| deepmemory |
Project context, data model, and CLI for Deep Memory — a knowledge compression project that maps the origins of humans so AIs can better understand and interact with them. Use when running extractions, placing files, scaffolding sources, or deploying signal work. |
1.2.0 |
| deepmemory-www |
Website administration for Deep Memory — rendering, database, HUD, event system, and server operations. Use when maintaining the website, managing the database, or debugging the live site. |
1.2.0 |
| extract |
Operator pipeline for processing per-video FFT signals into a denoised master signal using nlsp. Use when running a signal extraction run for any Deep Memory source. |
1.2.0 |
| ingest |
Acquire video transcripts from YouTube URLs using AssemblyAI (speaker diarization) or yt-dlp (auto-captions fallback), generate structured metadata with video info, and validate transcript format. Use when extracting transcripts from YouTube videos, processing video content for analysis, or when the user provides YouTube URLs that need transcription. |
1.0.0 |
| nlsp |
Natural Language Signal Processing — extract, merge, filter, and analyze semantic signals from text. Decomposes natural language into axioms, claims, and details, then compresses them into dense master signals. Part of the Deep Memory pipeline. |
1.1.0 |
| patch |
Patch video signals into master signals through FFT ingestion and signal extraction. Use when processing YouTube videos for a source - main agent handles FFT batch processing, operator sub-agent performs signal extraction (denoising + consensus building), producing the master signal. |
1.0.0 |