القائمة
Track token usage and costs across agents for budget management
يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.
اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
يتجاوز الأمر المباشر Prompt المخصّص للمراجعة. افحص المصدر قبل تشغيله.
npx skills add https://github.com/nexus-labs-automation/agent-observability --skill token-cost-trackingيبقى الأمر في سطر واحد. مرّر أفقيًا لمراجعته كاملًا قبل النسخ.
تفضّل نسخة محلية؟ نزّل الملفات المتاحة حاليًا لدى SkillsMP.
عرض SKILL.md
استنادا إلى تصنيف SOC المهني
| name | token-cost-tracking |
| description | Track token usage and costs across agents for budget management |
| triggers | ["track tokens","token usage","cost tracking","LLM costs","budget monitoring"] |
| priority | 1 |
Track token usage and costs across agents for budget management and optimization.
Every organization needs to answer:
# Per-call token tracking
span.set_attribute("llm.tokens.input", 1500)
span.set_attribute("llm.tokens.output", 350)
span.set_attribute("llm.tokens.total", 1850)
# Per-call cost
span.set_attribute("llm.cost_usd", 0.025)
# Attribution
span.set_attribute("cost.feature", "document_analysis")
span.set_attribute("cost.agent", "researcher")
span.set_attribute("cost.user_id", "user_abc") # Hashed
span.set_attribute("cost.org_id", "org_123")
Keep pricing updated (prices as of late 2024):
MODEL_PRICING = {
# Anthropic (per 1M tokens)
"claude-3-opus": {"input": 15.00, "output": 75.00},
"claude-3-5-sonnet": {"input": 3.00, "output": 15.00},
"claude-3-5-haiku": {"input": 0.80, "output": 4.00},
"claude-3-haiku": {"input": 0.25, "output": 1.25},
# OpenAI (per 1M tokens)
"gpt-4-turbo": {"input": 10.00, "output": 30.00},
"gpt-4o": {"input": 2.50, "output": 10.00},
"gpt-4o-mini": {"input": 0.15, "output": 0.60},
"gpt-3.5-turbo": {"input": 0.50, "output": 1.50},
# Embeddings (per 1M tokens)
"text-embedding-3-large": {"input": 0.13, "output": 0},
"text-embedding-3-small": {"input": 0.02, "output": 0},
: {: , : },
}
def calculate_cost(
model: str,
input_tokens: int,
output_tokens: int,
cached_tokens: int = 0,
cache_discount: float = 0.9 # 90% discount for cached
) -> float:
"""Calculate cost for an LLM call."""
pricing = MODEL_PRICING.get(model)
if not pricing:
return 0.0
# Cached tokens are discounted
effective_input = input_tokens - cached_tokens
cached_cost = (cached_tokens / 1_000_000) * pricing["input"] * (1 - cache_discount)
input_cost = (effective_input / 1_000_000) * pricing["input"]
output_cost = (output_tokens / 1_000_000) * pricing["output"]
return round(input_cost + cached_cost + output_cost, 6)
Track at multiple granularities:
span.set_attribute("llm.cost_usd", 0.025)
span.set_attribute("agent.total_tokens", 15000)
span.set_attribute("agent.total_cost_usd", 0.45)
span.set_attribute("agent.llm_calls", 5)
span.set_attribute("session.total_tokens", 45000)
span.set_attribute("session.total_cost_usd", 1.35)
span.set_attribute("session.agent_runs", 3)
Track in your observability platform, not in spans.
Set up alerts for:
BUDGET_THRESHOLDS = {
"per_call_max": 1.00, # Alert if single call > $1
"per_session_max": 10.00, # Alert if session > $10
"per_user_daily": 50.00, # Alert if user > $50/day
"org_daily": 1000.00, # Alert if org > $1000/day
}
def check_budget(cost: float, level: str, entity_id: str):
threshold = BUDGET_THRESHOLDS.get(f"{level}_max")
if threshold and cost > threshold:
log_budget_alert(level, entity_id, cost, threshold)
from langfuse import Langfuse
langfuse = Langfuse()
# Automatic token/cost tracking
trace = langfuse.trace(name="agent_run")
generation = trace.generation(
name="llm_call",
model="claude-3-5-sonnet",
usage={
"input": 1500,
"output": 350,
"unit": "TOKENS"
}
)
# Langfuse calculates cost automatically
from langsmith import Client
client = Client()
# Token tracking automatic via callbacks
# Cost calculation in LangSmith dashboard
from opentelemetry import trace
tracer = trace.get_tracer("agent")
with tracer.start_as_current_span("llm_call") as span:
span.set_attribute("llm.tokens.input", 1500)
span.set_attribute("llm.tokens.output", 350)
span.set_attribute("llm.cost_usd", calculate_cost(...))
Track cache hits for accurate costs:
span.set_attribute("llm.cache.hit", True)
span.set_attribute("llm.cache.tokens_saved", 1200)
span.set_attribute("llm.cache.cost_saved_usd", 0.018)
Track metrics that indicate optimization opportunities:
# Prompt efficiency
span.set_attribute("prompt.compression_ratio", 0.7)
span.set_attribute("prompt.could_use_haiku", True)
# Model selection
span.set_attribute("model.recommendation", "could_downgrade")
span.set_attribute("model.quality_requirement", "low")
llm-call-tracing - LLM instrumentationsession-conversation-tracking - Session aggregation