Context Builder Guide
  • Context Builder Guide
  • Welcome to Context Builder
    • 💡Quickstart Guide
    • ✨New to Generative AI?
  • BUILDER GUIDE
    • 📄DataSet
      • Data Loaders
      • Embeddings
      • Vector Databases
    • ⌚TOOL
      • Conversation chain
      • Conversation retriever chain
      • Self Checking
  • APP ANALYSIS
    • Monitoring
  • User Case
    • Simple Conversation
    • Product customer service
    • Mock Interview
  • User Agreement
    • Terms of Service
    • Privacy Policy
Powered by GitBook
On this page
  • Step One
  • Step Two
  • Step Three
  1. User Case

Simple Conversation

PreviousUser CaseNextProduct customer service

Last updated 1 year ago

Come and create a basic AI chat application together. All it requires is one Tool-Conversation to get started.

Step One

Click the "Add Task" button on the blank workflow page and add a Tool-Conversation.

Step Two

Complete the configuration for the Tool-Conversation on the right side.

You can change the model type; currently, Gpt-3.5 and Gpt4 are available.

You can adjust the model's settings, tweaking parameters to obtain different prompt outcomes, such as Temperature and Top_p.

  1. Temperature: In simple terms, the lower the temperature parameter, the more deterministic the model's output. Increasing this value might make the language model return more random outputs, meaning it might produce more diverse or creative outcomes. We're also boosting the weight of other possible tokens. Practically speaking, for tasks like Quality Assurance (QA), you can set a lower temperature to make the model return more factual and concise results. For creative tasks like poetry generation, you might want to increase the temperature slightly.

  1. Top_p: Similarly, using top_p (a technique known as nucleus sampling alongside temperature) can control the authenticity of the model's output. For accurate and factual answers, lower the parameter. For more diverse answers, increase it a bit. It's generally recommended to adjust only one of these parameters, not both.

Step Three

Debug and publish your app. It's advised to always debug before publishing your app. If the results aren't satisfactory, you can promptly adjust the model parameters and System Prompt.

With these simple three steps, you can create a basic AI chat application.

You can swap out different prompts to have the AI address various issues.

For instance:

  1. Language translation

  2. Writing copy｜articles｜outlines｜titles｜novels｜summary texts

  3. Role-playing in professional scenarios: Project Manager｜Product Manager｜Customer Service｜Interviewer/Mental Health Counselor｜Data Analyst｜Assistant

Here's a great example

System Prompt: "You talk to me like Harry Potter, make sure you know all about Harry Potter."