AI Assistant Customization

Memento Database offers a powerful feature that allows you to customize the AI Assistant to better suit your specific needs. This customization involves modifying the system prompts, which are the instructions sent to the AI models. By tailoring these prompts, you can optimize the AI’s responses for your particular use cases.

What are System Prompts?

System prompts are pre-defined instructions that guide the AI in generating responses. They set the context and parameters for the AI’s output, ensuring that the responses are relevant and formatted correctly for Memento Database.

How to Customize AI Assistant

You can customize the AI Assistant in two ways:

  1. Global Customization: Apply changes to all libraries
    • Go to the main menu
    • Select “Settings”
    • Choose “Customize AI Assistant”
  2. Library-Specific Customization: Customize for a particular library
    • Edit the desired library
    • Go to the “Main” tab
    • Select “Customize AI Assistant”

Customization Options

You can set custom prompts for three main functions:

  1. Creating entries
  2. Editing entries
  3. Analyzing entries

Additionally, you can specify the size of photos that will be sent to the AI for analysis.

Understanding the Customization Modes

1. Creating Entries

In this mode, Memento expects to receive an XML file from the AI model in the format:

<items>
<item>...</item>
...
</items>

The default prompt includes the following placeholders:

  • #{fields.names.xml}: Replaced with field names
  • #{library.name}: Replaced with the library name

Memento will parse this XML and create new entries based on the data provided. You can customize the prompt to guide the AI in generating appropriate content for your specific library type.

Example of a customized prompt for a book library:
Create entries for #{library.name}. Include the following details for each book: #{fields.names.xml}. Ensure the author’s name is in “Lastname, Firstname” format, and provide a brief synopsis for each book without revealing major plot twists. Generate 3 entries in the required XML format.[/sourcecode]

2. Editing Entries

In this mode, Memento expects to receive an XML file from the AI model in the format:


<item>
<field_name1>value1</field_name1>
<field_name2>value2</field_name2>
...
</item>

The default prompt includes the placeholder #{entry.xml}, which is replaced with the entry’s current field values in XML format. You can also use individual field placeholders like #{name of field} to insert the value of a specific field into the prompt.

You can add additional instructions to guide the AI in modifying or enhancing the entry, including specifying the desired tone of the response. Here are some examples:

  • Serious Tone: “Provide a formal and factual update to the following entry, maintaining a professional tone throughout.”
  • Humorous Tone: “Update the entry with a light-hearted and witty approach, adding amusing anecdotes or puns where appropriate.”
  • Data Enrichment: “Based on the existing information, research and add any missing details, ensuring all fields are comprehensive.”
  • Language Improvement: “Refine the language in all text fields, correcting any grammatical errors and improving clarity.”
  • Formatting: “Format long text fields with appropriate paragraph breaks and bullet points for readability.”
  • Field-Specific Instructions: “For the ‘description’ field, expand the content to at least 100 words while maintaining accuracy.”

Example of a customized prompt for a movie database:

Edit the following movie entry: #{entry.xml}
1. Expand the plot summary to at least 150 words without revealing major spoilers.
2. Add or update the ‘genres’ field with all relevant movie genres.
3. In the ‘cast’ field, ensure the top 5 billed actors are listed.
4. If the ‘awards’ field is empty, research and add any significant awards or nominations.
5. Format the ‘trivia’ field as a bulleted list of interesting facts about the movie’s production or reception.
Provide the updated entry in the required XML format.

Example of a customized prompt for a recipe database with a casual, friendly tone:

Edit the following recipe: #{entry.xml}
1. Spice up the recipe name to make it more appealing and fun.
2. In the ‘ingredients’ field, add a dash of humor to each item (e.g., “1 cup of flour, sifted with love”).
3. Rewrite the ‘instructions’ in a conversational, friendly tone as if you’re chatting with a buddy in the kitchen.
4. Add a ‘chef’s tip’ field with a quirky or unexpected suggestion.
Provide the updated entry in the required XML format, keeping it light and entertaining throughout!

3. Analyzing Entries

In this mode, the placeholder #{entry.text} is replaced with the entry’s field values as plain text. Memento will display the AI’s response as plain text, without any specific formatting requirements.

This flexibility allows you to customize the analysis prompt for a wide range of purposes. You can add instructions to guide the AI in providing specific types of analysis, insights, or recommendations based on the entry data.

Examples of customized analysis prompts:

Analyze the following travel entry: #{entry.text}
1. Identify the main highlights of the trip and suggest related destinations for future travel.
2. Provide insights on the cultural experiences mentioned and how they compare to global trends.
3. Analyze the expenses listed and offer budget optimization tips for similar trips.
4. Based on the activities described, recommend a personalized packing list for a future visit to this destination.
5. Suggest a catchy title for this travel entry that captures its essence.

For a Health and Fitness Library:

Analyze the following workout session: #{entry.text}
1. Calculate the total calories burned based on the exercise type and duration.
2. Suggest complementary exercises for a balanced fitness routine.
3. Provide recovery tips based on the muscle groups targeted in this workout.
4. Recommend adjustments to improve effectiveness or prevent potential injuries.

By customizing these prompts, you can turn the AI Assistant into a powerful tool for data analysis, decision support, and personalized recommendations across various domains and use cases in your Memento Database libraries.

Tips for Effective Customization

  • Be specific in your instructions to get more targeted responses.
  • Use clear language to guide the AI in generating the desired output format.
  • Experiment with different phrasings to find what works best for your needs.
  • Remember to include relevant placeholders to ensure the AI has access to necessary information.
  • If Memento fails to understand the AI’s response (e.g., due to improperly formatted XML), try the following:
    • Rephrase your request to be more specific about the required XML structure.
    • Add explicit instructions in your prompt about proper XML formatting.
    • Try a different AI model, as some may be better at following strict formatting requirements.
    • Break down complex requests into simpler, more manageable parts.

By customizing your AI Assistant, you can significantly enhance its effectiveness in managing your Memento Database libraries. Take some time to experiment with different prompts and find the configuration that works best for your specific use cases.

Was this article helpful?

Related Articles