10 KiB
Google AI (Gemini): Guardrails and Fix Patterns
🧭 Quick Return to Map
You are in a sub-page of LLM_Providers.
To reorient, go back here:
- LLM_Providers — model vendors and deployment options
- 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 compact field guide to stabilize Gemini calls inside RAG, agents, or long workflows. Use the checks below to localize failure, then jump to the exact WFGY fix page.
Core acceptance
- ΔS(question, retrieved) ≤ 0.45
- Coverage ≥ 0.70 for the target section
- λ remains convergent across 3 paraphrases
- E_resonance stable on long windows
Open these first
- Visual map and recovery: RAG Architecture & Recovery
- End-to-end retrieval knobs: Retrieval Playbook
- Why this snippet (traceability schema): Retrieval Traceability
- Ordering control: Rerankers
- Embedding vs meaning: Embedding ≠ Semantic
- Hallucination and chunk boundaries: Hallucination
- Long chains and entropy: Context Drift, Entropy Collapse
- Structural collapse and recovery: Logic Collapse
- Snippet and citation schema: Data Contracts
Typical breakpoints and the right fix
| Symptom you see | Likely cause | Fix page |
|---|---|---|
| High similarity yet wrong meaning | Metric or index mismatch | Embedding ≠ Semantic |
| Gemini cites the wrong paragraph | Chunk boundaries and trace loss | Hallucination, Retrieval Traceability |
| Answers flip across runs | λ instability on long threads | Context Drift, Entropy Collapse |
| Refuses or loops on safe content | Prompt contract not locked | Data Contracts |
| Good recall but bad ordering | Reranking missing | Rerankers |
| Corrected errors reappear | Re-entry without variance clamp | pattern_hallucination_reentry.md |
Fix in 60 seconds
- Measure ΔS
- Compute ΔS(question, retrieved) and ΔS(retrieved, expected anchor).
- Thresholds: stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60.
- Probe with λ_observe
- Vary k ∈ {5, 10, 20}. Flat high curve means metric or index mismatch.
- Reorder prompt headers. If ΔS spikes, lock the schema.
- Apply the module
- Retrieval drift → BBMC + Data Contracts.
- Reasoning collapse → BBCR bridge + BBAM variance clamp.
- Dead ends in long runs → BBPF alternate path with explicit step limits.
- Verify
- Coverage ≥ 0.70 on the target section.
- Three paraphrases keep ΔS ≤ 0.45 and λ convergent.
- Re-run with seed change and shuffled snippet order.
Gemini-specific gotchas
-
Tool and JSON calls
If the function schema is loose, Gemini may hallucinate fields. Lock schemas with Data Contracts and clamp variance with BBAM. -
Safety flips on neutral text
When the role block is not pinned, safety can overfire. Use a citation-first header from Retrieval Traceability and keep source boundaries explicit. -
Hybrid retrieval regressions
HyDE plus keyword can split queries. Check pattern_query_parsing_split.md and add a stable anchor paragraph to reduce drift. -
Long context smear
Large windows flatten meaning if chunks are not semantic. Rebuild with the chunking checklist and verify joins with ΔS probes.
Copy-paste prompt (safe)
read the WFGY TXT OS and Problem Map pages. extract ΔS, λ\_observe, E\_resonance and modules BBMC, BBPF, BBCR, BBAM.
given my gemini failure:
* symptom: \[brief]
* traces: \[ΔS(question, retrieved)=…, ΔS(retrieved, anchor)=…, λ states]
tell me:
1. which layer fails and why,
2. which fix page to open from this repo,
3. the minimal steps to push ΔS ≤ 0.45 and keep λ convergent,
4. how to verify with a reproducible test.
Escalation path
- Index or metric mismatch that stays flat after k sweeps → rebuild embeddings and check Vectorstore Fragmentation.
- Agent tools fighting each other → see Multi-Agent Problems and split memory namespaces.
- First prod call fails after deploy → check Pre-Deploy Collapse and Bootstrap Ordering.
🔗 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 —
Engineers, hackers, and open source builders who supported WFGY from day one.
⭐ WFGY Engine 2.0 is already unlocked. ⭐ Star the repo to help others discover it and unlock more on the Unlock Board.
