router: 4 optimizations — saturated flag fix, heavy tier MoE-first, better token est, session tracking

- Saturated flag now triggers on load saturation (was dead code)
- Heavy tier routes MoE(131K) first instead of Dense(98K)
- Token estimation uses JSON length/3.5 (was content/4)
- Cross-turn session tracking via X-Session-Id + Redis TTL 24h

router/router.py

@@ -117,7 +117,9 @@ def check_gpu_health(model):
 
 def available_models(): return [m for m in GPU_URLS if check_gpu_health(m)["status"] in ("healthy","saturated")]
 
-def estimate_tokens(msgs): return sum(len(str(m.get("content",""))) for m in msgs) // 4
+def estimate_tokens(msgs):
+    """Estimate token count from messages. Uses JSON length / 3.5 (closer to real tokenizer ratios for dense text)."""
+    return len(json.dumps(msgs, default=str)) // 3.5
 
 def is_gpu_busy(model):
     """Check if GPU is at or near max concurrent capacity."""
@@ -150,6 +152,9 @@ def route(rd, tier):
     allowed = TIER_MODELS.get(tier, ["qwen3.5-9b-vlm"])
     avail = [m for m in available_models() if m in allowed]
     if not avail: return {"model": allowed[0], "reason": "all_saturated", "saturated": True}
+    # Check if all available GPUs are at max capacity
+    if all(is_gpu_busy(m) for m in avail):
+        return {"model": avail[0], "reason": "all_saturated", "saturated": True}
     
     req = rd.get("model","auto")
     if req != "auto":
@@ -190,8 +195,8 @@ def route(rd, tier):
     
     # TIER 3: Heavy reasoning — extremely large context or very long conversations
     if t > 50000 or turns > 25:
-        # Dense first (98K, purpose-built for reasoning), then MoE/VLM 131K
-        candidates = [m for m in ["qwen3.6-27B-code","qwen3.6-35B-A3B","qwen3.5-9b-vlm"] if m in avail]
+        # MoE first (131K context handles heavy sessions), then Dense (98K reasoning), then Light (131K fallback)
+        candidates = [m for m in ["qwen3.6-35B-A3B","qwen3.6-27B-code","qwen3.5-9b-vlm"] if m in avail]
         result = select_best_gpu(candidates, "heavy_reasoning")
         if result: return result
     
@@ -265,6 +270,17 @@ def chat():
         # Allow agent to override queue timeout via header
         q_timeout = int(request.headers.get("X-Queue-Timeout", str(QUEUE_TIMEOUT)))
         
+        # Cross-turn context tracking: accumulate tokens per session
+        session_id = request.headers.get("X-Session-Id", "")
+        session_tokens = 0
+        if session_id and r:
+            try:
+                prev = int(r.get("session:" + session_id) or 0)
+                current = estimate_tokens(rd.get("messages",[]))
+                session_tokens = max(prev, current)  # context only grows
+                r.set("session:" + session_id, session_tokens, ex=86400)  # TTL 24h
+            except Exception: pass
+        
         d = route(rd, tier)
         queue_start = time.time()
         
@@ -307,7 +323,7 @@ def chat():
                 for raw in resp.iter_content(chunk_size=None, decode_unicode=True):
                     if raw: yield clean_unicode(raw)
             bcast()
-            ctx_remaining = GPU_CONTEXT.get(model, 65536) - estimate_tokens(rd.get("messages",[]))
+            ctx_remaining = GPU_CONTEXT.get(model, 65536) - max(session_tokens, estimate_tokens(rd.get("messages",[])))
             sse_resp = Response(stream_with_context(gen()), mimetype="text/event-stream")
             sse_resp.headers["X-Context-Remaining"] = str(max(0, ctx_remaining))
             sse_resp.headers["X-Context-Model"] = model
@@ -317,7 +333,7 @@ def chat():
             msg = c.get("message",{})
             if not msg.get("content") and msg.get("reasoning_content"):
                 msg["content"] = msg["reasoning_content"]
-        ctx_remaining = GPU_CONTEXT.get(model, 65536) - estimate_tokens(rd.get("messages",[]))
+        ctx_remaining = GPU_CONTEXT.get(model, 65536) - max(session_tokens, estimate_tokens(rd.get("messages",[])))
         data["routing"] = {"model":model,"reason":reason,"gpu":url,"tier":tier,"agent":agent,"latency_ms":lat,"active_gpu":gpu_active_count(model),"context_remaining": max(0, ctx_remaining)}
         resp = jsonify(data)
         resp.headers["X-Context-Remaining"] = str(max(0, ctx_remaining))