PHI // DRIFT: A Homeostatic Cognitive Architecture for AI Companions

PHI // DRIFT is a special type of software designed to give language models a persistent inner life. Instead of just answering questions, this tool creates a structured system where the AI remembers past conversations, tracks its emotional state, and manages its own needs. It works by assembling different parts of information into a single prompt for every interaction. This ensures that the AI behaves consistently over time rather than acting randomly or secretly executing code.

Benefits

PHI // DRIFT offers several unique advantages for users who want a more realistic and evolving AI companion.

• Persistent Memory: The system uses a memory module to store past interactions. This allows the AI to remember details from earlier conversations and use them in future chats.
• Emotional Awareness: The tool tracks five specific needs such as rest, connection, purpose, stimulation, and safety. These needs change over time, which helps the AI adjust its behavior to feel more human-like.
• Safe and Secure: Built-in security scans block malicious inputs and protect against harmful requests. The system also includes guardrails to prevent the AI from bypassing safety rules.
• Flexible Integration: It supports multiple AI models like Gemini, Groq, Kimi, and Ollama. Users can choose which model to use based on their needs and internet connection.
• Distributed Thinking: A consensus engine called the Hive Mind allows different parts of the system to discuss ideas before responding. This helps reduce errors and improve the quality of answers.

Use Cases

This product is useful for people who want a more advanced and personalized AI experience.

• Personal Companionship: Users can create a chatbot that remembers their stories and feelings. The AI can act as a friend, therapist, or listener who grows with the user over time.
• Creative Writing: Writers can use the tool to generate ideas or continue stories. The AI can maintain character consistency and remember plot points from previous sessions.
• Research and Analysis: Researchers can use the distributed cognition features to explore complex topics. The system can weigh different perspectives and provide balanced insights.
• Local Development: Developers can run the software on their own computers using local models. This is great for privacy-focused projects or those without reliable internet access.
• API Integration: Businesses can integrate the REST API into their own applications. This allows them to build custom tools that use the advanced cognitive architecture.

Pricing

Pricing information is not available for this product. It is an open-source project available under the Apache 2.0 license. Users can download and use it for free if they have the necessary technical skills to set it up.

Vibes

Public reception and reviews are not available for this product. The project is hosted on GitHub and is primarily used by developers and researchers interested in advanced AI architectures.

Additional Information

PHI // DRIFT is an open-source project hosted on GitHub at https://github.com/timeless-hayoka/infj-bot. The code is written in Python and requires Python version 3.12 or higher. Users need to install specific libraries including PyTorch for local embeddings and various API keys for different AI models.

The project includes a comprehensive test suite that checks for bugs and performance issues. Recent testing showed that removing certain features like the council system or shadow integration does not significantly change performance. However, removing the dynamic memory update feature resulted in a noticeable drop in prompt quality.

The system is designed to be extensible. Users can modify the code to add new features or change how the AI behaves. The project also provides a web interface for easier interaction and a command-line interface for advanced users.

The development team focuses on creating a robust and safe system. They have implemented strict security measures and regular testing to ensure reliability. The project aims to push the boundaries of what AI companions can do while maintaining safety and ethical standards.