everyoneisgross/mnemosynebot

a simple offline chat agent with dynamic contextual memory recall and "forgetting".

mnemosyneBOT

==============================

mnemosyneBOT, inspired by the Greek goddess of memory, is a sketch of a conversational agent using local GPT4ALL. It manages conversational context by integrating sentiment analysis and a memory decay mechanism. This allows it to gauge the vibe of user input and prioritize interactions based on their emotional significance and temporality. The bot's approach aims to create a more human-like conversational experience, viewing conversations as emotional narratives influenced by time, rather than isolated exchanges. This was just me following an idea until I had it run while getting my own memory aligned to the idea of vectors. 📎🧠💅

Features

---

Memory Decay Mechanism:

The agent uses a decay function to give more weight to recent interactions and reduce the influence of older interactions. This simulates the concept of "forgetting" in human memory, ensuring that the agent's responses are more aligned with recent context.

Sentiment-Based Context Selection:

Depending on the sentiment of the user's input, the agent chooses between different methods to retrieve relevant conversational context. This ensures that the agent's responses are more emotionally attuned to the user's sentiments.

Embedding & Data Handling:

The agent uses TF-IDF Vectorization to convert text data into embeddings. These embeddings are then used to calculate similarities and retrieve relevant conversational contexts.

History Handling:

The agent maintains a history of recent interactions to avoid repetitive responses and to provide contextually relevant answers.

Data Persistence:

Embeddings, vectorizers, and interaction data are persistently stored in files, allowing the agent to maintain its state across sessions.

GPT4ALL Integration:

The agent is built on top of the GPT4ALL python api, using a local model.