Skip to main content

Deeper tuning and advanced techniques for optimizing your AI (for demanding and technically savvy users)

This section is intended for users who want to go beyond the standard configuration and achieve even more precise behavior from their AI instance. These techniques require a deeper understanding of how BuddyPro works and a willingness to experiment. Before diving in, ensure you have a solid grasp of the fundamentals—especially the creation of a high-quality System Prompt and the preparation of relevant know-how.

28.1 Precise Manual Editing of Expert Subrole Definitions​

As mentioned in sections 11.6 and 16.3, you have the option to manually edit the content of the Google Docs that define each expert subrole, which BuddyPro automatically generates based on your know-how in the ROLES folder on your Google Drive.

When can this advanced technique be helpful?​

  • If, even after carefully refining the System Prompt and adding relevant know-how, you notice that the AI consistently underperforms in a specific expert subrole, gives suboptimal advice, or behaves in a way that contradicts your intended specialization. (Use the /lastRole command to identify the active role, and /investigateAnswer:{message} to deeply analyze the AI's response.)
  • If you want to explicitly embed specific directives, instructions, or links into a role definition that the AI should prioritize within that role's context. For example:
    • In the role "Copywriting Masterclass Advisor," you might add: "When a user expresses deeper interest in sales copy techniques, always recommend my e-book 'The Copywriter's Secret Weapons' with a link to [URL], and briefly mention its main benefits."
    • Or in the role "Facebook Ads Specialist for Beginners," you might specify: "When explaining how to set up the first campaign, always follow my 5-step 'FB Start' framework. Never forget to emphasize the importance of the Pixel!"

How to do it (with care and after thoughtful consideration)​

  1. Identify the specific Google Doc with the role definition you want to edit (e.g., Role_FacebookAdsBeginners.gdoc). Role names are also listed in the file ROLES/list_of_roles.gdoc.

  2. Create a backup of the file before making any changes!

  3. Carefully review the existing auto-generated content to understand the structure and logic of the role.

  4. Edit the text with caution. You can add new paragraphs, refine existing wording, or include explicit instructions. Keep your edits as clear, specific, and consistent as possible with the rest of the role definition and your global System Prompt.

  5. Save the changes in the Google Doc.

  6. To apply your manual changes in the AI's behavior, run the following command in Telegram in your BuddyPro instance's chat:

    Replace role_name_without_extension with the exact name of the role file (from the ROLES folder) without the .gdoc suffix. For example, if editing Role_FacebookAdsBeginners.gdoc, use:

    /updateRole:Role_FacebookAdsBeginners

    If the changes don't take effect, you can try the general cache-refresh command:

    /resetCache
warning

Manual role editing is a truly advanced technique for experts. Poorly executed or overly extensive changes can actually reduce the AI's consistency or functionality. Pavel Říha, for example, reports that 99.9% of the roles in his Online Strategist AI instance are generated automatically and have not needed manual edits. If in doubt, it's better to focus on optimizing your System Prompt and the quality of your know-how in sources. Most issues can be resolved through these standard approaches.

28.2 Changing Your AI's Voice for Audio Responses​

Your BuddyPro instance can respond not only in text but also via synthetic voice (e.g., when a user sends a voice message). You can choose from several available voices.

Use the following command to set or change the voice. Replace {voice} with the name of the desired voice (e.g., alloy, echo, fable, onyx, nova, shimmer).

/setVoice:{voice}
How to find available voices?

The most reliable source is the official OpenAI documentation, which includes a current list of available voices (both male and female) and sample audio. You can also ask your BuddyPro instance for available voices, but the documentation is the most up-to-date.

This feature adds an extra layer of personalization and can enhance user experience, especially for those who prefer listening over reading.

28.3 Experimenting with Advanced Diagnostic Commands​

For deep insights into your AI's "thinking" and detailed fine-tuning, you have access to a set of diagnostic commands. Using them requires patience and analytical skills.

  • /investigateAnswer:{message} — Responds to the given {message} and also displays the relevant parts of your know-how and role definition that were used in generating the answer. Great for verifying whether the AI is drawing from the correct sources.
  • /lastRole — Returns the name (ID) of the role that was used to generate the AI's most recent response. Useful for identifying which role might need manual revision.

By analyzing the outputs from these commands, you can identify subtle nuances in your AI's behavior and get ideas for further optimizing the System Prompt, the structure of your know-how, or expert subrole definitions. This is, however, a domain that requires experimentation and patience. Always proceed carefully and, ideally, test changes using a test profile first.

Last updated on