# Wetech regional SaaS default with customer confirmation - Owner: Scenario Router target selection and canonical customer confirmation. - Admission integration: LLM and deterministic classification recognize only the visible Wetech OP 01/02 shorthand as a supplied cluster reference. Coverage normalization clears only that mistaken missing alias, without changing Gate allowlists, assigning a canonical cluster, or removing other missing/safety boundaries. Router remains the mapping and customer-confirmation owner. - Source: explicit user instruction on 2026-09-05 following APID-29251's `diagnostic_wetech_target_missing` replay. Existing title, comments and both screenshots were present; this is not a historical-input repair. - Implementation: source PR #1344 merged; staging release run 33954784688 succeeded and app-delivery PRs #79322/#79323 merged. Before replay, live image SHA `4d005fb6230b1156d97a6565840ed6d44ff219cc` was verified with a Ready Pod, zero restarts and HTTP 200 root/health/readiness endpoints. - Local validation: TypeScript compilation and all 1092 tests passed after admission integration; 173 focused classifier/Gate/Router tests passed. The new topic is recalled for both 01/02 forms, Router receives its mapping and owns confirmation, and Evidence/Final Response do not receive default-selection authority. Updated the live-workspace topic inventory expectation. - Scope: `亚太01` / `亚太 01` means `asia-prod/SaaS01`, and the corresponding 02 forms mean `asia-prod/SaaS02`, only under the Wetech OP convention. - Preservation: explicit business/full names win; absent inventory, regional conflicts, other customers, Cloud, bare regions and other numbers do not receive a default. No Provider, Final Response or Judge ownership expansion. - Post-deployment verification on 2026-09-05: one Agent9 `mode=read` historical replay of APID-29251, run `755d9a9a-8c0c-492b-b738-665358dd05a6`, returned HTTP 200 in 509358 ms. The classifier preserved `cluster_alias` with `missingKeys=[]`; Router selected `asia-prod/SaaS02`; Evidence and Scenario Diagnostic executed before customer confirmation. Live classifier smoke also admitted both 01 and 02 shorthand forms without a missing cluster key. - The final customer draft says “本次暂按亚太 SaaS02 集群进行分析” and asks “请您确认：‘亚太02’是否指亚太 SaaS02 集群？” The result is `human_required`, with clarification required and only that cluster-intent question in `requestedContext`. This verifies mapping plus confirmation, not diagnostic causality or successful data acquisition. - Separate remaining blocker: primary CPU query attempt 1 failed validation (`invalid unit`); its corrected attempt 2 failed with `pzero_application_session_not_ready`. The handoff contains no observations; Scenario Diagnostic returned `inconclusive` with unresolved workload attribution. The final answer does not confirm ANALYZE as the root cause. This is a Provider-session failure, not a returned empty historical dataset and not renewed cluster-admission rejection. No Provider fix or second replay was included in this task. - Evaluation: `no_human_reply` / `not_compared`; the selected human comment arranges case merging rather than supplying an independent diagnosis. This result is not evidence of an AI-versus-human quality improvement. - Frozen answer fingerprint: `23088945ffe4b5105fe070dcd46e0efe587d5a4e8652eeae6e0b606289e9002e`. The API response and run evidence archive are retained locally under `~/.cache/aos-model-comparisons/20260905-wetech-shorthand-deployed-replay/`. Publication reports database/Jira/evaluation/Feishu all `not_requested`. - Status remains `optimized_unverified`: the bounded alias/confirmation behavior passed this read replay, but usable acquisition and the applicable Jira internal-comment refresh have not been verified. Next action requires a separately authorized Provider-session investigation and subsequent bounded replay/publication; do not rerun or publish implicitly.