This guide outlines the step-by-step process to manually construct prompts.

​
1. Accessing the Prompt Management Dashboard

  • Navigate to Future AGI Dashboard
  • Select Prompt from the left-hand sidebar to access the Prompt Management interface.
  • Click Add New Prompt or Write a prompt from scratch. This opens a blank workspace where you can create and evaluate your prompts.

​
2. Defining the User Input

  • In the Workbench section, click the input box and enter an instruction.
    • Example: “Write an email.”
  • For better AI responses, provide a detailed task description.
    • Example: “Write a professional email to a hiring manager highlighting key skills and experience.”
  • You can use variables for dynamic text insertion by referencing them in double curly braces, e.g., {{variable_name}}.

​
3. Configuring Model Settings

  • Click Model Settings on the top right corner.
  • Adjust the following parameters based on your requirements:
    • Temperature: Controls randomness in responses (0.5 for balanced outputs, higher values for creativity, lower for consistency).
    • Top P: Regulates response probability (keep at 1 for broad selection of tokens).
    • Max Tokens: Sets the response length (e.g., 8190 for longer outputs).
    • Presence & Frequency Penalty: Adjusts word repetition control.
  • Click Save to apply the changes.

​
4. Utilizing Variables for Dynamic Inputs

  • Click Variables on the top bar.
  • You can manually add data of the variable by double clicking on the cell.
  • You can also use data from existing datasets, select Import Data and choose the relevant dataset.
  • Map dataset columns to corresponding variables and confirm the assignment.
  • Click Apply to save changes.

​
5. Testing and Evaluating the Prompt

  • Click Save and Run to execute the prompt.
  • The response will appear in the Results section.
  • Review the output and ensure:
    • It follows the given instructions.
    • It incorporates dynamic variables correctly.
    • It maintains clarity and coherence.
  • If necessary, edit the prompt and rerun the test until the desired response is achieved.
  • Utilize the Evaluate feature to analyze the effectiveness of the prompt
Click here to learn more about Evaluation
Following this structured process ensures the creation of effective, reusable prompts tailored for AI-driven applications. By defining clear instructions, utilizing variables, testing iteratively, and refining outputs, users can generate high-quality AI responses aligned with specific operational needs.