Set Up Your AI Agent
Last updated
Last updated
Currently Epsilla supports two types of agents: Chat Agents and Smart Search Agents. Chat Agents provide an interactive experience similar to ChatGPT, generating context-aware responses for use cases such as customer support or domain-specific conversations, often tailored to specific knowledge bases. Smart Search Agents focus on efficient information retrieval, inspired by platforms like Perplexity. They excel in providing precise answers to user queries, making them ideal for use cases such as research, content discovery, and data-driven decision-making.
In your Epsilla Cloud dashboard, click on the Create Application button located on the navigation bar.
Or click on the Applications tab. This will lead you to the page where you can create and manage all your Applications.
Select Chat Agent as the application type from the available options.
Name Your Chat Agent: Enter a name for your chat agent that reflects its purpose.
Provide a Description: Add a description for your chat agent to indicate what kind of tasks or questions it is intended to handle.
Upload a Logo (Optional): For a personalized touch, upload a logo to represent your chat agent visually in the interface.
The chat agent will be created in 15 - 30 seconds. Now you can preview the Chat Agent and modify its configurations side by side:
Technically this is the system prompt to instruct the LLM to behave with desired personality. For instance, if you're creating a tax assistant, you could use a system prompt like:
"You are a professional CPA specializing in individual taxation. Utilize your expert knowledge of tax laws and regulations to ensure clients benefit from all applicable deductions and credits. Provide clear explanations to clients about their tax situations and the implications of their financial decisions."
This role setup is crucial as it helps ensure that the chat agent’s responses are aligned with your intended use case.
Link the Knowledge Base: Select the knowledge base you created and attach it to your chat agent. This ensures that the chat agent can ground its responses in the specific data you provided, making answers more relevant and accurate. You can add multiple knowledge bases to an AI agent.
Change Icon: Select a new icon for your agent by uploading a custom image. Ensure the icon reflects your agent's purpose and aligns with your brand identity.
Change Primary Color: Customize the primary color to match your branding.
Introduction Message: Set an introductory message that the chat agent will use to greet new users. This helps set the context of the conversation.
Chatbox Placeholder: Set the placeholder text in the chatbox.
Sample Questions: Add sample questions for users to quickly get started. For example, for a tax assistant, questions could be like "What are the standard deductions?" or "How do I report freelance income?"
Enable Follow-up Questions: Allow users to ask follow-up questions during a conversation, providing a more interactive and engaging experience. If enabled, you can futther customize the LLM model and prompt for generating the follow up questions.
How to Label a Chat: Config how to label a chat. By default Epsilla will use LLM to summarize the a chat.
Hide Epsilla Logo: For a paid tier, you can choose to hide Epsilla logo from the Chat Agent for white labeling.
Collect User Feedback: Whether allow users to provide feedback to the Chat Agent response.
Allow Message Sharing: Whether allow other users to share the chat with others.
You can test the chat agent in the right hand side Preview. Test the chat agent in the interface to ensure it responds appropriately. Verify that it can answer questions using the knowledge base and that the defined role aligns with its response:
Click the resize button to make the preview full screen:
