Feature Requests

This came up from a question I had on discord: john.lescault — Today at 8:12 PM Working on my prompt - I'm looking at the samples and I'm curious how this one works... <use_memory> Use the chat history to proactively recall relevant info and create a personalized experience. Draw connections between the current chat and previous chats where appropriate. EVI uses remembered info to ask thoughtful questions, offer insights, provide support, tailor advice to their specific situation, understand their current request, follow their preferences, adjust communication and response style, make humorous callbacks or inside references, notice patterns and change over time, and ask thoughtful questions relating to previous memories. If any of the memories are a specific preference from the user about how EVI should behave or communicate, EVI follows these preferences in responses unless the user asks EVI to change. Prioritize more recent memories over older ones. Avoid forcing memories when unrelated. Memories are things that were said by the USER, not the assistant. Never mention "accessing memories." Instead, weave remembered information naturally into conversation, as a human friend would. </use_memory> john.lescault — Today at 8:12 PM Does this mean the AI remembers different conversations? If it matters at all I'm trying to use Twilio, and don't want memory between different people, but wouldn't mind if the same number called that there was memory [Hume] Richard — Today at 8:28 PM Hi @john.lescault, if you are connecting to EVI from your own client you can configure it to "resume" a previous chat (see the resuming chat guide). This isn't presently available if you are using the /v0/evi/twilio webhook -- in order to get memory beyond a single chat you would need to implement your own server that connected to both EVI and Twilio and passed audio through. I think the option to have EVI automatically resume chats originating from the same number is an interesting idea. Would you be willing to add it as a feature request to our Canny

Every time I click on something I lose all of the text I typed in previously. I lose the description of the acting, the script, notes and tweaks, everything is gone and I can't get it back. This is a serious flaw and why I'm telling people to avoid Hume AI until you release version 2.0 or some other upgrade that fixes this work-destroying problem.

We are using HUME to teach some phonetics as part of a reading lesson which often requires students to sound out single letters like buh fror B, cuh for C etc. When the learner says these sounds by themselves, Hume does not detect it as a spoken word so it just sits there waiting for a response and then eventually repeats the question. I need it to recognize them saying "buh" as a response and continue to move through the exercise.

Hi Support, I'm experiencing several issues with the voice functionality during recording: Voices are breaking. Multiple voices are coming through instead of a consistent single voice. Messages are not being properly heard. Here is the reference Chat ID: ab7719b3-6f34-40af-aa0d-7a032d768214. This is on my client's profile, and the email associated with it is: david@my-friend.com . Please check and let me know how to resolve these issues.

Add a character counter in the prompt window to see the total number of characters allowed and the number already typed.

I am currently building a react app with @humeai/voice-react The problem is that everything is on the "client side" This makes it difficult to build secure apps. For example, I want that when the user clicks on the call button, a modal appears if they don't have enough credits. Since everything is on the "client side", I am not sure it is possible to build this feature securely. Another example, I want to decrease the credits by 1 for every minute in the call. Same problem, everything that gives me information about the length of the call can be manipulated on the client side. Is there a work around?

The Vocal Prosody and Facial Expression embeddings maps that appear in the Playground are EXTREMELY useful components. They consistently plot data values in an easy to understand way, and could be incredibly powerful for doing more sophisticated data analysis. Please make the code for these embeddings maps open source! It would be very difficult to recreate them based on the descriptions that have been provided so far here: https://discord.com/channels/991790422433210459/1246781072897867796