📚 Glossary

March 6, 2026 · View on GitHub

Quick Nav
Retrieval Playbook · Patterns · Eval


Core instruments

  • ΔS (delta-S)Semantic stress. ΔS = 1 − cos(I, G) where I is item embedding, G is ground/anchor.
    Thresholds: <0.40 stable · 0.40–0.60 transitional · ≥0.60 high risk.

  • λ_observeLayered observability state.
    Symbols: convergent · divergent · <> recursive · × chaotic.

  • E_resonanceCoherence indicator. Rolling mean of |residual| under BBMC; rising E with high ΔS ⇒ apply BBCR/BBAM.


Repair operators (WFGY modules)

  • BBMCBoundary-Bounded Memory Chunks. Reduce semantic residue vs anchors; align chunks/sections with tasks.
  • BBPFBranch-Bounded Prompt Frames. Explore multiple semantic paths safely; stabilize context windows.
  • BBCRBreak-Before-Crash Reset. Detect collapse, insert bridge node, restart cleanly.
  • BBAMAttention Modulation. Reduce variance in attention to avoid entropy melt on long contexts.

Retrieval, ranking & prompting

  • RRF (Reciprocal Rank Fusion) — Fuse ranks from multiple retrievers via 1/(k + rank).
  • MMR (Maximal Marginal Relevance) — Diversify candidates balancing relevance and novelty.
  • BM25 — Sparse lexical scoring for exact term match; strong for code/IDs.
  • HyDE — Hypothetical document expansion; creates a synthetic query/doc to improve recall.
  • Cross-encoder reranker — Jointly encodes [query ⊕ doc] for precision@k gains.

Patterns (named failure modes)

  • SCU (Symbolic Constraint Unlock) — Model merges sources or violates “cite-then-explain” schema. Fix: per-source fences + locked schema.
  • Query Parsing Split — Dense and sparse retrievers use different analyzers/tokenizers; hybrid breaks.
  • Vectorstore Fragmentation — Index contains “ghost” gaps; facts exist but never retrieved; fix shard/id audits.
  • Memory Desync — State flips between sessions/tabs; require mem_rev+mem_hash.
  • Role Drift — Multi-agent persona swap; tag agent IDs and lock via BBCR.

Multi-agent

  • Agent boundary — Contract that limits what an agent can read/write; prevents overwrite.
  • ΔS peer check — Measures divergence between agents’ plans to catch conflicts early.

Data & ops

  • Gold set — Small set (10–50) of realistic Q/A with citations; used for CI (recall@50, nDCG@10, ΔS).
  • Traceability — Provenance from answer → prompt → citations → chunks → source file.

Notation quickies

  • I — item (retrieved chunk) embedding; G — ground/anchor embedding.
  • ‖B‖ ≥ B_c — collapse threshold on residual magnitude (BBCR).
  • Convergent λ — layer is stable across paraphrases; divergent λ — layer is drifting.

🔗 Quick-Start Downloads (60 sec)

ToolLink3-Step Setup
WFGY 1.0 PDFEngine Paper1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>”
TXT OS (plain-text OS)TXTOS.txt1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly

Explore More

LayerPageWhat it’s for
⭐ ProofWFGY Recognition MapExternal citations, integrations, and ecosystem proof
⚙️ EngineWFGY 1.0Original PDF tension engine and early logic sketch (legacy reference)
⚙️ EngineWFGY 2.0Production tension kernel for RAG and agent systems
⚙️ EngineWFGY 3.0TXT based Singularity tension engine (131 S class set)
🗺️ MapProblem Map 1.0Flagship 16 problem RAG failure taxonomy and fix map
🗺️ MapProblem Map 2.0Global Debug Card for RAG and agent pipeline diagnosis
🗺️ MapProblem Map 3.0Global AI troubleshooting atlas and failure pattern map
🧰 AppTXT OS.txt semantic OS with fast bootstrap
🧰 AppBlah Blah BlahAbstract and paradox Q&A built on TXT OS
🧰 AppBlur Blur BlurText to image generation with semantic control
🏡 OnboardingStarter VillageGuided entry point for new users

If this repository helped, starring it improves discovery so more builders can find the docs and tools.
GitHub Repo stars