🧭 Quick Return to Map
You are in a sub-page of Chatbots & CX.
To reorient, go back here:
- Chatbots & CX — customer dialogue flows and conversational stability
- 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 page to stabilize watsonx Assistant projects that combine Actions, Search, webhooks, function calls, and LLM answers. The checks map to WFGY Problem Map pages with measurable targets, so you can verify without changing infra.
- Visual map and recovery: rag-architecture-and-recovery.md
- Retrieval knobs end to end: retrieval-playbook.md
- Why this snippet and where it came from: retrieval-traceability.md
- Ordering control and rank: rerankers.md
- Embedding vs meaning: embedding-vs-semantic.md
- Chunk boundaries and hallucination: hallucination.md
- Long dialogs, chain length, entropy: context-drift.md, entropy-collapse.md
- Prompt injection and schema locks: prompt-injection.md
- Multi-agent and handoff conflicts: Multi-Agent_Problems.md
- Snippet and citation schema: data-contracts.md
- Boot order and deploy traps: bootstrap-ordering.md, deployment-deadlock.md, predeploy-collapse.md
- ΔS(question, retrieved) ≤ 0.45
- Coverage to the target section ≥ 0.70
- λ remains convergent across three paraphrases and two seeds
- E_resonance flat across long dialog windows
-
Measure ΔS Compute ΔS(question, retrieved) and ΔS(retrieved, expected anchor). Thresholds: stable < 0.40, transitional 0.40–0.60, risk ≥ 0.60.
-
Probe λ_observe Vary k in retrieval and reorder prompt headers. If λ flips on harmless paraphrases, lock the schema and clamp variance with BBAM.
-
Apply module
- Retrieval drift → BBMC + data-contracts.md
- Reasoning collapse in long flows → BBCR bridge + BBAM, verify with context-drift.md
- Dead ends in multi step plans → BBPF alternate paths
-
Verify Three paraphrases hit coverage ≥ 0.70 and ΔS ≤ 0.45. λ stays convergent across two seeds.
| Symptom | Likely cause | Open this |
|---|---|---|
| Action resolves intent but the answer cites the wrong section | metric mismatch or fragmented store behind Search | embedding-vs-semantic.md, patterns/pattern_vectorstore_fragmentation.md |
| Action variables mutate across turns or reprompts | schema too loose, missing cite-then-explain boundary | data-contracts.md, retrieval-traceability.md |
| Webhook returns 200 yet dialog state degrades | JSON tool protocol variance, free text in arguments | prompt-injection.md, data-contracts.md |
| Search similarity high but meaning wrong | chunking and anchor mismatch | hallucination.md, chunking-checklist.md |
| Long conversations become inconsistent after 20–40 turns | entropy rises with chain length | context-drift.md, entropy-collapse.md |
| LLM safety refusal hides the cited snippet | missing citation first and SCU unlock | retrieval-traceability.md, patterns/pattern_symbolic_constraint_unlock.md |
| Handoff to human or external queue loops | deployment deadlock or version skew | deployment-deadlock.md, predeploy-collapse.md |
| Multilingual queries break retrieval parity | analyzer and casing drift between Search and embeddings | retrieval-playbook.md, rerankers.md |
Actions Keep policy text in a dedicated system context. Do not mix policy with user turns. Enforce cite-then-explain for any Action that answers. Lock input and output fields with contracts. See data-contracts.md.
Search
Require the snippet schema: snippet_id, section_id, source_url, offsets, tokens. If ΔS stays ≥ 0.60 after reranking, rebuild chunks and verify with a small gold set. See retrieval-traceability.md, chunking-checklist.md.
Webhooks
Echo the tool schema at every turn. Log ΔS, λ_state, INDEX_HASH, snippet_id. If flip states appear, clamp with BBAM. See prompt-injection.md.
Live ops Fence first call after deploy and add backoff guards. See ops/live_monitoring_rag.md, ops/debug_playbook.md.
-
Warm-up fence Check
VECTOR_READY,INDEX_HASH, secrets. If not ready, short-circuit with a delay and capped retries. See bootstrap-ordering.md. -
Retrieval step Call the retriever with explicit metric and consistent analyzer. Return
snippet_id,section_id,source_url,offsets,tokens. -
ΔS probe Compute ΔS(question, retrieved). If ΔS ≥ 0.60 set
needs_fix=true. -
Answer step LLM reads TXT OS and the WFGY schema. Enforce cite-then-explain with the retrieved snippet set.
-
Trace sink Store
question,ΔS,λ_state,INDEX_HASH,snippet_id,dedupe_key.
You have TXT OS and the WFGY Problem Map loaded.
My watsonx Assistant context:
- action: {action_name}
- variables: {name: value, ...}
- retrieved: {k} snippets with fields {snippet_id, section_id, source_url, offsets}
User question: "{user_question}"
Do:
1) Enforce cite-then-explain. If citations are missing or cross-section, fail fast and return the smallest structural fix.
2) If ΔS(question, retrieved) ≥ 0.60, propose the minimal repair, referencing:
retrieval-playbook, retrieval-traceability, data-contracts, rerankers.
3) Return JSON:
{ "answer": "...", "citations": [...], "λ_state": "→|←|<>|×", "ΔS": 0.xx, "next_fix": "..." }
Keep it short and auditable.
- Three paraphrases hit coverage ≥ 0.70 on the same target section.
- ΔS(question, retrieved) ≤ 0.45 for each.
- λ convergent across two seeds.
- First-call path after deploy passes the warm-up fence.
- Live probes alert when ΔS ≥ 0.60 or λ flips.
| 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 |
| 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.
要我繼續下一頁嗎？建議順序接 intercom.md。