README

connector-qwen

Qwen (Alibaba Cloud) AI connector for Ibexa DXP.

This bundle integrates the Qwen AI models via Alibaba Cloud's DashScope API (OpenAI-compatible) with the Ibexa AI framework (ibexa/connector-ai). It uses the openai-php/client SDK with a custom base URL, so there is no need for a hand-rolled HTTP client.

Requirements

• PHP >= 8.3
• Ibexa DXP ~5.0
• ibexa/connector-ai ~5.0
• Alibaba Cloud DashScope API key (get one at Model Studio Console)

Installation

Require the package:

composer require arfaram/connector-qwen

Register the bundle in config/bundles.php:

return [
    // ...
    ConnectorQwenBundle\ConnectorQwenBundle::class => ['all' => true],
];

Environment Variable

Set your DashScope API key in .env:

QWEN_API_KEY=sk-xxxxxxxxxxxxxxxxxxxxxxxx

⚠️ This must be a DashScope API key (starts with sk-), NOT an Alibaba Cloud AccessKey ID/Secret pair.

To get your DashScope API key:

• International: Go to Model Studio Console, navigate to API Keys
• China region: Go to DashScope Console → API Keys

⚠️ You must also activate each model you want to use in the Model Studio console under Model Square / Model Plaza.

Region Configuration

The bundle defaults to the international DashScope endpoint. If you're using the China region, override per SiteAccess:

ibexa:
    system:
        default:
            connector_qwen:
                qwen:
                    api_key: '%env(QWEN_API_KEY)%'
                    # International (default):
                    base_uri: 'https://dashscope-intl.aliyuncs.com/compatible-mode/v1'
                    # China region:
                    # base_uri: 'https://dashscope.aliyuncs.com/compatible-mode/v1'

Configuration

Available Qwen Models

⚠️ Retired models (do not use): qwen-max, qwen-plus, qwen-turbo, qwen2.5-* series are deprecated.

Text Models

| Model ID | Description |---|---|---| | qwen3.6-plus | Qwen 3.6 Plus, most capable (latest) | qwen3.6-plus-2026-04-02 | Qwen 3.6 Plus pinned to 2026-04-02 | qwen3.6-flash | Qwen 3.6 Flash, fast & balanced (latest) | qwen3.6-flash-2026-04-16 | Qwen 3.6 Flash pinned to 2026-04-16 | qwen3.5-plus | Qwen 3.5 Plus (latest) | qwen3.5-flash | Qwen 3.5 Flash (latest) | qwen3-max | Qwen 3 Max (latest) | qwen3.6-35b-a3b | Qwen 3.6 Open Source (35B-A3B) | qwen3.5-35b-a3b | Qwen 3.5 Open Source (35B-A3B) | qwen-long | Qwen Long, long context

Vision Models (Image-to-Text)

| Model ID | Description |---|---|---| | qwen3-vl-plus | Qwen 3 VL Plus, dedicated vision | qwen3-vl-plus-2025-09-23 | Qwen 3 VL Plus pinned to 2025-09-23 | qwen3.6-plus | Qwen 3.6 Plus, multimodal capable (latest) | qwen3.6-flash | Qwen 3.6 Flash, multimodal capable (latest) | qwen3.5-plus | Qwen 3.5 Plus, multimodal capable (latest) | qwen3.5-flash | Qwen 3.5 Flash, multimodal capable (latest) | qwen3.6-35b-a3b | Qwen 3.6 Open Source vision | qwen3.5-35b-a3b | Qwen 3.5 Open Source vision

Basic Configuration

Create config/packages/ibexa_connector_qwen.yaml:

ibexa_connector_qwen:
    text_to_text:
        default_model: qwen3.6-flash
        default_temperature: 1.0
        default_max_tokens: 4000
        models:
            qwen3.6-plus: 'Qwen 3.6 Plus (latest)'
            qwen3.6-flash: 'Qwen 3.6 Flash (latest)'
            qwen3.5-plus: 'Qwen 3.5 Plus (latest)'
            qwen3-max: 'Qwen 3 Max (latest)'
            qwen3.6-35b-a3b: 'Qwen 3.6 Open Source'
    image_to_text:
        default_model: qwen3-vl-plus
        default_temperature: 1.0
        default_max_tokens: 4000
        models:
            qwen3-vl-plus: 'Qwen 3 VL Plus (latest)'
            qwen3-vl-plus-2025-09-23: 'Qwen 3 VL Plus (2025-09-23)'
            qwen3.6-plus: 'Qwen 3.6 Plus (latest)'
            qwen3.6-flash: 'Qwen 3.6 Flash (latest)'
            qwen3.6-35b-a3b: 'Qwen 3.6 Open Source'

SiteAccess-Aware API Key Configuration

ibexa:
    system:
        default:
            connector_qwen:
                qwen:
                    api_key: '%env(QWEN_API_KEY)%'
        my_siteaccess:
            connector_qwen:
                qwen:
                    api_key: '%env(QWEN_API_KEY_MY_SITEACCESS)%'

Minimal Configuration (uses defaults)

ibexa_connector_qwen:
    text_to_text:
        default_model: qwen3.6-plus

All other values will use their defaults:

• default_temperature: 1.0
• default_max_tokens: 4000
• models: all current Qwen models as defined in the Configuration class

Production-Optimized Configuration

Use dated model versions in production for deterministic, reproducible results:

ibexa_connector_qwen:
    text_to_text:
        default_model: qwen3.6-flash-2026-04-16
        default_temperature: 0.7
        default_max_tokens: 4096
        models:
            qwen3.6-plus-2026-04-02: 'Qwen 3.6 Plus (2026-04-02)'
            qwen3.6-flash-2026-04-16: 'Qwen 3.6 Flash (2026-04-16)'
    image_to_text:
        default_model: qwen3-vl-plus-2025-09-23
        default_temperature: 0.5
        default_max_tokens: 2048
        models:
            qwen3-vl-plus-2025-09-23: 'Qwen 3 VL Plus (2025-09-23)'

License

MIT - See LICENSE for details.

Author

Ramzi Arfaoui - ramzi_arfa@hotmail.de