agentic-diet-engine / src /utils /token_audit.py
dnyandeobharambe
feat: initial clean architectural push
e8eed08
Raw History Blame Contribute Delete
1.59 kB
import os
from dotenv import load_dotenv
from langchain_google_genai import ChatGoogleGenerativeAI
from src.retrieval.hybrid_retriever import get_hybrid_retriever
from src.retrieval.reranker import get_reranked_docs
load_dotenv()
def audit_tokens():
llm = ChatGoogleGenerativeAI(model="gemini-3-flash-preview")
query = "give me protein recipes which can give me 25g protein"
# --- PATH A: Standard Semantic Search (Top 5) ---
# Simulating what happens without your optimizations
retriever = get_hybrid_retriever() # Uses the ensemble
semantic_docs = retriever.invoke(query)[:5]
semantic_text = "\n\n".join([d.page_content for d in semantic_docs])
# --- PATH B: Your Hybrid + Reranker (Top 2) ---
reranked_docs = get_reranked_docs(query)[:2]
hybrid_text = "\n\n".join([d.page_content for d in reranked_docs])
# --- CALCULATE TOKENS ---
# Gemini has a built-in method for precise counting
sem_tokens = llm.get_num_tokens(semantic_text)
hyb_tokens = llm.get_num_tokens(hybrid_text)
savings = sem_tokens - hyb_tokens
cost_reduction = (savings / sem_tokens) * 100 if sem_tokens > 0 else 0
print("="*40)
print("ARCHITECT'S TOKEN AUDIT")
print("="*40)
print(f"Query: {query}")
print("-"*40)
print(f"Standard Semantic Tokens: {sem_tokens}")
print(f"Your Hybrid + Rerank Tokens: {hyb_tokens}")
print("-"*40)
print(f"TOTAL TOKENS SAVED: {savings}")
print(f"ESTIMATED COST REDUCTION: {cost_reduction:.2f}%")
print("="*40)
if __name__ == "__main__":
audit_tokens()