refactor(core): consolidate into Triad-Pattern (Status, Usage, History) for SLM reliability

litellm_dashboard.py

@@ -2,10 +2,10 @@
 title: LiteLLM Dashboard
 author: Kagamine Rin & VAI
 author_url: https://rin.captcouch.xyz
-description: Hardcore LiteLLM observability — health, cost breakdown, and request history with Dynamic FinOps Audit.
+description: Consolidated Hardcore LiteLLM observability — status, usage, and history.
 required_open_webui_version: 0.4.0
 requirements: requests
-version: 1.4.0
+version: 1.5.0
 licence: MIT
 """
 
@@ -14,7 +14,7 @@ import json
 from datetime import datetime, timedelta
 from zoneinfo import ZoneInfo
 from pydantic import BaseModel, Field
-from typing import Optional, List, Dict, Any
+from typing import Optional, List, Dict, Any, Literal
 
 
 class Tools:
@@ -26,7 +26,7 @@ class Tools:
         """LiteLLM proxy connection settings."""
         litellm_url: str = Field(
             default="http://localhost:4000",
-            description="LiteLLM proxy base URL (e.g. http://localhost:4000)",
+            description="LiteLLM proxy base URL",
         )
         litellm_master_key: str = Field(
             default="",
@@ -46,20 +46,6 @@ class Tools:
             description="When True, attempts to fetch live pricing from /v1/model/info for the Audit Engine.",
         )
 
-    # Fallback registry for 2026 standard tier pricing
-    MODEL_PRICING_FALLBACK = {
-        "gemini/gemini-3.5-flash": {
-            "input": 1.50 / 1_000_000,
-            "input_cached": 0.15 / 1_000_000,
-            "output": 9.00 / 1_000_000,
-        },
-        "gemini/gemini-3-flash-preview": {
-            "input": 0.50 / 1_000_000,
-            "input_cached": 0.05 / 1_000_000,
-            "output": 3.00 / 1_000_000,
-        }
-    }
-
     def _headers(self) -> dict:
         if not self.valves.litellm_master_key:
             return {}
@@ -72,7 +58,6 @@ class Tools:
         """Sync internal price cache with LiteLLM's live model registry."""
         if not self.valves.enable_dynamic_pricing or self._price_cache:
             return
-        
         try:
             url = f"{self._base_url()}/v1/model/info"
             resp = requests.get(url, headers=self._headers(), timeout=10)
@@ -88,36 +73,7 @@ class Tools:
                             "output": info.get("output_cost_per_token", 0),
                             "reasoning": info.get("output_cost_per_reasoning_token", info.get("output_cost_per_token", 0))
                         }
-        except Exception:
-            pass
-
-    def _audit_cost(self, model: str, prompt: int, cached: int, completion: int, reported_cost: float, reasoning: int = 0) -> dict:
-        """Audit reported cost with dynamic or fallback pricing."""
-        self._fetch_live_prices()
-        pricing = self._price_cache.get(model) or self.MODEL_PRICING_FALLBACK.get(model)
-        
-        if not pricing:
-            return {"status": "unsupported", "display": ""}
-        
-        uncached = max(0, prompt - cached)
-        # Handle reasoning tokens separately if pricing allows
-        actual_output = max(0, completion - reasoning)
-        
-        calc = (cached * pricing.get("input_cached", 0)) + \
-               (uncached * pricing.get("input", 0)) + \
-               (actual_output * pricing.get("output", 0)) + \
-               (reasoning * pricing.get("reasoning", pricing.get("output", 0)))
-        
-        delta = reported_cost - calc
-        status = "mismatch" if abs(delta) > 0.005 else "ok"
-        symbol = "⚠️" if status == "mismatch" else "✓"
-        
-        return {
-            "audited": calc,
-            "delta": delta,
-            "status": status,
-            "display": f" [calc: ${calc:.4f} {symbol}]"
-        }
+        except Exception: pass
 
     def _resolve_date(self, date_str: str) -> str:
         date_str = date_str.strip().lower()
@@ -125,163 +81,157 @@ class Tools:
             today = datetime.now(ZoneInfo(self.valves.timezone)).date()
         except Exception:
             today = datetime.utcnow().date()
-
         if date_str == "today": return today.isoformat()
         if date_str == "yesterday": return (today - timedelta(days=1)).isoformat()
-        
         if "days ago" in date_str:
             try:
                 n = int(date_str.split()[0])
                 return (today - timedelta(days=n)).isoformat()
             except Exception: pass
-
         try:
             datetime.strptime(date_str, "%Y-%m-%d")
             return date_str
         except ValueError: pass
         return date_str
 
