docs(ms-ai-architect): Spor 1 — authority-URL valgt per reference-fil (manifest source-satt) [skip-docs]

docs/spor1-authority-selection.md

@@ -0,0 +1,202 @@
+# Spor 1 — Authority-URL-valg per reference-fil (Step 6 audit-logg)
+
+**Type:** methodology
+**Status:** Stable
+**Kategori:** Spor 1 korpus-migrering
+**Sist oppdatert:** 2026-07-04
+**Produsert av:** trekexecute Session 2, Step 6 (to-lags hovedkontekst — operatørvalg 2026-07-04)
+
+---
+
+Denne loggen dokumenterer valget av **THE authority `learn.microsoft.com`-URL** per non-advisor `reference`-fil, persistert som `source` i `scripts/kb-update/data/ref-type-manifest.json` (gitignored/regenererbar; den varige posten er denne loggen + `**Source:**`-headerne Step 9 stempler). Metode: **deterministiske tie-breaks først** (single citation → concept/overview over how-to → mest-sitert i korpus), deretter **hovedkontekst-vurdering** for gjenstående likestilte kandidater. Ingen gjetning: en fil uten forsvarlig MS-autoritet settes `deferred:true` (defer-not-guess, jf. `authority.mjs`).
+
+## Sammendrag
+
+| Bøtte | Antall | Metode |
+|---|---:|---|
+| Sourced (non-advisor reference) | 243 | header 6 · concept 64 · most-cited 141 · top-score 1 · residual-judgment 31 |
+| Deferred — ingen MS-sitering | 59 | `no-ms-citation` (typet reference på innhold i Step 4, men siterer 0 `learn.microsoft.com`-URL) |
+| Deferred — advisor-scope | 50 | `advisor-scope` (utenfor Step 6 sitt non-advisor-mandat; aldri mutert av applier; S-Cosmo) |
+| **Sum reference-entries** | **352** | |
+
+### Premiss-korreksjoner mot plan-estimatene (verifisert mot manifest, 2026-07-04)
+
+- Planen anslo «~253 reference / ~246 uten source». Ground truth: **302 non-advisor reference** (+ 50 advisor reference = 352 totalt).
+- Planen sa «7 filer har allerede `**Source:**`». Ground truth: **6** (header-metoden under).
+- **Advisor-håndtering (scope-avgjørelse):** Step 6-teksten scoper til «non-advisor», men Verify-predikatet skanner ALLE reference-entries. De 50 advisor reference-entriene settes derfor `deferred:true` (`advisor-scope`) — ikke gjettede kilder — så predikatet holder uten å utvide arbeidet inn på advisor-flaten (som uansett aldri muteres og er merket for Cosmo-fjerning).
+
+## Hovedkontekst-adjudikasjon — 31 residual-likestillinger
+
+Disse hadde flere kandidater likestilt på både score og korpus-sitering. Valgt URL er den primær-tema-kanoniske autoriteten, plukket FRA filens egen siteringsmengde (skriptet asserterer medlemskap).
