Handy Augmented AI chatbot - React App

Software
21

bookmarking for when github gets posted

1 Like
22

I have been hoping someone would be designing something like this for a while. I am pretty new to AI Chat stuff and had a great time making characters and interacting with them on Yodayo, but they recently stopped hosting any adult content which killed the site for me. How does one obtain an OpenAI API Key? From what I gather it costs money per use? I assume because of how much power it takes to run an AI it gets expensive? Keep up the good work with this project. I hope to try it soon.

23

Do you have a repo? It would be interesting to give it a look. If you intend to do a local version it would be pretty easy to add in support for other back ends without much change needed.

direct it to a local ip since most apps use an openai compatible api set up there isnā€™t much that would need to be changes.

I tried a similar project a couple months back and couldnā€™t figure out a way to convert the text to movements. how did you manage to get it working?

24

OpenAI has made their real time API available which includes real time voice to speech and vice versa, should be interesting to see

1 Like
25

Heres the github repo. The main files are

src/app.js
src/components/AI.js (LLM response parser)
src/components/Hardcode.js (LLM to Handy action function connection logic)
src/components/HandyFunctions.js (Handy Actions Functions)

Feel free to fork it and push changes.

Dm me if you have questions or concerns

26

posted below, would be cool if there were other custom handy projects other than faptap

27

Heres the github repo. The main files are

src/app.js
src/components/AI.js (LLM response parser)
src/components/Hardcode.js (LLM to Handy action function cnnection logic)
src/components/HandyFunctions.js (Handy Actions Functions)

Feel free to dm me and push the code regarding custom local models when ready. Iā€™ll take a look and add the option to this and future versions.

1 Like
28

posted below, I think parsing the key words isnā€™t that hard but getting the handy to activate can be problematic. The handy sdk readme is kind of a guide

29

Git is confusing dont worry but you should merge your version 2 branch into main. This will essentially take the changes of version 2 branch and apply them to the main branch.
so it will be like

git checkout main

Merge version-2 into main, taking all changes from version-2

git merge -X theirs version-2

Finalize and execute changes.

git push origin main
30

Yeah the main branch was for another project ill probably get rid of. In the meantime ive made some updates to the UI and added other features

31

Hi Carly, if the community likes this, you may share it on my behalf. Having issues with accounts

32

Yeah it cost money ($5) but it will last almost forever because cost per conversation is like .001 cents. You can easily get one from openai api website, chatgpt can give you a good guide or YouTube.
The AI is censored meaning it cant be too ā€œadultā€. Im working on ways to incorporate uncensored models.
I added characters since people seems to want that, let me know how it look

2 Likes
33

Looking forward for the option to connect oobabooga and use a local LLM instead of OpenAI! Great idea and project!

1 Like
34

The GitHub link 404s for me.

I did take this week to build out my own setup using Ollama for running local uncensored models and integrating handy control via custom tools parsing. Got some good results so far, but I donā€™t really have the hardware to run things instantly. For me the response stage takes around 30-45s which can be pretty rough.

35

I created a similar thing for mobile (little project of mine) using openrouter instead of openAI a couple of months ago, Iā€™d suggest looking into it (you can use openAI api for openrouter iirc) - you could have the user enter which model they want to use, they have some uncensored and free ones :slight_smile:

36

Has there been any further development on this? And is there a way to run this app locally instead of through a web gui?

Not sure if Iā€™m willing to sent private API keys over the web to an unknown source, but I think this is a cool project.

1 Like
37

Hey yeah Iā€™m actively making commits to the live site. Iā€™ll make another post once Im satisfied with the new features im adding as well as fixing the bugs.

since my last post I added Text-to-audio and text-to-speech, so you can talk to the llm using your voice and it can control the Handy. Its 60% there but I used it and I think it has a lot of potential for exploring new things.

I am trying to make the censored model free without the need for an API, Iā€™ve had people say that obtaining an API key is prohibitively complex.

Iā€™ll also be adding the option to use an uncensored model that can have X rated conversations by dynamically renting out a gpu and hosting a model.

Regarding the api key over the web, yeah its not good practice but its the simplest option to get this running. It looks like people use this app and I have not heard any complaints, I personally use it and havenā€™t seen issues. Iā€™ll try to get rid of the needs for an api key soon

Iā€™ll also add a settings to tune the functions as well

the disconnect&reset button is bugged, Iā€™ll fix it today

3 Likes
38

Yeah i had to take it down since i need to use the github for professional use. Im planning on adding features that might require robust security so idk if sharing the source code would be good also.

Ill be adding uncensored models (at the cheapest market rate i can find) for those that dont have NASA level gear. I add an option to use local models on PC if users feel this is their thing and want to invest in that kind of gear.

I might make accounts so users can save their character conversation.

1 Like
39

Just chiming in to say that I would be very interested in trying this with my local model setup. I donā€™t have anything too fancy, just a 12B model running entirely on a 4070, but it does the trick.

1 Like
40

Hey, Iā€™ve been following this project for a while and wanted to say thanks for your efforts in building it!

I totally understand needing to take it down from GitHub for professional appearances. It sucks that thereā€™s still such stigma around sex and sex-adjacent topics, but thatā€™s the reality weā€™re in right now and you should absolutely prioritize protecting your own livelihood.

Have you considered open sourcing it under an alt account with a pseudonym? Thatā€™s what I see many people doing in this space, for example the creator(s) of the popular porn organizer app Stash GitHub - stashapp/stash: An organizer for your porn, written in Go. Documentation: https://docs.stashapp.cc

Iā€™d personally love to contribute to your project like others have offered in the thread, but contributing wonā€™t be possible if itā€™s not open source anymore :disappointed:

Please know that adding features requiring robust security doesnā€™t mean a project canā€™t be open source. Security researchers have a saying about security through obscurity, which boils down to: hiding the implementation of security doesnā€™t make something more secure. In fact, there are many security-focused open source projects which benefit from being open source, as developers with a security background can verify the project is as secure as it seems and contribute improvements in that area if needed. See for example:

Iā€™d also be more than happy to help if you have questions about how to open source particular features securely :smile:

Regarding LLM options, Iā€™ve used Mancer in the past myself, they have a wide range of LLMs you can choose from, all unfiltered and private, under one API: https://mancer.tech/
Iā€™m not affiliated with them, but my experience has been great!

2 Likes