8.2 KiB
Chatbots & CX — Global Fix Map
🏥 Quick Return to Emergency Room
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
- WFGY Problem Map 1.0 — 16 reproducible failure modes
Think of this page as a sub-room.
If you want full consultation and prescriptions, go back to the Emergency Room lobby.
Chatbot and CX bugs are structural failures in dialog systems, where intent routing, slot/entity state, conversation memory, connector metadata, or knowledge retrieval breaks even when the underlying LLM is fine.
Most incidents come from environment drift, schema mismatches, cold-start latency, connector desync, and weak policy fences. This folder maps symptoms to vendor pages and WFGY structural fixes with measurable acceptance targets.
When to use this folder
- Bot flows break across environments (dev vs prod vs staging).
- Slot filling, entities, or context reset unexpectedly.
- Latency or cold starts cause dropped conversations.
- Omnichannel connectors desync with the core bot.
- Prompts bypass policies or hallucinate unsafe outputs.
- Regression after migration to a new vendor framework.
Acceptance targets
- Intent recognition F1 ≥ 0.85 across test set.
- ΔS(user_query, retrieved) ≤ 0.45 for all routes.
- Coverage of knowledge base ≥ 0.70 after repair.
- λ remains convergent across 3 paraphrases and 2 seeds.
- p95 latency ≤ 800 ms across channels, warm path.
- Zero PII leakage in logs or vector payloads.
Quick routes — per chatbot vendor
| Vendor / Platform | Fix Page |
|---|---|
| Amazon Lex | amazon_lex.md |
| Azure Bot Service | azure_bot_service.md |
| Dialogflow CX | dialogflow_cx.md |
| Freshchat | freshchat.md |
| Freshdesk | freshdesk.md |
| Intercom | intercom.md |
| Microsoft Copilot Studio | microsoft_copilot_studio.md |
| Rasa | rasa.md |
| Salesforce Einstein Bots | salesforce_einstein_bots.md |
| Twilio Studio | twilio_studio.md |
| Watsonx Assistant | watsonx_assistant.md |
| Zendesk | zendesk.md |
Symptom → exact fix
| Symptom | Likely cause | Open this |
|---|---|---|
| Slot filling fails randomly | Missing entity fallback, context reset | dialogflow_cx.md · rasa.md |
| Bot replies too slowly | Cold starts, webhook bottlenecks | amazon_lex.md · azure_bot_service.md |
| Knowledge base answers drift | Retrieval misaligned with store | watsonx_assistant.md · retrieval-traceability.md |
| Omnichannel flow desync | Connectors drop metadata | intercom.md · zendesk.md |
| Unsafe or off-policy replies | No policy fences or prompt injection | microsoft_copilot_studio.md · prompt-injection.md |
| Migration broke production | Schema mismatch, missing idempotency | serverless_ci_cd.md · env_bootstrap_and_migrations.md |
| Conversation history disappears | Stateless KV pattern missing | stateless_kv_queue_patterns.md |
Fix in 60 seconds
- Verify intents — run eval with test utterances, check F1 ≥ 0.85.
- Check state — ensure KV queue or context store persists between turns.
- Fence prompts — load prompt_injection.md.
- Measure latency — split warm vs cold path; provision concurrency if needed.
- Cross-channel sync — confirm connectors carry metadata (role, region, tenant).
Copy-paste prompt for chatbot incidents
You have TXT OS and the WFGY Problem Map loaded.
My chatbot incident:
- platform: [Dialogflow|Rasa|Intercom|etc.]
- symptom: [short description]
- eval: { F1, ΔS, coverage, λ states }
- infra: { cold_ms, warm_ms, concurrency, connectors }
- compliance: { pii_found: true|false, policy_eval }
Tell me:
1) which layer is failing and why,
2) the exact WFGY page to open,
3) the minimal steps to restore accuracy and latency,
4) a quick regression test to prevent repeat.
FAQ
Q: My bot answers correctly in dev but fails in production. Why? A: Likely env mismatch — analyzers, slots, or schema drift between environments. Check serverless_ci_cd.md.
Q: How do I stop hallucinations in Copilot Studio or Dialogflow? A: Enforce policy fences and cite-then-explain. See prompt_injection.md.
Q: Why is latency so different on first vs second request? A: Cold starts. See cold_start_concurrency.md.
Q: How can I ensure PII never leaks through connectors? A: Attach privacy_and_pii_edges.md and enforce payload contracts.
Q: Do I need to change my vendor to apply WFGY fixes? A: No. WFGY guardrails are store-agnostic and vendor-agnostic. You patch structure, not infra.
🔗 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 | Canonical framework entry point | View |
| Problem Map | Diagnostic map and navigation hub | View |
| Tension Universe Experiments | MVP experiment field | View |
| Recognition | Where WFGY is referenced or adopted | View |
| AI Guide | Anti-hallucination reading protocol for tools | View |
If this repository helps, starring it improves discovery for other builders.