+
+| Fil | Valgt authority | Begrunnelse |
+|---|---|---|
+| ms-ai-engineering / agent-orchestration/agent-365-governance-and-deployment.md | https://learn.microsoft.com/copilot/microsoft-365/agent-essentials/m365-agents-blueprint | M365 agents blueprint governs deployment |
+| ms-ai-engineering / agent-orchestration/agent-memory-and-context-management.md | https://learn.microsoft.com/azure/foundry/agents/concepts/what-is-memory | canonical "what is memory" for Foundry agents |
+| ms-ai-engineering / agent-orchestration/semantic-kernel-agents-implementation.md | https://learn.microsoft.com/agent-framework/overview/agent-framework-overview | Agent Framework overview = authoritative agent-implementation entry |
+| ms-ai-engineering / azure-ai-services/ai-services-api-best-practices.md | https://learn.microsoft.com/azure/architecture/patterns/rate-limiting-pattern | rate-limiting architecture pattern anchors API resilience best-practices |
+| ms-ai-engineering / azure-ai-services/document-intelligence-custom-models.md | https://learn.microsoft.com/azure/ai-services/document-intelligence/train/custom-model | parent custom-model doc (sub-types derive from it) |
+| ms-ai-engineering / azure-ai-services/language-services-question-answering.md | https://learn.microsoft.com/azure/ai-services/language-service/question-answering/overview | question-answering overview is the canonical entry |
+| ms-ai-engineering / azure-ai-services/translator-custom-neural-models.md | https://learn.microsoft.com/azure/ai-services/translator/custom-translator/overview | custom-translator overview is the canonical entry |
+| ms-ai-engineering / azure-ai-services/translator-document-translation.md | https://learn.microsoft.com/azure/ai-services/translator/document-translation/overview | document-translation overview (on-topic) over the generic translator overview |
+| ms-ai-engineering / data-engineering/cross-cloud-data-integration.md | https://learn.microsoft.com/fabric/governance/external-data-sharing-overview | external data sharing = cross-cloud integration authority |
+| ms-ai-engineering / data-engineering/data-sampling-labeling.md | https://learn.microsoft.com/azure/machine-learning/how-to-label-data | canonical data-labeling doc (all-howto pool) |
+| ms-ai-engineering / data-engineering/data-versioning-lineage.md | https://learn.microsoft.com/fabric/governance/lineage | Fabric governance lineage = canonical lineage authority |
+| ms-ai-engineering / data-engineering/microsoft-purview-governance.md | https://learn.microsoft.com/purview/unified-catalog | Purview Unified Catalog is the governance hub |
+| ms-ai-engineering / data-engineering/real-time-streaming-ai.md | https://learn.microsoft.com/fabric/real-time-intelligence/overview | broadest Real-Time Intelligence overview |
+| ms-ai-engineering / data-engineering/zero-etl-fabric-patterns.md | https://learn.microsoft.com/fabric/mirroring/overview | Fabric mirroring = zero-ETL authority |
+| ms-ai-engineering / mlops-genaiops/mlops-security-access-control.md | https://learn.microsoft.com/azure/machine-learning/concept-enterprise-security | ML enterprise-security concept anchors access-control |
+| ms-ai-engineering / rag-architecture/contextual-retrieval.md | https://learn.microsoft.com/azure/search/cognitive-search-custom-skill-interface | current custom-skill interface over the previous-versions variant |
+| ms-ai-engineering / rag-architecture/metadata-management-filtering.md | https://learn.microsoft.com/azure/search/search-what-is-an-index | index schema is the root authority for filterable metadata |
+| ms-ai-governance / monitoring-observability/alerting-strategies-escalation.md | https://learn.microsoft.com/azure/azure-monitor/alerts/best-practices-alerts | alerts best-practices = alerting-strategy authority |
+| ms-ai-governance / monitoring-observability/distributed-tracing-ai-pipelines.md | https://learn.microsoft.com/azure/azure-monitor/app/opentelemetry-overview | OpenTelemetry overview = distributed-tracing authority |
