID=meta-llama/Meta-Llama-3-8B-Instruct
HF_MAX_LENGTH=512
HF_TEMPERATURE=0.7
HF_STREAM=false
HF_TIMEOUT=60
HF_USE_OPENAI=true

Support

• Documentation: GitHub Repository
• Issues: GitHub Issues
• Discussions: GitHub Discussions

Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

This plugin is open-sourced software licensed under the MIT license.

Made with ❤️ by the Filaforge Team