Home AI Agent Agent Setting

Agent Setting

Last updated on Oct 21, 2024

Once your agent is successfully created, you'll be automatically directed to the settings page where you can configure its functionalities.

Configuring Prompts

Prompts play a crucial role in directing the AI to deliver precise and professional responses. Incorporate prompt variables, such as {{input}}, to dynamically replace them with user-provided values, ensuring the responses are tailored and accurate.

Integrating a Knowledge Base

To enable an agent to produce content grounded in personalized context, utilize the 'Knowledge' feature. By clicking the [Edit] button in the [Knowledge Base] section, you can incorporate an existing knowledge base to bolster the agent's contextual understanding.

Enhancing with Tools

For ChatAgent-type agents, explore the [Tools] section where you can add necessary tools to expand their capabilities. These tools can significantly augment LLM functions, enabling tasks like internet searches, performing complex calculations, or creating images, thus enhancing the agent's real-world interaction capabilities.

Chat Enhance Feature

Visit the [Chat Enhance] section and select [Add Feature] to access the agent toolbox.

  • Conversation Opener: Set the AI to initiate dialogue by providing a welcoming sentence or inquiry. Craft these openers to encourage user interaction, convey the agent's background, and ease users into starting a conversation.

  • Follow-up Suggestions: Enable this option to let the AI propose three subsequent questions based on the dialogue, steering future interactions smoothly.

  • Citation and Attribution: Activating this feature ensures the AI references information from the knowledge base in its responses. Detailed citation data, including original text, segment number, and matching score, can be viewed beneath each answer.

Debugging and Previewing

Use the [Preview] function to simulate user interactions and examine the agent's response quality. If you find the output lacking, you may need to revisit and refine prompts, tweak model parameters, or adjust other settings to achieve optimal performance.