""" title: LiteLLM Dashboard author: Kagamine Rin & VAI author_url: https://rin.captcouch.xyz description: Professional LiteLLM observability — High-density health, cost breakdown, and request history. required_open_webui_version: 0.4.0 requirements: requests version: 1.5.1 licence: MIT """ import requests import json from datetime import datetime, timedelta from zoneinfo import ZoneInfo from pydantic import BaseModel, Field from typing import Optional, List, Dict, Any, Literal class Tools: def __init__(self): self.valves = self.Valves() self._price_cache = {} class Valves(BaseModel): """LiteLLM proxy connection settings.""" litellm_url: str = Field( default="http://localhost:4000", description="LiteLLM proxy base URL (e.g. http://localhost:4000)", ) litellm_master_key: str = Field( default="", description="LiteLLM master API key for authentication", json_schema_extra={"input": {"type": "password"}}, ) default_hide_zero_spend: bool = Field( default=True, description="Default behavior for zero-spend model filtering.", ) timezone: str = Field( default="UTC", description="IANA Timezone name (e.g. America/New_York).", ) enable_dynamic_pricing: bool = Field( default=True, description="When True, attempts to fetch live pricing from /v1/model/info for the Audit Engine.", ) def _headers(self) -> dict: if not self.valves.litellm_master_key: return {} return {"Authorization": f"Bearer {self.valves.litellm_master_key}"} def _base_url(self) -> str: return self.valves.litellm_url.rstrip("/") def _fetch_live_prices(self): 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) if resp.status_code == 200: data = resp.json().get("data", []) for m in data: name = m.get("model_name") info = m.get("model_info", {}) if name and info: self._price_cache[name] = { "input": info.get("input_cost_per_token", 0), "input_cached": info.get("cache_read_input_token_cost", 0), "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 _resolve_date(self, date_str: str) -> str: date_str = date_str.strip().lower() try: 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: 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") or 0)) + \ (uncached * (pricing.get("input") or 0)) + \ (actual_output * (pricing.get("output") or 0)) + \ (reasoning * (pricing.get("reasoning") or pricing.get("output") or 0)) delta = reported_cost - calc symbol = "⚠️" if abs(delta) > 0.005 else "✓" return f" [calc: ${calc:.4f} {symbol}]" def litellm_status(self) -> str: """ 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 Infrastructure Status — {self._base_url()}"] lines.append("=" * 60) try: r = requests.get(f"{self._base_url()}/health/readiness", timeout=5) if r.status_code == 200: 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") try: 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"\nHealthy Models: {len(healthy)} | Unhealthy: {len(unhealthy)}") lines.append("-" * 60) for ep in healthy: m = ep.get("model", "unknown") display = m.replace("gemini/", "").replace("openai/", "") lines.append(f" ✅ {display:30s} — rpm: {ep.get('rpm','?'):>3}, timeout: {ep.get('timeout','?')}s") for ep in unhealthy: lines.append(f" ❌ {ep.get('model', 'unknown')}") except Exception as e: lines.append(f"\n❌ Health Check Failed: {str(e)}") return "\n".join(lines) 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 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) 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 Snapshot (Lifetime)", "=" * 50, f"Total Lifetime Spend: ${total:.2f}\n"] lines.append(f"{'Model':30s} | {'Spend':>10s}") lines.append("-" * 45) for m in models[:8]: display = m['model'].split("/")[-1] lines.append(f" {display:28s} | ${m['total_spend']:9.2f}") return "\n".join(lines) except Exception: pass 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 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 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) + "]" meta = data.get("metadata", {}) lines = [f"LiteLLM Usage Report: {r_start} to {r_end}"] lines.append("=" * 75) lines.append(f"Total Spend: ${meta.get('total_spend', 0):.2f}{get_spark(spends)}") lines.append(f"Total Tokens: {meta.get('total_tokens', 0):,} ({meta.get('total_prompt_tokens',0):,} prompt + {meta.get('total_completion_tokens',0):,} completion)") lines.append(f"Requests: {meta.get('total_successful_requests', 0)} successful, {meta.get('total_failed_requests', 0)} failed") # Model Breakdown model_totals = {} for day in results: models = day.get("breakdown", {}).get("models", {}) for m_name, m_data in models.items(): metrics = m_data.get("metrics", {}) if m_name not in model_totals: model_totals[m_name] = {"spend": 0, "reqs": 0, "tokens": 0} model_totals[m_name]["spend"] += metrics.get("spend", 0) model_totals[m_name]["reqs"] += metrics.get("successful_requests", 0) model_totals[m_name]["tokens"] += metrics.get("total_tokens", 0) lines.append(f"\n{'Model':30s} | {'Spend':>10s} | {'Requests':>12s} | {'Tokens':>12s}") lines.append("-" * 75) sorted_models = sorted(model_totals.items(), key=lambda x: x[1]["spend"], reverse=True) hidden = 0 for m_name, m_stats in sorted_models: if m_stats["spend"] < 0.001 and self.valves.default_hide_zero_spend: hidden += 1 continue display = m_name.split("/")[-1][:30] lines.append(f" {display:28s} | ${m_stats['spend']:9.2f} | {m_stats['reqs']:10d} ok | {m_stats['tokens']:10,d}") if hidden: lines.append(f" ... ({hidden} models with sub-$0.001 spend hidden)") lines.append(f"\n{'Endpoint':30s} | {'Spend':>10s} | {'Requests':>12s}") lines.append("-" * 60) ends = results[0].get("breakdown", {}).get("endpoints", {}) if results else {} for path, edata in sorted(ends.items(), key=lambda x: x[1].get("metrics",{}).get("spend",0), reverse=True): m = edata.get("metrics", {}) lines.append(f" {path:28s} | ${m.get('spend',0):9.4f} | {m.get('successful_requests',0):10d} ok") return "\n".join(lines) def litellm_history(self, limit: int = 20, filter_model: Optional[str] = None) -> str: """Get recent request logs with audit data and thought detection.""" if not self.valves.litellm_master_key: return "❌ Master key required." 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 filter_model: logs = [l for l in logs if filter_model.lower() in l.get("model", "").lower()] recent = logs[:max(1, min(50, limit))] lines = [f"LiteLLM History — {self._base_url()}", "=" * 105] for i, entry in enumerate(recent, 1): ts = entry.get("startTime", "?")[5:16] model = entry.get("model", "unknown").split("/")[-1][:25] spend = entry.get("spend", 0) prompt = entry.get("prompt_tokens", 0) or 0 completion = entry.get("completion_tokens", 0) or 0 usage = entry.get("usage", {}) or {} thought = usage.get("completion_tokens_details", {}).get("reasoning_tokens", 0) thought_str = f" [💭{thought}]" if thought else "" 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) status = "✅" if not entry.get("metadata", {}).get("exception") else "❌" p_s = f"{prompt/1000:.1f}k" if prompt >= 1000 else str(prompt) c_s = f"{completion/1000:.1f}k" if completion >= 1000 else str(completion) lines.append(f"{i:2d}. {ts} {status} | {model:25s} | ${spend:.4f}{audit:18s} | {p_s:>6s}→{c_s:<6s}{thought_str}{cache_str}") return "\n".join(lines)