GEMINI-3X-PREVIEW-PROFILE-REFRESH (AI Runtime)

Release digest

GEMINI-3X-PREVIEW-PROFILE-REFRESH (AI Runtime) — The optional Gemini preview profile now uses the official gemini-3.5-flash model for fast chat, sentiment, and ADK paths while keeping gemini-3.1-pro-preview as the report-model preview path. The stable production profile remains on the Gemini 2.5 family until a separate deployment decision explicitly changes GEMINI_LOCATION and accepts the preview profile's location tradeoffs.
Gemini 3.x runtime calls now omit legacy sampling parameters such as temperature, while Gemini 2.5 and 2.0 fallback models continue to receive the existing sampling configuration.
Added dedicated governance policy docs for agent-role ownership and source-of-truth update decisions.

Added focused regression coverage for Gemini 3.x sampling-config behavior and re-ran the model fallback, google.genai text, and chat model runtime tests.
Rebuilt the public knowledge base and refreshed the knowledge-base snapshot after user-facing AI model documentation updates.
Updated canonical project docs, AI runtime architecture, deployment/env inventory, infrastructure docs, governance index, and public knowledge-base AI pages to describe the stable/preview model profiles and the new governance role/source-update policies. Public docs decision: user-facing docs were updated because the model profile descriptions are visible product/runtime configuration guidance; no legal, investment, audit, safety, or guarantee claims were added.