+    def _audit_cost(self, model: str, prompt: int, cached: int, completion: int, reported_cost: float, reasoning: int = 0) -> str:
+        """Audit reported cost with dynamic pricing."""
+        self._fetch_live_prices()
+        pricing = self._price_cache.get(model)
+        if not pricing: return ""
+        uncached = max(0, prompt - cached)
+        actual_output = max(0, completion - reasoning)
+        calc = (cached * pricing.get("input_cached", 0)) + \
+               (uncached * pricing.get("input", 0)) + \
+               (actual_output * pricing.get("output", 0)) + \
+               (reasoning * pricing.get("reasoning", pricing.get("output", 0)))
+        delta = reported_cost - calc
+        symbol = "⚠️" if abs(delta) > 0.005 else "✓"
+        return f" [calc: ${calc:.4f} {symbol}]"
+
     def litellm_status(self) -> str:
         """
-        Get an ultra-lean snapshot of proxy health and total spend.
-        Uses /health/readiness and /global/spend/models for maximum token efficiency.
+        Comprehensive system check: Proxy readiness, DB connectivity, and Model health.
+        Use this to verify the infrastructure is up and which models are available.
         """
-        lines = [f"LiteLLM Status Snapshot — {self._base_url()}"]
-        lines.append("=" * 40)
+        lines = [f"LiteLLM Infrastructure Status — {self._base_url()}", "=" * 50]
         
-        # 1. Readiness (Infrastructure)
+        # 1. Readiness
         try:
             r = requests.get(f"{self._base_url()}/health/readiness", timeout=5)
             if r.status_code == 200:
-                data = r.json()
-                lines.append(f"Proxy: ✅ {data.get('status')} | DB: {data.get('db')} | v{data.get('litellm_version')}")
-            else:
-                lines.append(f"Proxy Readiness: ⚠️ Status {r.status_code}")
-        except Exception as e:
-            lines.append(f"Proxy Readiness: ❌ Connection Error")
+                d = r.json()
+                lines.append(f"Proxy: ✅ {d.get('status')} | DB: {d.get('db')} | v{d.get('litellm_version')}")
+            else: lines.append(f"Proxy Readiness: ⚠️ Status {r.status_code}")
+        except Exception: lines.append("Proxy Readiness: ❌ Connection Error")
 
-        # 2. Spend Snapshot (Token Efficient)
-        if self.valves.litellm_master_key:
-            try:
-                s = requests.get(f"{self._base_url()}/global/spend/models", headers=self._headers(), timeout=5)
-                if s.status_code == 200:
-                    models = s.json()
-                    total = sum(m.get("total_spend", 0) for m in models)
-                    top = sorted(models, key=lambda x: x.get("total_spend", 0), reverse=True)[:3]
-                    lines.append(f"Lifetime Spend: ${total:.2f}")
-                    if top:
-                        lines.append("Top Models: " + ", ".join([f"{m['model']}: ${m['total_spend']:.2f}" for m in top]))
-            except Exception:
-                pass
-        
-        return "\n".join(lines)
-
-    def litellm_health(self) -> str:
-        """Check LiteLLM proxy health and authentication status."""
-        url = f"{self._base_url()}/health"
+        # 2. Model Health
         try:
-            resp = requests.get(url, headers=self._headers(), timeout=15)
-        except Exception as e:
-            return f"❌ Connection failed: {str(e)}"
-
-        if resp.status_code != 200:
-            return f"❌ Error {resp.status_code}: {resp.text[:200]}"
-
-        data = resp.json()
-        healthy = data.get("healthy_endpoints", [])
-        unhealthy = data.get("unhealthy_endpoints", [])
+            h = requests.get(f"{self._base_url()}/health", headers=self._headers(), timeout=10)
+            if h.status_code == 200:
+                hd = h.json()
+                healthy = hd.get("healthy_endpoints", [])
+                unhealthy = hd.get("unhealthy_endpoints", [])
+                lines.append(f"Models: {len(healthy)} healthy, {len(unhealthy)} unhealthy")
+                for ep in healthy[:15]: # Cap display to save tokens
+                    m = ep.get("model", "unknown")
+                    lines.append(f"  ✅ {m:25s} | rpm: {ep.get('rpm','?'):>3}")
+                if len(healthy) > 15: lines.append(f"  ... and {len(healthy)-15} more healthy models.")
+                for ep in unhealthy:
+                    lines.append(f"  ❌ {ep.get('model', 'unknown')}")
+        except Exception: lines.append("Model Health: ❌ Failed to fetch")
         
