grapeot/context-infrastructure

What it solves

It addresses the problem of AI providing generic, "correct but useless" responses by providing a structured framework for building a personalized context infrastructure. It allows users to move AI from a general-purpose tool to a personalized assistant that understands the user's specific identity, preferences, and accumulated knowledge.

How it works

The system operates as a reference implementation (a blueprint) rather than a plug-and-play tool. It organizes information into a three-layer structure:

• Presentation Layer: Contains distilled axioms and skills accumulated over a year, serving as examples of how to structure knowledge.
• Reusable Layer: Provides templates for identity (SOUL.md), user preferences (USER.md), and communication styles (COMMUNICATION.md), as well as code for a memory system (ai_heartbeat).
• Non-Reusable Layer: The specific experiences and insights that the user must collect from their own behavior data.

It utilizes a memory system with daily observation scripts (observer.py) and weekly reflection scripts (reflector.py) that can be configured via cron jobs to maintain a continuous loop of learning and memory accumulation.

Who it’s for

Individuals who want to personalize their AI assistants (using tools like Claude Code, Cursor, or OpenCode) and are willing to invest the time to collect their own behavioral data to create a high-context AI experience.

Highlights

• Reference Implementation: A complete structural blueprint of a system that has been operational for a year.
• Three-Layer Memory System: A structured approach to L1/L2 memory via OBSERVATIONS.md.
• Automated Memory Maintenance: Includes scripts for daily observations and an "AI heartbeat" for weekly reflections.
• Extensible Skill Ecosystem: Supports integration with external public skill repositories for capabilities like web search, Google Docs, and payment analysis.