Add routecheck service and CommuteTool fast tool

routecheck/ — FastAPI service (port 8090):
  - Image captcha (PIL: arithmetic problem, noise, wave distortion)
  - POST /api/captcha/new + /api/captcha/solve → short-lived token
  - GET /api/route?from=lat,lon&to=lat,lon&token=... → Yandex Routing API
  - Internal bypass via INTERNAL_TOKEN env var (for CommuteTool)
  - HTTPS proxy forwarded to reach Yandex API from container

CommuteTool (fast_tools.py):
  - Matches commute/traffic/arrival time queries
  - Calls routecheck /api/route with ROUTECHECK_TOKEN
  - Hardcoded route: Balashikha home → Moscow center
  - Returns traffic-adjusted travel time + delay annotation

Needs: YANDEX_ROUTING_KEY + ROUTECHECK_TOKEN in .env

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

agent.py

@@ -24,7 +24,7 @@ from langchain_core.tools import Tool
 from vram_manager import VRAMManager
 from router import Router
 from agent_factory import build_medium_agent, build_complex_agent
-from fast_tools import FastToolRunner, WeatherTool
+from fast_tools import FastToolRunner, WeatherTool, CommuteTool
 import channels
 
 # Bifrost gateway — all LLM inference goes through here
@@ -38,6 +38,8 @@ COMPLEX_MODEL = os.getenv("DEEPAGENTS_COMPLEX_MODEL", "qwen3:8b")
 SEARXNG_URL = os.getenv("SEARXNG_URL", "http://host.docker.internal:11437")
 OPENMEMORY_URL = os.getenv("OPENMEMORY_URL", "http://openmemory:8765")
 CRAWL4AI_URL = os.getenv("CRAWL4AI_URL", "http://crawl4ai:11235")
+ROUTECHECK_URL = os.getenv("ROUTECHECK_URL", "http://routecheck:8090")
+ROUTECHECK_TOKEN = os.getenv("ROUTECHECK_TOKEN", "")
 
 MAX_HISTORY_TURNS = 5
 _conversation_buffers: dict[str, list] = {}
@@ -122,6 +124,7 @@ _memory_search_tool = None
 # Fast tools run before the LLM — classifier + context enricher
 _fast_tool_runner = FastToolRunner([
     WeatherTool(searxng_url=SEARXNG_URL),
+    CommuteTool(routecheck_url=ROUTECHECK_URL, internal_token=ROUTECHECK_TOKEN),
 ])
 
 # GPU mutex: one LLM inference at a time