While large language models (LLMs) demonstrate exceptional performance on complex tasks, they still face limitations—such as inability to access up-to-date information, tendency to generate hallucinations, and difficulty performing precise calculations. To address these challenges, you can integrate plugins into your LLM applications to extend their capabilities and execute more sophisticated tasks.
Introduction to Plugins
A plugin is a collection of tools. Each plugin may contain multiple tools (APIs), with each tool implementing a specific function. Bailian supports official plugins, third-party plugins, and custom plugins.
Official Plugins
The Component Marketplace includes preconfigured official Bailian plugins. You can invoke them directly without configuring input or output parameters. For usage instructions, see Official and Third-Party Plugins.
Third-Party Plugins
In addition to official plugins, third-party plugins span domains including commercial services, image/video processing, and education—and have undergone performance testing. Once enabled, they can be invoked immediately without additional configuration. For usage instructions, see Official and Third-Party Plugins.
Custom Plugins
You can create custom plugins and integrate them into your LLM applications to enable flexible, tailored functionality extensions.
Plugin Effectiveness Example
| Sample Input | Output Without Plugins | Output With Plugins |
|---|---|---|
| What is 12313 × 13232? | The LLM application cannot accurately compute complex mathematical problems and may produce incorrect results. The correct result is 162,925,616. | The LLM application gains robust computational capability and generates accurate calculation results. |
Supported Models
| Model | Model Identifier |
|---|---|
| Qwen-Turbo | qwen-turbo |
| Qwen-Plus | qwen-plus |
| Qwen-Max | qwen-max |
| Qwen-VL-Max | qwen-vl-max |
| Qwen-VL-Plus | qwen-vl-plus |
Plugin compatibility may vary across models. For the most up-to-date compatibility status, refer to actual execution results in the console.
Plugin Invocation Mechanism
Invoking a plugin fundamentally means invoking one or more tools within that plugin. Bailian supports plugin invocation via Agent Applications, Workflow Applications, and the Assistant API.
When invoking plugins through an Agent Application or the Assistant API, the LLM determines whether to call a tool based on:
- the user’s input,
- the tool name, and
- the tool description.
- If a tool call is required, the LLM selects the appropriate tool. After internal tool execution, the plugin returns the result, which is merged with the original user input and fed back into the LLM to generate the final response.
- If no tool call is needed, the LLM generates and outputs the response directly.