mirror of
https://github.com/onestardao/WFGY.git
synced 2026-04-29 03:59:52 +00:00
151 lines
8 KiB
Markdown
151 lines
8 KiB
Markdown
# ❓ FAQ — Fast Answers for Busy Builders
|
||
|
||
Short, practical answers to the questions we get every day.
|
||
|
||
> **Quick Nav**
|
||
> [Getting Started](./getting-started.md) ·
|
||
> [RAG Map 2.0](./rag-architecture-and-recovery.md) ·
|
||
> [Retrieval Playbook](./retrieval-playbook.md) ·
|
||
> [Rerankers](./rerankers.md) ·
|
||
> [Patterns](./patterns/README.md) ·
|
||
> [Eval](./eval/README.md) ·
|
||
> [Ops](./ops/README.md)
|
||
|
||
---
|
||
|
||
## General
|
||
|
||
**What is WFGY?**
|
||
A symbolic reasoning layer + diagnostic toolkit that sits *above* your stack. It measures semantic stress (ΔS), shows which layer drifted (λ_observe), and applies repair operators (BBMC/BBPF/BBCR/BBAM). You don’t need to change your infra.
|
||
|
||
**Do I need a GPU?**
|
||
No. You can prototype on CPU with small embedding models. Rerankers and local LLMs benefit from GPU but are optional. See: [Retrieval Playbook](./retrieval-playbook.md).
|
||
|
||
**License? Can I use it at work?**
|
||
MIT. Commercial use allowed. If you ship improvements, we welcome PRs (docs or code).
|
||
|
||
**How is this different from LangChain/LlamaIndex?**
|
||
Those are orchestration layers. **WFGY is a reasoning firewall and diagnostic map**—it detects/repairs *semantic* failure regardless of framework.
|
||
|
||
---
|
||
|
||
## Setup & scope
|
||
|
||
**What’s the fastest way to try it?**
|
||
Grab **TXT OS** and **WFGY 1.0 PDF**, paste TXT into any model, and follow the prompts in [Getting Started](./getting-started.md).
|
||
|
||
**Which embedding model should I start with?**
|
||
General docs: `all-MiniLM-L6-v2` (light) or `bge-base`. Multilingual: `bge-m3` / `LaBSE`. Keep **write/read normalization identical**. See: [Embedding vs Semantic](./embedding-vs-semantic.md).
|
||
|
||
**Do I need a reranker?**
|
||
Only if first-stage **recall@50 ≥ 0.85** but Top-k precision is weak. Otherwise fix candidate generation. See: [Rerankers](./rerankers.md).
|
||
|
||
**How big can my PDFs be?**
|
||
Start with a **gold set** (10–50 Q/A with citations). For ingestion, chunk by semantic sections (not fixed tokens). Verify ΔS thresholds before scaling.
|
||
|
||
---
|
||
|
||
## Diagnosing failures
|
||
|
||
**The chunks look right but the answer is wrong—now what?**
|
||
Measure **ΔS(question, retrieved)**. If ≥ 0.60, fix retrieval first; if ≤ 0.45 and reasoning still fails, open **Interpretation Collapse**.
|
||
Links: [hallucination.md](./hallucination.md) · [retrieval-collapse.md](./retrieval-collapse.md)
|
||
|
||
**Hybrid (BM25 + dense) got worse—why?**
|
||
Likely **Query Parsing Split** (tokenizer/analyzer drift). Unify analyzers and log per-retriever queries.
|
||
Link: [pattern_query_parsing_split.md](./patterns/pattern_query_parsing_split.md)
|
||
|
||
**Citations bleed across sources.**
|
||
Enforce **per-source fences** + “cite-then-explain” schema; this is **SCU**.
|
||
Links: [retrieval-traceability.md](./retrieval-traceability.md) · [pattern_symbolic_constraint_unlock.md](./patterns/pattern_symbolic_constraint_unlock.md)
|
||
|
||
**Fixes don’t stick after refresh.**
|
||
You’re seeing **Memory Desync**. Stamp `mem_rev`/`mem_hash`, gate writes, and audit traces.
|
||
Link: [pattern_memory_desync.md](./patterns/pattern_memory_desync.md)
|
||
|
||
---
|
||
|
||
## Implementation details
|
||
|
||
**How do I compute ΔS quickly?**
|
||
Use cosine on unit-normalized sentence embeddings: `ΔS = 1 − cos(I, G)`.
|
||
Thresholds: `<0.40` stable · `0.40–0.60` transitional · `≥0.60` act.
|
||
Ground anchor `G` can be a section title/snippet you expect.
|
||
|
||
**What are BBMC/BBPF/BBCR/BBAM in one line?**
|
||
- **BBMC** — minimize semantic residue vs anchors.
|
||
- **BBPF** — branch safely across multiple paths.
