feat(gateway): configurable max_tool_loops per schedule, bump default to 50

refactor(auth): remove separate admin token, use API token for all endpoints

feat(gateway): per-schedule model override and OpenClaw import upsert

fix: CI failures — clippy, TypeScript types, Windows tests, doctest

docs: add LLM router implementation plan

docs: LLM router design — embedding-based smart routing

Merge pull request #11 from sblanchard/worktree-llm-router

fix: record routing decisions, validate routing_profile, fix clearing profile

test(providers): add smart router integration tests

fix(dashboard): show router card error state and loading indicator

feat(dashboard): add routing profile dropdown to schedule forms

feat(dashboard): add LLM Router card to Settings page

fix(dashboard): make avg_latency_ms optional, fix routerDecisions response shape

feat(dashboard): add router API client methods and types

feat(gateway): initialize smart router at startup from config