-        lines = [f"LiteLLM Health Check — {self._base_url()}"]
-        lines.append("=" * 50)
-        lines.append(f"Healthy: {len(healthy)} | Unhealthy: {len(unhealthy)}")
-        
-        for ep in healthy:
-            m = ep.get("model", "unknown")
-            lines.append(f" ✅ {m:25s} | rpm: {ep.get('rpm','?'):>3} | timeout: {ep.get('timeout','?')}s")
-        for ep in unhealthy:
-            lines.append(f" ❌ {ep.get('model', 'unknown')}")
-            
         return "\n".join(lines)
 
-    def litellm_cost(self, start_date: str, end_date: str, hide_zero_spend: bool = False) -> str:
-        """Get cost breakdown with Sparkline trend visualization."""
-        if not self.valves.litellm_master_key:
-            return "❌ Master key required."
-
-        effective_hide = hide_zero_spend or self.valves.default_hide_zero_spend
+    def litellm_usage(self, start_date: str = "today", end_date: str = "today", detail_level: Literal["summary", "full"] = "summary") -> str:
+        """
+        Get fiscal usage and spend data.
+        :param start_date: YYYY-MM-DD or shortcut (today, yesterday, 7 days ago)
+        :param end_date: YYYY-MM-DD or shortcut
+        :param detail_level: 'summary' for quick spend snapshot, 'full' for daily breakdown and sparklines.
+        """
+        if not self.valves.litellm_master_key: return "❌ Master key required."
         r_start = self._resolve_date(start_date)
         r_end = self._resolve_date(end_date)
 
-        url = f"{self._base_url()}/user/daily/activity"
+        if detail_level == "summary":
+            try:
+                s = requests.get(f"{self._base_url()}/global/spend/models", headers=self._headers(), timeout=10)
+                if s.status_code == 200:
+                    models = sorted(s.json(), key=lambda x: x.get("total_spend", 0), reverse=True)
+                    total = sum(m.get("total_spend", 0) for m in models)
+                    lines = [f"LiteLLM Spend Summary (Lifetime)", "=" * 40, f"Total Lifetime Spend: ${total:.2f}"]
+                    for m in models[:5]:
+                        lines.append(f"  - {m['model']}: ${m['total_spend']:.2f}")
+                    return "\n".join(lines)
+            except Exception: pass
+
+        # Full Report
         try:
+            url = f"{self._base_url()}/user/daily/activity"
             resp = requests.get(url, headers=self._headers(), params={"start_date": r_start, "end_date": r_end}, timeout=30)
             if resp.status_code != 200: return f"❌ Error {resp.status_code}"
             data = resp.json()
         except Exception as e: return f"❌ Failed: {str(e)}"
 
         results = data.get("results", [])
-        if not results: return "No data found."
-
-        # Sparkline Generation
+        if not results: return f"No usage found for {r_start} to {r_end}."
+        
         spends = [d.get("metrics", {}).get("spend", 0) for d in results]
         def get_spark(vals):
-            if not vals: return ""
+            if len(vals) < 2: return ""
             chars = " ▂▃▄▅▆▇█"
             low, high = min(vals), max(vals)
             rng = (high - low) or 1
-            return "".join(chars[int((v - low) / rng * 7)] for v in vals)
+            return " [" + "".join(chars[int((v - low) / rng * 7)] for v in vals) + "]"
 
