Answer-style apps

Allganize Alli Works 유저가이드 (EN)

Answer-style apps

Answer-type apps are builders that allow anyone to easily create an app based on a desired idea or prompt using an LLM. The generated results can be copied or downloaded as documents for use.

To create an answer-type app, go to App Management > Create New App > Create Answer-Type App.

App Name: Set the name of the app to be shown to users. Description: Enter a brief overview and description of the app to clearly communicate to users what they can do with the app. Input Settings: There are six types of input settings; these determine how the user will provide input.

As an example, let's assume we're creating a translation app and add three fields.

1) Document Set the input type to 'Document (default)' so that users can directly upload the file they want to translate. Configure it so that the uploaded document is stored in a variable named file1.

*If you want the app to use files already uploaded to Dashboard > Knowledge Base, turn on the 'Attach document from knowledge base' option and select the desired document using 'Set document(s) the user can choose'. You can configure the answer-type app to target a specific uploaded document.

2) Paragraph Text

You can also specify that users input text directly instead of uploading a file. Set the input name to 'Content' and configure it so the input is stored in a variable named content. At this time, using the 'Placeholder input' feature lets you add example sentences or explanations of what to enter.

3) Single Select

Create a language selection field. Provide three language options: Korean, English, and Japanese, and configure it so the selected language is stored in a variable named language.

Once configuring the items to be received from the user is complete, you can select the model that will run the app.

Choose the most suitable model for the app among the activated models. If you want to use a deactivated model, please contact the person in charge.

Prompt: Write a simple translation app prompt that converts the input text into the appropriate language according to the translation style selected by the user.

System Prompthelps set the behavior of the LLM. For example, you can provide a persona to the LLM model or give specific instructions on how the LLM should behave throughout the conversation using this prompt.

User Promptis where you can provide the request or question the LLM should respond to. It is recommended to write queries and questions in the user prompt.

Markdown Output ButtonWhen pressed, a prompt to output in markdown is automatically added when generating the output. Markdown output is a syntax for displaying text-based information neatly and readably.

Choose how to handle long inputs: This selects the method to use when processing long input data. Below are descriptions of each option: 1) Process without splitting input until token limit

• Process the input at once without splitting it.

• Suitable when the input does not exceed the model's maximum token count.

2) Map (split input processing and then concatenate results as-is)

• Split the input into multiple pieces, process each piece, and then concatenate the results as-is.

• Use when the input is long but you need to process all content without separate summarization or reorganization.

3) Map-Reduce (split input processing and then merge results)

• Split the input into small pieces, process each piece (map), then gather and summarize or merge the results (reduce).

• Efficient for handling large amounts of text or documents; it processes by repeatedly applying summarization prompts so as not to exceed the model's maximum token count.

After testing the desired behavior, create the app by clicking the 'Publish' button at the top right.