noteOS

This repository is part of MetaMemoryWorks —
a file-based architecture for persistent AI memory.

🌐 http://metamemoryworks.com
📄 Architecture: https://github.com/johannes42x/MetaMemoryArchitecture

noteOS is a file-based, memory-driven universal note and knowledge logging system for LLMs. It provides a structured method for receiving arbitrary user input, routing it into the correct log, normalizing it into a deterministic entry schema, and appending it to persistent long-term memory.

noteOS is part of the MetaMemoryWorks architecture, which defines a general method for persistent, file-driven cognitive state management in LLM environments.

---

1. Purpose

noteOS enables large language models to:

• accept any kind of user input (speech, text, uploads, OCR, PDFs, screenshots),
• normalize these inputs into structured entries,
• route them into the correct log (e.g., notebook, work log, project log, conflict log, idea log),
• store them in persistent append-only memory files,
• and retrieve or summarize information deterministically based on accumulated memory.

noteOS is not a standalone program. It is both:

1. a concrete implementation that provides structured persistent memory for all note-like information, and
2. a formal method for building file-based logging systems with deterministic behavior.

---

2. Core Method

1. Input Acquisition Free-form input (typed, spoken, uploaded) is received. noteOS may perform OCR when required.

2. Routing Layer noteOS determines the correct target log using:

   • explicit user target (highest priority),
   • LLM inference when confident,
   • fallback to residual log when unclear.

3. Normalization Inputs are transformed into a unified structured entry format. noteOS gives the LLM explicit instructions to avoid inferring missing data. Unknown or ambiguous values remain null unless the user provides them.

4. Persistent Storage Entries are appended using deterministic update rules (append-only, tail markers, sequential IDs).

5. Deterministic Retrieval & Organization Summaries, listings, reorganizations, and meta-operations rely strictly on stored entries.

---

3. Implementation Notes

• noteOS may perform OCR; extraction from uploads is supported either internally or by upstream systems.

• Each log contains:

  • header (metadata, cues, schema),
  • meta (state, last assigned ID),
  • entry_template,
  • entries.

• LLMs can be unpredictable; noteOS therefore defines strict instructions to reduce inference, encouraging explicit null values instead of guessed data.

• Logs may be large. To manage this, noteOS employs snippet-handling, which increases reliability when only partial file contexts are visible.

• Routing and engine specifications define the deterministic behavior and safety constraints.

• noteOS provides persistent memory through the ability to embed arbitrarily many logs for different domains (notebooks, projects, conflicts, research, ideas).

---

4. Method Specification and License

The protected method is defined in:

• METHOD_SPEC_noteOS_v1.0.md
• ENGINE_SPEC_noteOS_v1.0.md
• noteOS_routing.json
• noteOS_engine_header.json

Usage and redistribution terms are defined in:

• LICENSE_noteOS_v1.0.md

Any system that receives arbitrary input, normalizes it, routes it, and stores it in append-only logs is considered an implementation of this method.

---

5. Status

noteOS v0.9 provides:

• universal structured logging
• deterministic routing + engine integration
• append-only persistent memory
• batch append and preview workflows
• snippet-handling for large logs
• flexible template-based log architecture

---

6. Contact

For licensing or integration inquiries:

kontakt@metamemoryworks.de