First Commit

.env

@@ -0,0 +1,2 @@
+BUILD_VERSION=20231101.1
+PRICE_UPDATE=2023-11-01

.gitignore

@@ -0,0 +1,5 @@
+__pycache__
+venv
+
+.DS_Store
+*.pyc

app.py

@@ -0,0 +1,70 @@
+# Description: LLM Cost Calculator Web App
+# Author: Vincent CIBELLI (https://www.linkedin.com/in/vincentcibelli/)
+# Date: 2023-11-01
+# Last Modified: 2023-11-01
+# License: MIT License
+
+import os
+import streamlit as st
+import pandas as pd
+from dotenv import load_dotenv
+from babel.numbers import format_currency
+from services.llm_service import LLMService
+from services.table_service import TableService
+
+
+load_dotenv()
+
+llm_service = LLMService()
+
+st.set_page_config(
+    page_title="LLM Cost Calculator",
+    page_icon="💸",
+    layout="wide",
+    menu_items={
+        'About': "## LLM Cost Calculator\n#### Build {}\nCreated with ❤️ by [Vincent CIBELLI](https://www.linkedin.com/in/vincentcibelli/)".format(os.environ.get("BUILD_VERSION"))
+    }
+)
+
+st.header("💸 LLM Cost Calculator 💸", divider="green")
+
+with st.expander("Last price update ($ currency): {}.".format(os.getenv("PRICE_UPDATE")), expanded=False):
+    st.table(TableService.format_source_table(llm_service.get_list()))
+
+st.subheader("Your project information:")
+with st.expander("#### Average use of your LLM:", expanded=True):
+    col1, col2, col3 = st.columns(3)
+
+    with col1:
+        request_number = st.number_input("Average request number per hour", min_value=0, value=0, step=100)
+
+    with col2:
+        col2_1, col2_2 = st.columns(2)
+        
+        with col2_1:
+            request_size = st.number_input("Average input token size", min_value=0, value=0, step=1000)
+
+        with col2_2:
+            response_size = st.number_input("Average output token size", min_value=0, value=0, step=1000)
+
+    with col3:
+        col3_1, col3_2 = st.columns(2)
+
+        with col3_1:
+            run_hours = st.number_input("Running hours per day", min_value=0, max_value=24, value=0, step=1)
+        
+        with col3_2:
+            run_day = st.number_input("Running days per week", min_value=0, max_value=7, value=0, step=1)
+
+with st.expander("#### LLM models to compare:", expanded=True):
+    llm_chat_list = st.multiselect("Select Chat LLM models", options=[llm.name for llm in llm_service.get_list_by_category("chat")])
+    llm_text_list = st.multiselect("Select Text LLM models", options=[llm.name for llm in llm_service.get_list_by_category("text")])
+
+result = []
+for llm in llm_chat_list + llm_text_list:
+    result.append(llm_service.get_result(llm_service.get_by_name(llm), request_number, request_size, response_size, run_hours, run_day))
+
+if len(result) > 0:
+    st.subheader("Your Cost Result:")
+
+    st.table(TableService.format_result_table(result))

data/llm.json

@@ -0,0 +1,13 @@
+[
+    { "id": "gpt-35-turbo-4k", "name": "GPT 3.5 Turbo 4K", "category": "chat", "type": "saas", "input_price": 0.0015, "output_price": 0.002 },
+    { "id": "gpt-35-turbo-16k", "name": "GPT 3.5 Turbo 16K", "category": "chat", "type": "saas", "input_price": 0.003, "output_price": 0.004 },
+    { "id": "gpt4-8k", "name": "GPT4 8K", "category": "chat", "type": "saas", "input_price": 0.03, "output_price": 0.06 },
+    { "id": "gpt4-32k", "name": "GPT4 32K", "category": "chat", "type": "saas", "input_price": 0.06, "output_price": 0.12 },
+    { "id": "claude-instant", "name": "Claude Instant", "category": "text", "type": "saas", "input_price": 0.00163, "output_price": 0.00551 },
+    { "id": "claude-2", "name": "Claude 2", "category": "text", "type": "saas", "input_price": 0.01102, "output_price": 0.03268 },
+    { "id": "palm2-chat-bison", "name": "PaLM 2 for Chat (Bison)", "category": "chat", "type": "saas", "input_price": 0.002, "output_price": 0.002},
+    { "id": "palm2-text-bison", "name": "PaLM 2 for Text (Bison)", "category": "text", "type": "saas", "input_price": 0.002, "output_price": 0.002 },
+    { "id": "gpt-35-turbo-4k-finetuned", "name": "GPT 3.5 Turbo - Fine-Tuned", "category": "chat", "type": "saas", "input_price": 0.012, "output_price": 0.016 },
+    { "id": "titan-text–lite", "name": "AWS Titan Text – Lite", "category": "text", "type": "saas", "input_price": 0.0003, "output_price": 0.0004 },
+    { "id": "titan-text–express", "name": "AWS Titan Text – Express", "category": "text", "type": "saas", "input_price": 0.0013, "output_price": 0.0017 }
+]

