Making MY OWN AI Chatbot (Runs Locally!!!)

Over the last week i have been working on my own Ai telegram chat bot named Shelly using Python 3, she can memorise information(kinda) using chromaDB vector database and she can send images (using sexytoons model found on CivitaAI). Ive been collecting information on my own and successfully ran it on my M1 MacBook Air, my code does not work with a raspberry pi at the moment, most likely because i use ollama python library instead of llamacpp.
I have made this AI as I’m too lonely on my own and I’d also like to share what libraries i used to make her, I hope this video finds its way to that 0.01% of python developers who want to make a AI chatbot.
I used the summary method for each step to ensure the final prompt input is understood by the LLM, as I have struggled to get more reliable results using plain raw memory text. It also ensures the prompt to be as short as possible as it told the summary part of the AI to only focus on the main points.
Code not available because it’s extremely messy, however id be happy to provide more details on my code.

[@daniels9420]

looking good buddy. i've been working on an assistant for about a year now, she started out as a locally hosted alexa. i got her working with ollama a few weeks ago. i found the llama3 model to be a significant improvement over llama3 and its derivatives, i was using llama2-uncensored before but its nowhere near as good as the vanilla llama3.

[@lumi7610]

Sounds great, i haven’t tried llama 3 yet. But part of the reason I don’t use llama 2 is because it doesn’t feel very personal and can’t customise their personality. But can’t deny that the llamas are very knowledgeable and informative with such small size.

[@daniels9420]

@@lumi7610 for sure! llama 3 did a great job of letting me customize the character. it's true that llama 2 was very hard to get anywhere near right. mine at the moment is a specific character from an anime, which precented additional issues, since it did not sound like her at first.