TextGen WebUI: Guardrails and Fix Patterns

🧭 Quick Return to Map

You are in a sub-page of LocalDeploy_Inference.
To reorient, go back here:

LocalDeploy_Inference — on-prem deployment and model inference

WFGY Global Fix Map — main Emergency Room, 300+ structured fixes

WFGY Problem Map 1.0 — 16 reproducible failure modes

Think of this page as a desk within a ward.
If you need the full triage and all prescriptions, return to the Emergency Room lobby.

A stabilization guide for oobabooga/text-generation-webui (TextGen WebUI). This tool is popular for local or Docker-based inference but often drifts under plugins, extensions, or mixed model backends. Use this page to detect and repair instability without changing infra.

Open these first

Architecture & recovery: RAG Architecture & Recovery
Retrieval knobs: Retrieval Playbook
Trace schema: retrieval-traceability.md
Ordering & warm-up fixes: bootstrap-ordering.md, deployment-deadlock.md, predeploy-collapse.md
Embedding check: embedding-vs-semantic.md

Core acceptance

ΔS(question, retrieved) ≤ 0.45
Coverage ≥ 0.70 to target snippet
λ remains convergent across 3 paraphrases and 2 seeds
Stable across UI plugins (no flip-flops when toggling extensions)

Typical WebUI breakpoints and fix

Symptom	Likely cause	Fix
Runs fine in notebook, but WebUI returns divergent outputs	Extension modifies prompt schema	data-contracts.md, retrieval-traceability.md
Citations missing when using API mode	WebUI truncates field names	retrieval-traceability.md
ΔS > 0.60 only when multiple users connect	Concurrency split in worker threads	context-drift.md, entropy-collapse.md
JSON schema output fails with extensions	UI prompt injection / malformed tools	prompt-injection.md, logic-collapse.md
First run after model load unstable	Race in init hooks	bootstrap-ordering.md

Fix in 60 seconds

Warm-up: Always run a dummy inference after model load. Log INDEX_HASH.
Schema lock: Enforce snippet_id, tokens, offsets in every response.
λ probe: Run 3 paraphrases, 2 seeds. Clamp with BBAM if divergent.
Trace logging: Save ΔS and λ per extension toggle.
Verify: Replay dataset with and without extensions enabled. ΔS ≤ 0.45 required.

Copy-paste diagnostic prompt

I am running text-generation-webui with {extensions_list}.  
Question: "{user_question}"  

Please return:
- ΔS vs retrieved snippet  
- λ across paraphrases  
- Citations preserved? (yes/no)  
- If ΔS ≥ 0.60, point to minimal WFGY module to apply

🔗 Quick-Start Downloads (60 sec)

Tool	Link	3-Step Setup
WFGY 1.0 PDF	Engine Paper	1️⃣ Download · 2️⃣ Upload to your LLM · 3️⃣ Ask “Answer using WFGY + <your question>”
TXT OS (plain-text 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 →
Problem Map 1.0	Initial 16-mode diagnostic and symbolic fix framework	View →
Problem Map 2.0	RAG-focused failure tree, modular fixes, and pipelines	View →
Semantic Clinic Index	Expanded failure catalog: prompt injection, memory bugs, logic drift	View →
Semantic Blueprint	Layer-based symbolic reasoning & semantic modulations	View →
Benchmark vs GPT-5	Stress test GPT-5 with full WFGY reasoning suite	View →
🧙‍♂️ Starter Village 🏡	New here? Lost in symbols? Click here and let the wizard guide you through	Start →

👑 Early Stargazers: See the Hall of Fame ⭐ WFGY Engine 2.0 is already unlocked. ⭐ Star the repo to help others discover it and unlock more on the Unlock Board.

9.3 KiB Raw Blame History Unescape Escape