+| ms-ai-governance / monitoring-observability/observability-for-copilot-extensions.md | https://learn.microsoft.com/microsoft-cloud/dev/copilot/isv/observability-for-ai | directly "observability for AI" (Copilot ISV) |
+| ms-ai-governance / norwegian-public-sector-governance/accessibility-requirements-wcag-norway.md | https://learn.microsoft.com/compliance/regulatory/offering-wcag-2-1 | WCAG 2.1 compliance offering = accessibility authority |
+| ms-ai-governance / norwegian-public-sector-governance/digdir-ai-governance-structure.md | https://learn.microsoft.com/azure/cloud-adoption-framework/scenarios/ai/govern | CAF AI Govern = canonical AI-governance authority |
+| ms-ai-governance / norwegian-public-sector-governance/digital-accessibility-action-plan.md | https://learn.microsoft.com/compliance/regulatory/offering-wcag-2-1 | WCAG 2.1 compliance offering = accessibility standards authority |
+| ms-ai-governance / norwegian-public-sector-governance/ros-analyse-ai-systems.md | https://learn.microsoft.com/azure/well-architected/security/threat-model | Well-Architected threat-model = risk-analysis authority |
+| ms-ai-infrastructure / bcdr/compliance-requirements-bcdr.md | https://learn.microsoft.com/azure/compliance | Azure compliance hub = compliance-requirements authority |
+| ms-ai-infrastructure / bcdr/data-replication-patterns-ai.md | https://learn.microsoft.com/azure/reliability/concept-redundancy-replication-backup | reliability redundancy/replication concept = replication-patterns authority |
+| ms-ai-infrastructure / bcdr/network-resilience-patterns-ai.md | https://learn.microsoft.com/azure/ddos-protection/ddos-protection-overview | DDoS protection = primary network-resilience authority in pool |
+| ms-ai-security / ai-security-engineering/entra-agent-id-zero-trust.md | https://learn.microsoft.com/entra/id-governance/agent-id-governance-overview | Entra Agent ID governance overview (directly on-topic) |
+| ms-ai-security / ai-security-engineering/zero-trust-ai-services.md | https://learn.microsoft.com/security/zero-trust/apply-zero-trust-azure-services-overview | apply Zero Trust to Azure services overview = canonical |
+| ms-ai-security / cost-optimization/licensing-compliance-cost-avoidance.md | https://learn.microsoft.com/azure/cloud-adoption-framework/scenarios/ai/plan | CAF AI Plan = broad licensing/cost authority |
+| ms-ai-security / performance-scalability/rate-limit-management.md | https://learn.microsoft.com/azure/foundry/openai/quotas-limits | Azure OpenAI quotas-limits = rate-limit authority |
+
+## Deterministisk-løste (212)
+
+**Header (beholdt eksisterende `**Source:**`, 6):**
+
+- ms-ai-engineering / rag-architecture/late-chunking-patterns.md → https://learn.microsoft.com/azure/foundry/openai/tutorials/embeddings
+- ms-ai-security / ai-security-engineering/ai-red-team-operations-practical.md → https://learn.microsoft.com/security/ai-red-team/training
+- ms-ai-security / cost-optimization/batch-processing-cost-reduction.md → https://learn.microsoft.com/azure/foundry/openai/how-to/batch
+- ms-ai-security / cost-optimization/model-selection-price-performance.md → https://learn.microsoft.com/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure
+- ms-ai-security / cost-optimization/ptu-vs-paygo-economics.md → https://learn.microsoft.com/azure/foundry/openai/concepts/provisioned-throughput
+- ms-ai-security / performance-scalability/batch-api-usage-optimization.md → https://learn.microsoft.com/azure/foundry/openai/how-to/batch-blob-storage
+
+**Deterministiske tie-breaks (206):** single/concept/top-score/most-cited — full liste i manifestets `authorityMethod`-felt per entry. Metode-fordeling over.
+
+## Deferred — ingen MS-sitering (59, `unsourced`-bøtte)
+
+Reference-typet på innhold (Step 4), men uten inline `learn.microsoft.com`-sitering å binde autoritet til. Ingen gjetning — corpus judge-pass (neste brief) henter + verifiserer + kilde-setter disse.
+
+- ms-ai-engineering / agent-orchestration/agent-compliance-and-audit-trails.md
+- ms-ai-engineering / agent-orchestration/agent-cost-optimization-strategies.md
+- ms-ai-engineering / agent-orchestration/agent-ecosystem-and-plugin-marketplace.md
+- ms-ai-engineering / agent-orchestration/agent-feedback-and-learning-loops.md
