fix: routing — remove turn limit from default tier, no gaps

router/router.py

@@ -172,23 +172,23 @@ def route(rd, tier):
         result = select_best_gpu(fallback, "lightweight_fallback")
         if result: return result
     
-    # TIER 2: Simple conversations — low token, short context → VLM preferred
-    if not sys and t <= 1000 and turns <= 4 and "qwen3.5-9b-vlm" in avail:
+    # TIER 2: Simple conversations — short context, any prompt → VLM preferred
+    if t <= 1000 and turns <= 4 and "qwen3.5-9b-vlm" in avail:
         if not is_gpu_busy("qwen3.5-9b-vlm"):
             return {"model":"qwen3.5-9b-vlm","reason":"simple_conv"}
         # VLM busy — try Dense
         if "qwen3.6-27B-code" in avail and not is_gpu_busy("qwen3.6-27B-code"):
             return {"model":"qwen3.6-27B-code","reason":"simple_conv_fallback"}
     
-    # TIER 3: Heavy reasoning — large context, system prompts, long conversations
-    if t > 4000 or sys or turns > 8:
+    # TIER 3: Heavy reasoning — large context or very long conversations
+    if t > 4000 or turns > 8:
         candidates = [m for m in ["qwen3.6-27B-code","qwen3.6-35B-A3B","qwen3.5-9b-vlm"] if m in avail]
         result = select_best_gpu(candidates, "heavy_reasoning")
         if result: return result
     
-    # TIER 4: Default — Dense preferred for medium tasks, MoE as workhorse, VLM as overflow
-    if turns <= 6 and t <= 4000:
-        # Medium complexity — try Dense first, then MoE, then VLM
+    # TIER 4: Default — Dense preferred for medium tasks, MoE as workhorse, VLM overflow
+    if t <= 4000:
+        # Catch everything that isn't heavy (≤4000 tokens, any turn count)
         candidates = [m for m in ["qwen3.6-27B-code","qwen3.6-35B-A3B","qwen3.5-9b-vlm"] if m in avail]
         result = select_best_gpu(candidates, "medium_task")
         if result: return result