Skip to main content

Configuring Preamble and Postamble Settings in Unstract

Understanding Preamble and Postamble

  • Preamble: Text that is added before your prompt questions to guide the LLM's approach to extraction
  • Postamble: Text that is added after your prompt questions to specify how responses should be formatted

Steps for Configuring Preamble Settings

  1. Access the Settings menu by clicking the gear icon (⚙️) in the top-right corner of the Unstract interface.

  2. In the Settings sidebar on the left, click on "Preamble".

  3. You'll see the "Preamble Settings" panel with a text area containing the default preamble:

    • The default text instructs the model to pay close attention to context, language, structure, and cross-references
    • It emphasizes using only information from the provided context and not external knowledge
  4. Modify the preamble text to suit your specific requirements:

    • You can customize the instructions to better align with your extraction needs
    • The preamble sets the context and guidelines for how the LLM should approach all prompts
  5. You can reset to the default prompt by clicking "Reset with default prompt" if needed.

  6. Click the "Save" button to apply your changes.

img Prompt Studio project

Steps for Configuring Postamble Settings

  1. Access the Settings menu by clicking the gear icon (⚙️) in the top-right corner.

  2. In the Settings sidebar on the left, click on "Postamble".

  3. You'll see the "Postamble Settings" panel with a text area containing the default postamble:

    • The default postamble instructs the model to "not include any explanation in the reply. Only include the extracted information in the reply."
  4. Modify the postamble text according to your requirements:

    • You can specify how you want the extracted information to be formatted
    • Add instructions for response style, brevity, or specific output format
  5. You can reset to the default prompt by clicking "Reset with default prompt" if needed.

  6. Click the "Save" button to apply your changes.

img Prompt Studio project

Best Practices for Preamble and Postamble

  • Preamble: Focus on setting extraction guidelines, context handling, and general approach
  • Postamble: Focus on response formatting, output style, and presentation of results