10 KiB
AWS CodeWhisperer: Guardrails and Fix Patterns
🧭 Quick Return to Map
You are in a sub-page of DevTools_CodeAI.
To reorient, go back here:
- DevTools_CodeAI — AI-assisted coding and developer productivity
- 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.
Use this guide when completions or chat inside CodeWhisperer feel flaky, tool steps loop, or RAG-style answers cite the wrong things. The fixes below map to WFGY pages with measurable targets so you can verify quickly and avoid infra changes.
Open these first
- Visual map and recovery: RAG Architecture & Recovery
- End-to-end retrieval knobs: Retrieval Playbook
- Why this snippet, not another: Retrieval Traceability
- Ordering control for top-k: Rerankers
- Embedding vs meaning drift: Embedding ≠ Semantic
- Long dialogs and chain fatigue: Context Drift, Entropy Collapse
- Prompt injection and schema locks: Prompt Injection
- Multi-agent conflicts and handoffs: Multi-Agent Problems
- Cold boot and deploy ordering: Bootstrap Ordering, Deployment Deadlock, Pre-Deploy Collapse
- Snippet and citation schema: Data Contracts
Core acceptance
- ΔS(question, retrieved) ≤ 0.45
- Coverage ≥ 0.70 to the correct section
- λ remains convergent across three paraphrases and two seeds
- E_resonance flat across the dialog window
Typical CodeWhisperer breakpoints → exact fix
-
Region or account skew between your IDE plugin, credentials, and model endpoint. Verify region and identity consistently. If first call in a fresh boot fails, fix ordering. Open: Bootstrap Ordering, Pre-Deploy Collapse
-
IDE chat cites the wrong file or wrong snippet after retrieval. Lock the snippet contract and require cite-then-explain. Open: Retrieval Traceability, Data Contracts
-
High similarity yet wrong answer when CodeWhisperer consults docs. Suspect metric or index mismatch, or fragmented store. Open: Embedding ≠ Semantic, Vectorstore Fragmentation
-
Hybrid retrieval gets worse than single retriever in chat plans. Stabilize query split and lock reranking deterministically. Open: Query Parsing Split, Rerankers
-
Tool loop or agent handoff stalls when chat triggers build, test, or docs tools. Split memory namespaces, apply timeouts, and fence writes by
mem_revandmem_hash. Open: Multi-Agent Problems -
Security or policy blocks cause silent fallbacks that change outputs. Make refusal paths explicit and keep the schema locked to avoid hidden branches. Open: Prompt Injection
Fix in 60 seconds
-
Measure ΔS Compute ΔS(question, retrieved) and ΔS(retrieved, anchor section). Stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60.
-
Probe λ_observe Re-order headers minimally and vary k as 5, 10, 20. If ΔS stays flat and high, rebuild metric and normalize. If λ flips on harmless paraphrase, clamp with BBAM.
-
Apply the module Retrieval drift → BBMC + Data Contracts Reasoning collapse → BBCR bridge + BBAM, then verify with Logic Collapse Dead ends in long chains → BBPF alternate paths
-
Verify Coverage ≥ 0.70 on three paraphrases. λ convergent on two seeds. E_resonance flat over ten-step dialogs.
IDE checklist for stable runs
-
Warm-up fence before chat or retrieval. Confirm
INDEX_HASH,VECTOR_READY, and current credentials. See: Bootstrap Ordering -
Idempotency for any write step triggered by chat tools. Compute
dedupe_key = sha256(source_id + revision + index_hash)and drop duplicates. -
Cite-then-explain as a hard rule in the prompt template. Forbid cross-section reuse unless explicitly allowed by contract.
-
Observability probes inside the IDE task. Log ΔS and λ states for retrieve, assemble, reason. Alert when ΔS ≥ 0.60 or λ turns divergent.
-
Regression gate before you trust the session. See: RAG Precision/Recall
Copy-paste prompt for CodeWhisperer Chat
You have TXTOS and the WFGY Problem Map loaded.
My task:
- symptom: [one line]
- traces: ΔS(question,retrieved)=..., ΔS(retrieved,anchor)=..., λ states across 3 paraphrases
Do:
1) identify which layer fails and why,
2) point me to the exact WFGY page,
3) give minimal steps to push ΔS ≤ 0.45 and keep λ convergent,
4) return a short JSON plan with {citations, steps, ΔS, λ_state, next_fix}.
Use BBMC, BBPF, BBCR, BBAM when relevant. Enforce cite-then-explain.
When to escalate
-
ΔS stays ≥ 0.60 after chunking and metric fixes Rebuild with the semantic chunking checklist and verify on a small gold set. Open: Chunking Checklist
-
Answers flip between identical runs in the same IDE session Investigate memory and version skew. Open: Pre-Deploy Collapse
🔗 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
| Layer | Page | What it’s for |
|---|---|---|
| ⭐ Proof | WFGY Recognition Map | External citations, integrations, and ecosystem proof |
| ⚙️ Engine | WFGY 1.0 | Original PDF tension engine and early logic sketch (legacy reference) |
| ⚙️ Engine | WFGY 2.0 | Production tension kernel for RAG and agent systems |
| ⚙️ Engine | WFGY 3.0 | TXT based Singularity tension engine (131 S class set) |
| 🗺️ Map | Problem Map 1.0 | Flagship 16 problem RAG failure taxonomy and fix map |
| 🗺️ Map | Problem Map 2.0 | Global Debug Card for RAG and agent pipeline diagnosis |
| 🗺️ Map | Problem Map 3.0 | Global AI troubleshooting atlas and failure pattern map |
| 🧰 App | TXT OS | .txt semantic OS with fast bootstrap |
| 🧰 App | Blah Blah Blah | Abstract and paradox Q&A built on TXT OS |
| 🧰 App | Blur Blur Blur | Text to image generation with semantic control |
| 🏡 Onboarding | Starter Village | Guided entry point for new users |
If this repository helped, starring it improves discovery so more builders can find the docs and tools.