+- ms-ai-engineering / agent-orchestration/agent-latency-optimization.md
+- ms-ai-engineering / agent-orchestration/agent-monitoring-observability.md
+- ms-ai-engineering / agent-orchestration/agent-routing-and-specialization.md
+- ms-ai-engineering / agent-orchestration/agent-security-threat-modeling.md
+- ms-ai-engineering / agent-orchestration/copilot-agent-integration-patterns.md
+- ms-ai-engineering / agent-orchestration/declarative-vs-imperative-agent-design.md
+- ms-ai-engineering / agent-orchestration/multi-tenant-agent-isolation.md
+- ms-ai-engineering / api-management/circuit-breaker-ai-resilience.md
+- ms-ai-engineering / api-management/load-balancing-openai-instances.md
+- ms-ai-engineering / api-management/semantic-caching-apim.md
+- ms-ai-engineering / api-management/token-rate-limiting-policies.md
+- ms-ai-engineering / multi-modal/accessibility-multimodal-ai.md
+- ms-ai-engineering / multi-modal/audio-video-transcription-workflow.md
+- ms-ai-engineering / multi-modal/azure-video-indexer-patterns.md
+- ms-ai-engineering / multi-modal/cv-llm-integration.md
+- ms-ai-engineering / multi-modal/dalle-image-generation.md
+- ms-ai-engineering / multi-modal/document-vision-processing.md
+- ms-ai-engineering / multi-modal/gpt4o-vision-architecture.md
+- ms-ai-engineering / multi-modal/image-classification-understanding.md
+- ms-ai-engineering / multi-modal/multimodal-content-safety.md
+- ms-ai-engineering / multi-modal/multimodal-evaluation-metrics.md
+- ms-ai-engineering / multi-modal/multimodal-prompt-engineering.md
+- ms-ai-engineering / multi-modal/multimodal-rag-architecture.md
+- ms-ai-engineering / multi-modal/ocr-pipeline-architecture.md
+- ms-ai-engineering / multi-modal/real-time-audio-api.md
+- ms-ai-engineering / multi-modal/speech-to-ai-pipelines.md
+- ms-ai-engineering / multi-modal/text-to-speech-citizen.md
+- ms-ai-engineering / multi-modal/video-analysis-patterns.md
+- ms-ai-engineering / multi-modal/whisper-speech-recognition.md
+- ms-ai-governance / monitoring-observability/application-insights-llm-monitoring.md
+- ms-ai-governance / monitoring-observability/azure-monitor-setup-ai-workloads.md
+- ms-ai-governance / norwegian-public-sector-governance/norwegian-nlp-benchmarks.md
+- ms-ai-governance / responsible-ai/ai-act-microsoft-tools-mapping.md
+- ms-ai-infrastructure / hybrid-edge/azure-arc-ai-management.md
+- ms-ai-infrastructure / hybrid-edge/azure-confidential-computing-ai.md
+- ms-ai-infrastructure / hybrid-edge/azure-iot-hub-ai-pipeline.md
+- ms-ai-infrastructure / hybrid-edge/azure-local-ai-workloads.md
+- ms-ai-infrastructure / hybrid-edge/data-sovereignty-norway-public-sector.md
+- ms-ai-infrastructure / hybrid-edge/edge-ai-inferencing-patterns.md
+- ms-ai-infrastructure / hybrid-edge/edge-to-cloud-data-synchronization.md
+- ms-ai-infrastructure / hybrid-edge/hybrid-rag-architecture.md
+- ms-ai-infrastructure / hybrid-edge/iot-operations-ai-integration.md
+- ms-ai-infrastructure / hybrid-edge/kubernetes-edge-aks-edge.md
+- ms-ai-infrastructure / hybrid-edge/network-constrained-ai-deployment.md
+- ms-ai-infrastructure / hybrid-edge/offline-first-ai-applications.md
+- ms-ai-infrastructure / hybrid-edge/on-premises-slm-phi-deployment.md
+- ms-ai-infrastructure / hybrid-edge/onnx-runtime-edge-deployment.md
+- ms-ai-infrastructure / hybrid-edge/regulatory-compliance-edge-ai.md
+- ms-ai-infrastructure / hybrid-edge/sovereign-cloud-norway.md
+- ms-ai-infrastructure / hybrid-edge/windows-ai-apc-capabilities.md
+- ms-ai-security / ai-security-engineering/secure-model-deployment-hardening.md
+- ms-ai-security / performance-scalability/auto-scaling-ai-infrastructure.md
+- ms-ai-security / performance-scalability/cdn-edge-caching-ai.md
+- ms-ai-security / performance-scalability/latency-optimization-azure-openai.md
+- ms-ai-security / performance-scalability/streaming-response-patterns.md
+
+## Deferred — advisor-scope (50)
+
+Advisor reference-entries. Utenfor Step 6 sitt non-advisor-mandat; applier (Step 9) muterer dem aldri (hard advisor-fence). Kilde-valg utsatt (evt. mooted av Cosmo-fjerning R13/R14).
