mirror of
https://github.com/onestardao/WFGY.git
synced 2026-04-28 11:40:07 +00:00
252 lines
12 KiB
Markdown
252 lines
12 KiB
Markdown
<!--
|
||
Search Anchor:
|
||
memory long context global fix map
|
||
llm long context stability
|
||
multi session memory drift
|
||
memory collapse map
|
||
ghost context bug
|
||
state fork multi tab
|
||
pattern memory desync
|
||
long conversation drift
|
||
multi day chat loses task
|
||
support thread rewrites history
|
||
ocr transcript jitter over time
|
||
attention melt in long windows
|
||
entropy collapse long context
|
||
context drift late in chain
|
||
cache hazard cross tab
|
||
refresh changes answer
|
||
model recalls phantom text
|
||
same task id different state
|
||
same ticket different summary
|
||
token window rollover bug
|
||
task state lost after restart
|
||
facts flip after model switch
|
||
citations correct but reasoning wrong
|
||
capitalization spacing drift in transcript
|
||
long meeting notes inconsistent answers
|
||
agent memory desync between tools
|
||
|
||
When to use this folder:
|
||
dialog grows past 50k tokens and answers degrade
|
||
multi day support or ops thread loses state
|
||
facts change after tab refresh or model change
|
||
citations look right but reasoning goes flat or chaotic
|
||
ocr export text slowly drifts from original scan
|
||
multi session agent runs disagree on same task id
|
||
user comes back next day and system forgets constraints
|
||
model remembers details from previous unrelated session
|
||
you see ghost messages that were never shown to user
|
||
context cut in middle of sentence at window joins
|
||
|
||
Key metrics and targets:
|
||
retrieval coverage >= 0.70 to intended section
|
||
delta s question retrieved <= 0.45
|
||
delta s at window joins <= 0.50
|
||
lambda observe convergent across three paraphrases
|
||
no state fork for same task_id across tabs or agents
|
||
no ghost context after clearing cache and reload
|
||
ocr exports stable across runs on the same file
|
||
capitalization and spacing variance bounded
|
||
memory fences respected across restarts
|
||
traceability log shows a single continuous story
|
||
|
||
Core pages in this folder:
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/memory-coherence.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/entropy-collapse.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/context-drift.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/pattern_memory_desync.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/ghost-context.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/state-fork.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/ocr-parsing-checklist.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/ocr-jitter.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/retrieval-traceability.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/data-contracts.md
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/chunking-checklist.md
|
||
|
||
Related structural fixes:
|
||
ProblemMap/rag-architecture-and-recovery.md
|
||
ProblemMap/retrieval-playbook.md
|
||
ProblemMap/retrieval-traceability.md
|
||
ProblemMap/data-contracts.md
|
||
ProblemMap/context-drift.md
|
||
ProblemMap/entropy-collapse.md
|
||
ProblemMap/GlobalFixMap/Chunking/README.md
|
||
ProblemMap/GlobalFixMap/OCR_Parsing/README.md
|
||
ProblemMap/SemanticClinicIndex.md
|
||
|
||
Memory and context scenarios:
|
||
long running research chat forgets earlier assumptions
|
||
multi step planning session answers differently next day
|
||
customer support ticket summary changes after refresh
|
||
same meeting transcript yields different action items
|
||
legal or medical thread must stay consistent over weeks
|
||
agent orchestration keeps desyncing shared memory store
|
||
cache warm vs cold produces different story
|
||
sandbox vs production answers differ with same logs
|
||
long novel or book chat loses early chapters
|
||
training or tutoring session loops or forgets progress
|
||
|
||
Signals to check:
|
||
coverage good but lambda unstable over time
|
||
delta s low locally but high at joins
|
||
citations stable yet narrative rewrites earlier events
|
||
thread repeats same question with different answers
|
||
user corrects model but mistake returns later
|
||
state differs by browser tab or device
|
||
ocr text diff shows small random changes over runs
|
||
session id or task id missing from logs
|
||
traceability logs have gaps between days
|
||
embedding analyzer changed between deployments
|
||
|
||
Normalization and contracts:
|
||
always require task_id session_id snippet_id
|
||
log model version and deployment id
|
||
lock same embeddings and analyzers across sessions
|
||
store snapshots for long context boundary checks
|
||
tie memory store updates to explicit commits not every turn
|
||
-->
|
||
|
||
<!--
|
||
Cross folder jumps:
|
||
ProblemMap/GlobalFixMap/MemoryLongContext/README.md
|
||
ProblemMap/GlobalFixMap/Reasoning/README.md
|
||
ProblemMap/GlobalFixMap/Chunking/README.md
|
||
ProblemMap/GlobalFixMap/OCR_Parsing/README.md
|
||
ProblemMap/rag-architecture-and-recovery.md
|
||
ProblemMap/retrieval-playbook.md
|
||
ProblemMap/retrieval-traceability.md
|
||
ProblemMap/data-contracts.md
|
||
ProblemMap/context-drift.md
|
||
ProblemMap/entropy-collapse.md
|
||
ProblemMap/SemanticClinicIndex.md
|
||
-->
|
||
|
||
|
||
# Memory & Long-Context — Global Fix Map
|
||
|
||
<details>
|
||
<summary><strong>🏥 Quick Return to Emergency Room</strong></summary>
|
||
|
||
<br>
|
||
|
||
> You are in a specialist desk.
