Implement prompt caching, speculative decoding config, parallel slots, and dynamic pricing metrics

backend/routers/system.py

@@ -91,6 +91,7 @@ def self_update() -> dict:
 
 
 from services.token_stats import get_stats
+from services.llamaswap import list_models
 
 @router.get("/system/token-stats")
 def token_stats() -> dict:
@@ -99,9 +100,48 @@ def token_stats() -> dict:
     c = stats.get("completion_tokens", 0)
     total = p + c
     
-    # Juni 2026 API-Preise (deutlich gestiegen):
-    # Premium-Modelle (Claude 4 / GPT-5): 15,00 $ / 1M Input-Tokens und 75,00 $ / 1M Output-Tokens
-    saved_usd = (p * 15.0 + c * 75.0) / 1_000_000.0
+    # Map model IDs and aliases to their respective roles for pricing resolution
+    role_map = {}
+    try:
+        for m in list_models():
+            role_map[m["name"].lower()] = m.get("role")
+            for alias in m.get("aliases", []):
+                role_map[alias.lower()] = m.get("role")
+    except Exception:
+        pass
+
+    # Dynamic pricing tiers based on model class in June 2026
+    PRICING = {
+        "heavy": (15.0, 75.0),
+        "coder": (3.0, 15.0),
+        "hermes": (1.0, 5.0),
+        "fast": (0.15, 0.60),
+        "scout": (0.15, 0.60),
+        "vision": (0.15, 0.60),
+        "reasoning": (0.15, 0.60),
+    }
+
+    modeled_p = 0
+    modeled_c = 0
+    saved_usd = 0.0
+    
+    models_data = stats.get("models") or {}
+    for m_name, m_tokens in models_data.items():
+        mp = m_tokens.get("prompt", 0)
+        mc = m_tokens.get("completion", 0)
+        modeled_p += mp
+        modeled_c += mc
+        
+        role = role_map.get(m_name, m_name)
+        rate_in, rate_out = PRICING.get(role, (0.15, 0.60))
+        saved_usd += (mp * rate_in + mc * rate_out) / 1_000_000.0
+
+    # Baseline/legacy tokens calculated at premium rates ($15.00 / $75.00)
+    # to preserve historical savings value prior to model-specific logging
+    baseline_p = max(0, p - modeled_p)
+    baseline_c = max(0, c - modeled_c)
+    saved_usd += (baseline_p * 15.0 + baseline_c * 75.0) / 1_000_000.0
+    
     saved_eur = saved_usd * 0.92  # 1 USD = 0.92 EUR
     
     return {