+
+- ms-ai-advisor / architecture/cost-models.md
+- ms-ai-advisor / architecture/licensing-matrix.md
+- ms-ai-advisor / architecture/migration-patterns.md
+- ms-ai-advisor / architecture/regional-availability-verification.md
+- ms-ai-advisor / architecture/security.md
+- ms-ai-advisor / copilot-extensibility/adaptive-cards-copilot-responses.md
+- ms-ai-advisor / copilot-extensibility/copilot-analytics-and-usage-insights.md
+- ms-ai-advisor / copilot-extensibility/copilot-api-rate-limiting-resilience.md
+- ms-ai-advisor / copilot-extensibility/copilot-connectors-design-patterns.md
+- ms-ai-advisor / copilot-extensibility/copilot-context-window-optimization.md
+- ms-ai-advisor / copilot-extensibility/copilot-dlp-and-governance.md
+- ms-ai-advisor / copilot-extensibility/copilot-extensibility-security-patterns.md
+- ms-ai-advisor / copilot-extensibility/copilot-orchestration-multi-agent.md
+- ms-ai-advisor / copilot-extensibility/copilot-prompt-engineering-governance.md
+- ms-ai-advisor / copilot-extensibility/copilot-studio-localization-globalization.md
+- ms-ai-advisor / copilot-extensibility/copilot-studio-nlp-configuration.md
+- ms-ai-advisor / copilot-extensibility/copilot-studio-topics-and-entities.md
+- ms-ai-advisor / copilot-extensibility/custom-engine-agents-development.md
+- ms-ai-advisor / copilot-extensibility/declarative-agents-fundamentals.md
+- ms-ai-advisor / copilot-extensibility/declarative-agents-grounding-strategies.md
+- ms-ai-advisor / copilot-extensibility/enterprise-governance-copilot-deployment.md
+- ms-ai-advisor / copilot-extensibility/m365-copilot-plugins-ecosystem.md
+- ms-ai-advisor / copilot-extensibility/mcp-protocol-copilot-studio.md
+- ms-ai-advisor / copilot-extensibility/microsoft-graph-api-copilot-integration.md
+- ms-ai-advisor / copilot-extensibility/power-automate-copilot-integration.md
+- ms-ai-advisor / copilot-extensibility/sharepoint-copilot-agents.md
+- ms-ai-advisor / copilot-extensibility/teams-copilot-message-extensions.md
+- ms-ai-advisor / development/agent-framework.md
+- ms-ai-advisor / platforms/azure-ai-foundry.md
+- ms-ai-advisor / platforms/copilot-studio.md
+- ms-ai-advisor / platforms/m365-copilot.md
+- ms-ai-advisor / platforms/model-catalog-2026.md
+- ms-ai-advisor / platforms/power-platform.md
+- ms-ai-advisor / prompt-engineering/adversarial-prompting-and-jailbreaks.md
+- ms-ai-advisor / prompt-engineering/chain-of-thought-prompting.md
+- ms-ai-advisor / prompt-engineering/domain-specific-prompt-optimization.md
+- ms-ai-advisor / prompt-engineering/error-handling-and-fallback-prompting.md
+- ms-ai-advisor / prompt-engineering/few-shot-learning-techniques.md
+- ms-ai-advisor / prompt-engineering/function-calling-and-tool-use.md
+- ms-ai-advisor / prompt-engineering/grounding-and-knowledge-injection.md
+- ms-ai-advisor / prompt-engineering/multi-turn-conversation-management.md
+- ms-ai-advisor / prompt-engineering/multimodal-prompt-design.md
+- ms-ai-advisor / prompt-engineering/prompt-testing-and-evaluation.md
+- ms-ai-advisor / prompt-engineering/real-time-reasoning-performance.md
+- ms-ai-advisor / prompt-engineering/reasoning-models-o1-o3-optimization.md
+- ms-ai-advisor / prompt-engineering/role-playing-and-persona-techniques.md
+- ms-ai-advisor / prompt-engineering/structured-output-formatting.md
+- ms-ai-advisor / prompt-engineering/system-message-design-patterns.md
+- ms-ai-advisor / prompt-engineering/temperature-sampling-and-parameters.md
+- ms-ai-advisor / prompt-engineering/token-optimization-and-efficiency.md
+
+## Reproduserbarhet
+
+- Deterministiske valg reproduseres av `scripts/kb-update/lib/url-normalize.extractUrls` + tie-break-reglene over mot samme korpus + `url-registry.json`.
+- De 31 residual-valgene er menneske-adjudikert (låst i denne loggen + manifestets `authorityMethod:"residual-judgment"` + `authoritySelectedBy`).
+- Manifestet er gitignored (regenererbart); den varige autoritetsposten er denne loggen og — etter Step 9 — `**Source:**`-headerne i korpuset.