Summary

Neocortex AI Memory addresses the challenge of traditional AI memory systems being overwhelmed by irrelevant data, leading to degraded performance. It offers a human-brain-inspired approach to AI memory, intelligently pruning noise and reinforcing crucial information. This system targets developers building AI applications, providing a lean, focused, and rapidly accessible memory layer that enhances AI responses with a massive context window at a low cost.

How It Works

Neocortex mimics human cognitive processes by aggressively filtering noise and allowing low-value memories to decay naturally, while strengthening frequently accessed knowledge. Its core mechanisms include "Intelligent Noise Filters" where memories fade without interaction, and "Interaction-Aware" systems that prioritize memories based on user engagement signals like views and replies. A key differentiator is "Conscious Recall," which proactively surfaces the most relevant memories by combining recent activity with time-based decay, offering LLMs a focused context slice rather than an undifferentiated data dump. This approach avoids overwhelming LLMs and reduces token costs.

Quick Start & Requirements

Installation is straightforward via pip install tinyhumansai for the Python SDK. The system requires an API key obtained from tinyhumans.ai. Notably, Neocortex is currently in closed alpha, necessitating direct contact with founders@tinyhumans.ai for access. It supports numerous SDKs (Python, JS, Go, Rust, etc.) and integrates with popular frameworks like LangGraph and CrewAI. Official documentation is available at tinyhumans.gitbook.io/neocortex/.

Highlighted Details

• Supports a context window exceeding 10 million tokens at speeds up to 4000 tokens/second.
• Claims low operational costs ($0.2/Mn tokens).
• Achieves top-tier benchmark scores in RAGAS (Answer Relevancy 0.97, Context Precision 0.75) and TemporalBench (100% recency accuracy).
• Features "Conscious Recall" for proactive, contextually relevant memory surfacing.

Maintenance & Community

Community channels include Discord, Reddit, and X. Comprehensive documentation is available via GitBook.

Licensing & Compatibility

The provided README does not specify a software license. This lack of explicit licensing information may present restrictions for commercial use or integration into closed-source projects.

Limitations & Caveats

The project is currently in a closed alpha phase, requiring users to reach out for access and API keys. The absence of a stated software license in the documentation is a significant caveat for potential adopters.