         meta = data.get("metadata", {})
-        lines = [f"LiteLLM Cost Report: {r_start} to {r_end}"]
-        lines.append("=" * 60)
-        lines.append(f"Total Spend: ${meta.get('total_spend', 0):.2f}  [{get_spark(spends)}]")
-        lines.append(f"Requests:    {meta.get('total_successful_requests', 0)} ok, {meta.get('total_failed_requests', 0)} fail")
+        lines = [f"LiteLLM Usage Report: {r_start} to {r_end}", "=" * 60]
+        lines.append(f"Spend: ${meta.get('total_spend', 0):.2f}{get_spark(spends)}")
+        lines.append(f"Requests: {meta.get('total_successful_requests', 0)} ok, {meta.get('total_failed_requests', 0)} fail")
         
-        # Endpoint breakdown
-        lines.append("\nBy Endpoint:")
+        lines.append("\nBreakdown by Endpoint:")
         ends = results[0].get("breakdown", {}).get("endpoints", {}) if results else {}
         for path, edata in ends.items():
             m = edata.get("metrics", {})
             lines.append(f"  {path:20s} | ${m.get('spend',0):.4f} | {m.get('successful_requests',0)} reqs")
-
         return "\n".join(lines)
 
-    def litellm_history(self, limit: int = 20, show_failed: bool = True) -> str:
-        """Request history with Thought Token detection and Dynamic Audit."""
+    def litellm_history(self, limit: int = 20, filter_model: Optional[str] = None) -> str:
+        """
+        Get recent request logs with audit data and thought detection.
+        :param limit: Number of requests (1-50)
+        :param filter_model: Optional string to filter by model name (e.g. 'gpt-4')
+        """
         if not self.valves.litellm_master_key: return "❌ Master key required."
-        
-        limit = max(1, min(50, limit))
         try:
             resp = requests.get(f"{self._base_url()}/spend/logs", headers=self._headers(), timeout=30)
             if resp.status_code != 200: return f"❌ Error {resp.status_code}"
             logs = sorted(resp.json(), key=lambda x: x.get("startTime", ""), reverse=True)
         except Exception as e: return f"❌ Failed: {str(e)}"
 
-        if not show_failed:
-            logs = [l for l in logs if not l.get("metadata", {}).get("exception")]
+        if filter_model:
+            logs = [l for l in logs if filter_model.lower() in l.get("model", "").lower()]
         
-        recent = logs[:limit]
-        lines = [f"LiteLLM History (Most Recent {len(recent)})", "=" * 90]
+        recent = logs[:max(1, min(50, limit))]
+        lines = [f"LiteLLM History (Most Recent {len(recent)})", "=" * 95]
 
         for i, entry in enumerate(recent, 1):
-            ts = entry.get("startTime", "?")[5:19]
+            ts = entry.get("startTime", "?")[5:16]
             model = entry.get("model", "unknown")[:25]
             spend = entry.get("spend", 0)
             prompt = entry.get("prompt_tokens", 0) or 0
             completion = entry.get("completion_tokens", 0) or 0
             
-            # Thought / Reasoning tokens
             usage = entry.get("usage", {}) or {}
             thought = usage.get("completion_tokens_details", {}).get("reasoning_tokens", 0)
             thought_str = f" [💭{thought}]" if thought else ""
-            
-            # Cache
             cached = entry.get("cache_read_input_tokens", 0) or 0
             cache_str = f" 💾{int(cached/prompt*100)}%" if prompt and cached else ""
             
-            audit = self._audit_cost(entry.get("model"), prompt, cached, completion, spend, reasoning=thought)
+            audit_display = self._audit_cost(entry.get("model"), prompt, cached, completion, spend, reasoning=thought)
+            status = "❌" if entry.get("metadata", {}).get("exception") else "✅"
             
-            line = f"{i:2d}. {ts} | {model:25s} | ${spend:.4f}{audit['display']} | {prompt}→{completion}{thought_str}{cache_str}"
-            lines.append(line)
-
+            lines.append(f"{i:2d}. {ts} {status} | {model:25s} | ${spend:.4f}{audit_display} | {prompt}→{completion}{thought_str}{cache_str}")
+        
         return "\n".join(lines)