[Sebas1705]

Docs-only follow-up to #10 and #12.

• AGENTS.md: image build is dispatched on merge, deploys are manual-only, no staging host.
• Architecture (target-architecture, c4-views, repository-map, domain-model, glossary, use-cases): only the ai-engine-api llama target remains as runtime state; gateway and BFF are stateless; RoutingTarget/preset model replaced by LlamaTarget (fields from llama_target_store.py).
• ADR 0005 and ADR 0001: dated 'Superseded in part' notes; decision text untouched.
• Guides (ai-content-generation, ai-observability, backoffice-operational-diagnostics, capability matrix, functional overview, validation smoke matrix, local edge integration, others): presets, service-target overrides, Kubernetes overview and gateway target removed; WF-06 reflects smoke-edge.sh (llama target only).
• Operations (threat-model, environments-and-secrets, deployment-strategy, cicd-workflow-map, functionality baseline, minimum-standards, branch protection, standards adoption): auto-deploy statements marked thesis-era with a status note.

Verified: facts checked against origin/main of bff-backoffice routes, ai-engine llama_target_store.py and platform-infra workflows/smoke scripts; relative-link check (Node port of validate-docs.yml) passes on 83 files.