models/llm.py

@@ -0,0 +1,8 @@
+class LLMModel:
+    def __init__(self, id, name, category, type, input_price, output_price):
+        self.id = id
+        self.name = name
+        self.category = category
+        self.type = type
+        self.input_price = input_price
+        self.output_price = output_price

requirements.txt

@@ -0,0 +1,45 @@
+altair==5.1.2
+attrs==23.1.0
+Babel==2.13.1
+blinker==1.6.3
+cachetools==5.3.2
+certifi==2023.7.22
+charset-normalizer==3.3.2
+click==8.1.7
+gitdb==4.0.11
+GitPython==3.1.40
+idna==3.4
+importlib-metadata==6.8.0
+Jinja2==3.1.2
+jsonschema==4.19.2
+jsonschema-specifications==2023.7.1
+markdown-it-py==3.0.0
+MarkupSafe==2.1.3
+mdurl==0.1.2
+numpy==1.26.1
+packaging==23.2
+pandas==2.1.2
+Pillow==10.1.0
+protobuf==4.24.4
+pyarrow==14.0.0
+pydeck==0.8.1b0
+Pygments==2.16.1
+python-dateutil==2.8.2
+python-dotenv==1.0.0
+pytz==2023.3.post1
+referencing==0.30.2
+requests==2.31.0
+rich==13.6.0
+rpds-py==0.10.6
+six==1.16.0
+smmap==5.0.1
+tenacity==8.2.3
+toml==0.10.2
+toolz==0.12.0
+tornado==6.3.3
+typing_extensions==4.8.0
+tzdata==2023.3
+tzlocal==5.2
+urllib3==2.0.7
+validators==0.22.0
+zipp==3.17.0

services/llm_service.py

@@ -0,0 +1,28 @@
+import json
+from decimal import Decimal
+from models.llm import LLMModel
+
+class LLMService:
+    def __init__(self):
+        with open('data/llm.json', 'r') as f:
+            self.llm_list = json.load(f, object_hook=lambda d: LLMModel(**d))
+            self.llm_list = sorted(self.llm_list, key=lambda x: x.name)
+
+    def get_list(self) -> list[LLMModel]:
+        return self.llm_list
+    
+    def get_list_by_category(self, category) -> list[LLMModel]:
+        return [llm for llm in self.llm_list if llm.category == category]
+    
+    def get_by_name(self, name) -> LLMModel:
+        return next((llm for llm in self.llm_list if llm.name == name), None)
+    
+    def get_result(self, llm: LLMModel, request_number: int, request_size: int, response_size: int, run_hours: int, run_day: int) -> dict:
+        return {
+            "model": llm.name,
+            "category": llm.category,
+            "price_per_request": Decimal(llm.input_price * (request_size / 1000) + llm.output_price * (response_size / 1000)),
+            "price_per_day": Decimal((llm.input_price * (request_size / 1000) + llm.output_price * (response_size / 1000)) * request_number * run_hours),
+            "price_per_week": Decimal((llm.input_price * (request_size / 1000) + llm.output_price * (response_size / 1000)) * request_number * run_hours * run_day),
+            "price_per_month": Decimal((llm.input_price * (request_size / 1000) + llm.output_price * (response_size / 1000)) * request_number * run_hours * run_day * 4.3)
+        }

services/table_service.py

@@ -0,0 +1,33 @@
+import pandas as pd
+from babel.numbers import format_currency
+from models.llm import LLMModel
+
+class TableService:
+    def format_source_table(data: list[LLMModel]) -> pd.DataFrame:
+        src = []
+        for llm in data:
+            src.append({
+                "name": llm.name,
+                "category": llm.category,
+                "type": llm.type,
+                "input_price": llm.input_price,
+                "output_price": llm.output_price
+            })
+
+        table_source = pd.DataFrame(src)
+        table_source = table_source.set_index("name")
+        table_source = table_source.rename(columns={"name": "Model", "category": "Category", "type": "Type", "input_price": "Input price in $ (1000 tokens)", "output_price": "Output price in $ (1000 tokens)"})
+
+        return table_source
+    
+    def format_result_table(data: list[LLMModel]) -> pd.DataFrame:
+        table_result = pd.DataFrame(data)
+        table_result = table_result.set_index("model")
+        table_result = table_result.rename(columns={"model": "Model", "category": "Category", "price_per_request": "Price per request", "price_per_day": "Price per day", "price_per_week": "Price per week", "price_per_month": "Price per month"})
+        table_result["Price per request"] = table_result["Price per request"].apply(lambda x: format_currency(x, currency="USD", locale="en_US"))
+        table_result["Price per day"] = table_result["Price per day"].apply(lambda x: format_currency(x, currency="USD", locale="en_US"))
+        table_result["Price per week"] = table_result["Price per week"].apply(lambda x: format_currency(x, currency="USD", locale="en_US"))
+        table_result["Price per month"] = table_result["Price per month"].apply(lambda x: format_currency(x, currency="USD", locale="en_US"))
+        table_result = table_result.sort_values(["Category", "Price per request"])
+
+        return table_result