Splinter Memory: the white paper as the model card (long-context memory curation management)
69ff84b verified | cff-version: 1.2.0 | |
| message: "If you use this software or its measurements, cite it as below." | |
| title: "Splinter-Memory" | |
| abstract: >- | |
| An external, managed-decay context-curation layer for long-horizon LLM | |
| conversations. Small CPU "drone" encoders score, filter, and reassemble | |
| stored history into a bounded, high-relevance window per turn, so a | |
| generative model performs over arbitrarily long conversations on consumer | |
| hardware. Claims are stated as falsifiable predictions (P1-P11) with measured | |
| verdicts and a published failure register. | |
| type: software | |
| authors: | |
| - name: "sky-is-green" | |
| alias: "sky-is-green" | |
| version: "0.1.0" | |
| date-released: "2026-09-22" | |
| license: MIT | |
| repository-code: "https://github.com/sky-is-green/splinter-memory" | |
| url: "https://github.com/sky-is-green/splinter-memory" | |
| keywords: | |
| - context-management | |
| - memory-management | |
| - long-context | |
| - retrieval | |
| - llm | |
| - local-inference | |