CoDY the AI Therapist

There are three essential components in the model: ‎1. Co:here Classify model that has been trained on almost 500,000 real training data ‎‎(sentence, emotion) and is capable of using NLP to truncate and classify large ‎prompts into a range of emotions. ‎2. Co:here Embedding + Sklearn model that transforms the user prompt to a list of ‎floats (embeds) and performs semantic search in a massive data base of real, ‎labeled, embedded sentences to find the closest matching example.‎ ‎3. Co:here Generate model that receives user input, classified emotions, previous ‎calls, and some hard coded psychological information to generate an accurate ‎therapist response. The model has memory of past inputs and thus can create a ‎coherent conversation.‎ These components work together in the back end, while the front end is a webpage ‎that user can access to send information and read model's responses.‎ Ultimately, the project aims to create an ongoing conversation with Co:DY and user, ‎providing emotional support, helping users work through problems, and offering ‎tailored advice.‎