Print

AI Chat – Tuning

Updated OnApril 23, 2025
0 out Of 5 Stars
5 Stars 0%
4 Stars 0%
3 Stars 0%
2 Stars 0%
1 Stars 0%
Views3

This page provides a clear guide to tuning the AI Chat feature in Echo Knowledge Base. It covers essential steps and best practices to customize the AI’s responses, ensuring accurate and relevant answers for your users.

Tuning

Embeddings Tests

On the Embeddings tab, you can switch from Edit to AI Search mode.

In AI Search mode, you can query the embeddings to check if the AI can answer your questions. If no answers are found or the answers are incorrect, you may need to add more information to your embeddings or optimize your model (see Optimizations below).

Tuning

Embeddings - Minimum Score

The "minimum score" in Pinecone refers to the lowest score you’re willing to accept for a result to be considered relevant. For example, if you set a minimum score of 70, Pinecone will only show results with a score of 70 or higher, filtering out less relevant matches.

This helps you focus on the most useful results, especially when dealing with lots of data. However, setting the minimum score too high might exclude some good results, so it’s a balance you adjust based on what you need.

Default is 35 but experiment with lower or higher values.

Tuning

Chatbot - Instructions

The context parameter is critical, as it defines how the AI should act, including setting boundaries on responses or incorporating specific details like your company profile and links. In many scenarios, well-crafted context alone can create an exceptional chatbot experience.

Examples of AI Chat Instructions:

  • "Act as a friendly support agent for [Your Company Name], providing accurate information about our services and avoiding unrelated topics."
  • "You are a knowledgeable assistant for [Product Name]. Share details from our website [Your Website URL] and decline non-product-related questions."
  • "Be a professional representative of [Your Brand], using our company bio [Insert Bio] and linking to [Your Links] when relevant."
  • "Serve as a customer service bot for [Your Company], focusing on our offerings and politely refusing to answer off-topic queries."
  • "You are an expert assistant for [Service/Product Name], incorporating our mission [Insert Mission] and guiding users to [Your Website URL]."
Tuning

Chatbot - Local Memory

Local Memory: Store the conversation in the user's browser. For a given Chatbot ID, the dialogue persists and can be resumed across different pages, independent of whether conversations are stored on the server.

Was this article helpful?
0 out Of 5 Stars
5 Stars 0%
4 Stars 0%
3 Stars 0%
2 Stars 0%
1 Stars 0%
5
How can we improve this article?
Please submit the reason for your vote so that we can improve the article.
Need help?
Tags:
Table of Contents