|
||
> For full triage and doctors on duty, return here:
|
||
>
|
||
> - [**WFGY Global Fix Map** — main Emergency Room, 300+ structured fixes](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/README.md)
|
||
> - [**WFGY Problem Map 1.0** — 16 reproducible failure modes](https://github.com/onestardao/WFGY/blob/main/ProblemMap/README.md)
|
||
>
|
||
> Think of this page as a sub-room.
|
||
> If you want full consultation and prescriptions, go back to the Emergency Room lobby.
|
||
</details>
|
||
|
||
Stabilize **long windows** and **multi-session memory**.
|
||
This map helps you repair drift, collapse, forks, and ghost contexts when conversations or documents stretch far beyond the usual size.
|
||
|
||
---
|
||
|
||
## What this page is
|
||
- A **beginner-friendly checklist** for long contexts and multi-day sessions.
|
||
- **Copy-paste guardrails** that stop drift and collapse before they spread.
|
||
- **Concrete metrics** with ΔS and λ_observe so you know if your system is stable.
|
||
|
||
---
|
||
|
||
## When to use
|
||
- Dialog grows past **50k–100k tokens** and answers degrade.
|
||
- Facts flip after **tab refresh** or **model switch**.
|
||
- Citations look right but reasoning goes **flat or chaotic**.
|
||
- OCR transcripts look fine but **capitalization or spacing drift**.
|
||
- Multi-day support threads **lose task state** or **rewrite history**.
|
||
|
||
---
|
||
|
||
<!--
|
||
Anchor Menu:
|
||
open: memory coherence guide ProblemMap/GlobalFixMap/MemoryLongContext/memory-coherence.md
|
||
open: entropy collapse long context guide ProblemMap/GlobalFixMap/MemoryLongContext/entropy-collapse.md
|
||
open: context drift long window guide ProblemMap/GlobalFixMap/MemoryLongContext/context-drift.md
|
||
open: pattern memory desync guide ProblemMap/GlobalFixMap/MemoryLongContext/pattern_memory_desync.md
|
||
open: ghost context guide ProblemMap/GlobalFixMap/MemoryLongContext/ghost-context.md
|
||
open: state fork guide ProblemMap/GlobalFixMap/MemoryLongContext/state-fork.md
|
||
open: ocr parsing checklist ProblemMap/GlobalFixMap/MemoryLongContext/ocr-parsing-checklist.md
|
||
open: ocr jitter guide ProblemMap/GlobalFixMap/MemoryLongContext/ocr-jitter.md
|
||
open: long context retrieval traceability ProblemMap/GlobalFixMap/MemoryLongContext/retrieval-traceability.md
|
||
open: long context data contracts ProblemMap/GlobalFixMap/MemoryLongContext/data-contracts.md
|
||
open: long context chunking checklist ProblemMap/GlobalFixMap/MemoryLongContext/chunking-checklist.md
|
||
|
||
jump: rag architecture and recovery visual map ProblemMap/rag-architecture-and-recovery.md
|
||
jump: retrieval playbook knobs and metrics ProblemMap/retrieval-playbook.md
|
||
jump: global retrieval traceability and data contracts ProblemMap/retrieval-traceability.md ProblemMap/data-contracts.md
|
||
jump: general context drift and entropy collapse ProblemMap/context-drift.md ProblemMap/entropy-collapse.md
|
||
jump: chunking global fix map ProblemMap/GlobalFixMap/Chunking/README.md
|
||
jump: ocr parsing global fix map ProblemMap/GlobalFixMap/OCR_Parsing/README.md
|
||
jump: semantic clinic index ProblemMap/SemanticClinicIndex.md
|
||
-->
|
||
|
||
## Orientation: quick routes
|
||
|
||
| Page | What it solves | Typical symptom |
|
||
|------|----------------|-----------------|
|
||
| [memory-coherence.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/memory-coherence.md) | Memory fences & continuity | Threads repeat or contradict |
|
||
| [entropy-collapse.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/entropy-collapse.md) | Attention melt in long windows | Outputs drift into filler |
|
||
| [context-drift.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/context-drift.md) | Long reasoning drift | Correct early, wrong later |
