Charm 2.1 Ai Companion
Email me if anyone is interested in this...
A. Core Architecture
Charm is built in Python, with multiple subsystems for advanced interactive capabilities. It utilizes a SQLite3 database for structured chat history and long-term memory. Each memory summary is transformed into a high-dimensional semantic vector, enabling retrieval of past interactions based on meaning. Memory recall commands are available for easy access to specific memories.
B. Vision Subsystem
The vision subsystem processes image inputs and analyzes visual patterns, emotions, and creative insights. It encodes images in Base64 format and utilizes image recognition algorithms for analysis.
C. Speech Subsystem
This subsystem integrates with a Text-to-Speech engine, allowing text responses to be converted into audio files with configurable voice settings.
D. News Subsystem
Charm retrieves real-time news data from sources like BBC News and provides summaries, key insights, and follow-up questions on the news. Users can trigger news commands to stay informed.
E. Sentiment Analysis
Charm uses VADER and TextBlob to evaluate text sentiment, producing a hybrid sentiment score. The system returns a tuple including the hybrid score, full VADER scores, and TextBlob polarity.
F. Multi-Agent Pipeline
Charm employs specialized agents for conversation processing, including ReflectionAgent, PlanningAgent, and ResponderAgent, to facilitate efficient and effective communication.
G. Consciousness Monitoring (Conscious Core)
The system continuously computes an internal “consciousness score” as a heuristic measure for internal dynamics, helping to modulate response tone and guide multi-agent outputs.
H. Wikipedia Integration
Users can access Wikipedia summaries by issuing specific commands such as "get wiki article [topic]" or using the shortcut "wiki: [query]". This integration provides a wealth of knowledge on various topics.
I. Overall Command Overview
Charm offers various commands for memory recall, news retrieval, and Wikipedia summaries. It also automatically analyzes sentiment and consciousness to enhance responses and guide interactions with users.
J. Training Data Logging
All interactions with Charm are logged into a training database for continuous refinement and improvement of future interactions. The stored data includes user input, assistant responses, timestamps, and optional ratings.
