Skip to content

Latest commit

 

History

History
91 lines (70 loc) · 4.91 KB

10_1_1_Product Platform Quick Start.md

File metadata and controls

91 lines (70 loc) · 4.91 KB
Raw

Product Platform Quick Start

In this section, we will show you how to:

  • Start the agentUniverse productized service
  • Configure the agentUniverse product modules
  • Quickly get started with productized features
  • Debug and experience agents online, optimizing agent performance

Environment and Application Engineering Preparation

Application Engineering Preparation

We have placed the product module samples in the agentUniverse’s sample_standard_app project. You can view them here. This part can be configured in the background through YAML, and of course, these functions can be automatically created and managed through the product page.

Installing Dependencies

Using pip

pip install magent-ui ruamel.yaml

Configuration File

If you have previously used aU’s sample project, please add the following information to the config.toml file configuration to include the product module path, as configured in sample_standard_app:

# Ignore the context content.
[CORE_PACKAGE]
# Scan and register product components for all paths under this list, with priority over the default.
product = ['sample_standard_app.app.core.product']
# Ignore the context content.

If you are using aU for the first time, you can directly use the latest sample project’s config.toml.

Private Configuration File

Of course, when using the agent, you need to pre-configure the various LLM model keys/Tool keys, otherwise, the overall process of the agent cannot be connected. This part is currently not provided in the product page configuration management. Please reuse aU’s original key configuration method. In future versions, we will directly provide key management capabilities for each model in the product.

Using the agentUniverse Product Platform

Starting the Product Service

Run the product_application file located in sample_standard_app/app/bootstrap to start with one click.

img.png

After a successful start, it will automatically redirect to the product homepage, which includes system presets and your custom Agent/Tool/Knowledge product modules. agentuniverse_product_homepage

Experience the Agent

As shown in the image above, click the chat button on the right of the peer multi-agent group to enter the conversation page.

The conversation management system includes the last 10 agent conversation history records, allowing you to directly engage in multi-turn dialogues and experience the capabilities of the peer multi-agent group (default is streaming dialogue, and the multi-agent group includes the intermediate thinking process). agentuniverse_product_agent_chat

Debugging the Agent

On the product homepage, click the edit button on the left of the agent to enter the online debugging page.

You can debug the agent’s Prompt/Tool/Knowledge/LLM online. Click the save button, and aU-product will automatically save the configuration to the corresponding yaml file. agentuniverse_product_agent_editor

Click the debug button in the upper right corner of the image above to view the Trace information, including token consumption, call chain, and latency of the agent’s specific invocation process. agentuniverse_product_agent_trace

Configuring agentUniverse Product Modules

Creating Product Modules

id: demo_rag_agent
nickname: rag智能体
type: AGENT
opening_speech: |
  欢迎使用rag chatbot！根据您的提问，我将结合实时信息及我个人掌握的知识，给您提供合理的解答。
  你可以这样问我：
  问题1: 巴黎奥运会中国获得了几块奖牌
  问题2: 巴菲特为什么减持苹果股票
avatar: ../../../resources/rag_agent_logo.png
metadata:
  class: AgentProduct
  module: agentuniverse_product.base.agent_product
  type: PRODUCT

As shown in the yaml file above:

  • product id corresponds to name in the agent yaml
  • type corresponds to AGENT
  • metadata is configured as the value shown above
  • Other parameters can be configured as needed (the same applies to tool/knowledge modules)

Configuring Global Config

In the global config.toml, configure the package scan path corresponding to the product module. With one click, you can start the product platform and display the corresponding agent/tool/knowledge information.

More

The agentUniverse-product platform is continuously iterating, and more convenient and useful features will be available for community users to experience in the future. We also welcome your valuable suggestions.

This feature is jointly launched by difizen and agentUniverse.

Let’s explore and progress together!