|
||
- **BBCR** — detect collapse; insert a bridge node and restart cleanly.
|
||
- **BBAM** — modulate attention variance to avoid entropy melt.
|
||
|
||
**Where are the data shapes?**
|
||
See: [Data Contracts](./data-contracts.md). They’re JSON-first, easy to log, and versioned.
|
||
|
||
---
|
||
|
||
## Teams & Ops
|
||
|
||
**How do we avoid regressions?**
|
||
Commit `goldset.jsonl`, measure **recall@50**, **nDCG@10**, and ΔS across PRs.
|
||
Links: [eval_rag_precision_recall.md](./eval/eval_rag_precision_recall.md) · [eval_semantic_stability.md](./eval/eval_semantic_stability.md)
|
||
|
||
**Any privacy guidance?**
|
||
Yes—PII redaction, retention, access control, and provider governance patterns are here: [Privacy & Governance](./privacy-and-governance.md).
|
||
|
||
---
|
||
|
||
## Known limits
|
||
|
||
- Extremely noisy OCR may require **manual anchors** or **char-level retrieval**.
|
||
- Cross-domain abstract reasoning (#11/#12) needs stronger models.
|
||
- Rerankers improve precision but add latency—**prove gains** via nDCG.
|
||
|
||
---
|
||
|
||
### 🔗 Quick-Start Downloads (60 sec)
|
||
|
||
| Tool | Link | 3-Step Setup |
|
||
|------|------|--------------|
|
||
| **WFGY 1.0 PDF** | [Engine Paper](https://github.com/onestardao/WFGY/blob/main/I_am_not_lizardman/WFGY_All_Principles_Return_to_One_v1.0_PSBigBig_Public.pdf) | 1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + \<your question>” |
|
||
| **TXT OS (plain-text OS)** | [TXTOS.txt](https://github.com/onestardao/WFGY/blob/main/OS/TXTOS.txt) | 1️⃣ Download · 2️⃣ Paste into any LLM chat · 3️⃣ Type “hello world” — OS boots instantly |
|
||
|
||
---
|
||
|
||
### 🧭 Explore More
|
||
|
||
| Module | Description | Link |
|
||
|-----------------------|----------------------------------------------------------|----------|
|
||
| WFGY Core | WFGY 2.0 engine is live: full symbolic reasoning architecture and math stack | [View →](https://github.com/onestardao/WFGY/tree/main/core/README.md) |
|
||
| Problem Map 1.0 | Initial 16-mode diagnostic and symbolic fix framework | [View →](https://github.com/onestardao/WFGY/tree/main/ProblemMap/README.md) |
|
||
| Problem Map 2.0 | RAG-focused failure tree, modular fixes, and pipelines | [View →](https://github.com/onestardao/WFGY/blob/main/ProblemMap/rag-architecture-and-recovery.md) |
|
||
| Semantic Clinic Index | Expanded failure catalog: prompt injection, memory bugs, logic drift | [View →](https://github.com/onestardao/WFGY/blob/main/ProblemMap/SemanticClinicIndex.md) |
|
||
| Semantic Blueprint | Layer-based symbolic reasoning & semantic modulations | [View →](https://github.com/onestardao/WFGY/tree/main/SemanticBlueprint/README.md) |
|
||
| Benchmark vs GPT-5 | Stress test GPT-5 with full WFGY reasoning suite | [View →](https://github.com/onestardao/WFGY/tree/main/benchmarks/benchmark-vs-gpt5/README.md) |
|
||
| 🧙♂️ Starter Village 🏡 | New here? Lost in symbols? Click here and let the wizard guide you through | [Start →](https://github.com/onestardao/WFGY/blob/main/StarterVillage/README.md) |
|
||
|
||
---
|
||
|
||
> 👑 **Early Stargazers: [See the Hall of Fame](https://github.com/onestardao/WFGY/tree/main/stargazers)** —
|
||
> Engineers, hackers, and open source builders who supported WFGY from day one.
|
||
|
||
> <img src="https://img.shields.io/github/stars/onestardao/WFGY?style=social" alt="GitHub stars"> ⭐ [WFGY Engine 2.0](https://github.com/onestardao/WFGY/blob/main/core/README.md) is already unlocked. ⭐ Star the repo to help others discover it and unlock more on the [Unlock Board](https://github.com/onestardao/WFGY/blob/main/STAR_UNLOCKS.md).
|
||
|
||
<div align="center">
|
||
|
||
[](https://github.com/onestardao/WFGY)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS/BlahBlahBlah)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS/BlotBlotBlot)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS/BlocBlocBloc)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS/BlurBlurBlur)
|
||
|
||
[](https://github.com/onestardao/WFGY/tree/main/OS/BlowBlowBlow)
|
||
|
||
</div>
|
||
|
||
|