|
||
| [pattern_memory_desync.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/pattern_memory_desync.md) | Cross-tab & cache hazards | State flips after refresh |
|
||
| [ghost-context.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/ghost-context.md) | Stale buffers & residue | Model recalls “phantom” text |
|
||
| [state-fork.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/state-fork.md) | Divergent memory forks | Same task_id, different answers |
|
||
| [ocr-parsing-checklist.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/ocr-parsing-checklist.md), [ocr-jitter.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/ocr-jitter.md) | OCR-specific noise | Exported text drifts vs. original |
|
||
| [retrieval-traceability.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/retrieval-traceability.md), [data-contracts.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/data-contracts.md) | Traceability & audit | Citations break or vanish |
|
||
| [chunking-checklist.md](https://github.com/onestardao/WFGY/blob/main/ProblemMap/GlobalFixMap/MemoryLongContext/chunking-checklist.md) | Chunk stability at joins | Mid-sentence cuts or overlaps |
|
||
|
||
---
|
||
|
||
## Acceptance targets
|
||
- Retrieval coverage **≥ 0.70** to the intended section
|
||
- ΔS(question, retrieved) **≤ 0.45** and joins **≤ 0.50**
|
||
- λ_observe remains **convergent** across 3 paraphrases
|
||
- No state fork across tabs or agents for the same `task_id`
|
||
|
||
---
|
||
|
||
## 60-second fix checklist
|
||
|
||
1. **Lock metrics** — same analyzer & embeddings across sessions.
|
||
2. **Enforce memory fences** — require `task_id`, `session_id`, and `snippet_id`.
|
||
3. **Probe ΔS and λ** — run 3 paraphrases × 2 seeds.
|
||
4. **Patch drift** — realign chunks, re-check OCR, drop ghost spans.
|
||
5. **Audit consistency** — run traceability logs and verify continuity across restarts.
|
||
|
||
---
|
||
|
||
### 🔗 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 · 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 |
|
||
|
||
---
|
||
|
||
<!-- WFGY_FOOTER_START -->
|
||
|
||
### Explore More
|
||
|
||
| Layer | Page | What it’s for |
|
||
| --- | --- | --- |
|
||
| ⭐ Proof | [WFGY Recognition Map](/recognition/README.md) | External citations, integrations, and ecosystem proof |
|
||
| ⚙️ Engine | [WFGY 1.0](/legacy/README.md) | Original PDF tension engine and early logic sketch (legacy reference) |
|
||
| ⚙️ Engine | [WFGY 2.0](/core/README.md) | Production tension kernel for RAG and agent systems |
|
||
| ⚙️ Engine | [WFGY 3.0](/TensionUniverse/EventHorizon/README.md) | TXT based Singularity tension engine (131 S class set) |
|
||
| 🗺️ Map | [Problem Map 1.0](/ProblemMap/README.md) | Flagship 16 problem RAG failure taxonomy and fix map |
|
||
| 🗺️ Map | [Problem Map 2.0](/ProblemMap/wfgy-rag-16-problem-map-global-debug-card.md) | Global Debug Card for RAG and agent pipeline diagnosis |
|
||
| 🗺️ Map | [Problem Map 3.0](/ProblemMap/wfgy-ai-problem-map-troubleshooting-atlas.md) | Global AI troubleshooting atlas and failure pattern map |
|
||
| 🧰 App | [TXT OS](/OS/README.md) | .txt semantic OS with fast bootstrap |
|
||
| 🧰 App | [Blah Blah Blah](/OS/BlahBlahBlah/README.md) | Abstract and paradox Q&A built on TXT OS |
|
||
| 🧰 App | [Blur Blur Blur](/OS/BlurBlurBlur/README.md) | Text to image generation with semantic control |
|
||
| 🏡 Onboarding | [Starter Village](/StarterVillage/README.md) | Guided entry point for new users |
|
||
|
||
If this repository helped, starring it improves discovery so more builders can find the docs and tools.
|
||
[](https://github.com/onestardao/WFGY)
|
||
|
||
<!-- WFGY_FOOTER_END -->
|
||
|