v1.1.0 — TRAE model discovery and visibility

TRAE CN model management for DeepSeek Harness 0.2.0-rc.2.

• Discover the active account's genuine Remote/SOLO models, capacities and reasoning capabilities; no compatibility aliases masquerade as separate models.
• Configure each model's advertised Low/High/Xhigh strengths. Mixed and non-reasoning catalogs do not inherit an incompatible High default.
• Search, per-model and bulk show/hide controls with persisted preferences and live DSH Settings API synchronization.
• Preserve preferences across refresh/restart; hiding every model is reversible. Hiding removes a model from DSH's configured catalog, so sessions using it should switch first.
• Restore confirmed model names after restart and retain existing private Python environments/accounts on update.
• Native account controls, Desktop-compatible console links and owned relay lifecycle remain available.

Windows, macOS and Ubuntu CI validate 11 Node tests, 129 Python protocol/capability tests, package privacy audit and real DSH install/discovery/visibility/persistence tests using a simulated upstream. An isolated DSH UI verified the controls and the native picker. Actual model entitlement/quota depends on the TRAE account; tests do not perform billable live inference.

Download dsh-trae-relay.tgz and SHA256SUMS.txt. See README for setup and upgrade instructions. The release package contains no personal account data or local configuration.