| import streamlit as st
|
|
|
| st.set_page_config(page_title="AI Tool Integration Evaluator", layout="centered")
|
|
|
| custom_css = """
|
| <style>
|
| @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;600;700&display=swap');
|
|
|
| html, body, [class*="st-"], .stMarkdown, p, h1, h2, h3, h4, h5, h6, label, div {
|
| font-family: 'Inter', sans-serif !important;
|
| color: #00205b !important;
|
| }
|
| h1 { font-size: 2.2rem !important; }
|
| h2 { font-size: 1.8rem !important; }
|
| h3 { font-size: 1.5rem !important; font-weight: 600 !important; }
|
| p, .stMarkdown p { font-size: 1.15rem !important; }
|
| .stRadio label p { font-size: 1.15rem !important; }
|
| .stNumberInput input { font-size: 1.15rem !important; color: #00205b !important; }
|
| [data-testid="stMetricValue"] { font-size: 2rem !important; color: #00205b !important; }
|
| [data-testid="stMetricLabel"] { font-size: 1.2rem !important; color: #00205b !important; }
|
| </style>
|
| """
|
| st.markdown(custom_css, unsafe_allow_html=True)
|
|
|
| if "step" not in st.session_state: st.session_state.step = 0
|
| if "N" not in st.session_state: st.session_state.N = 0
|
| if "U" not in st.session_state: st.session_state.U = 0
|
| if "model_choice" not in st.session_state: st.session_state.model_choice = ""
|
| if "FEin" not in st.session_state: st.session_state.FEin = 0.0
|
| if "FEout" not in st.session_state: st.session_state.FEout = 0.0
|
| if "Tin" not in st.session_state: st.session_state.Tin = 0
|
| if "Tout" not in st.session_state: st.session_state.Tout = 0
|
| if "Tin_type" not in st.session_state: st.session_state.Tin_type = ""
|
| if "Tout_type" not in st.session_state: st.session_state.Tout_type = ""
|
| if "R" not in st.session_state: st.session_state.R = 0
|
| if "F" not in st.session_state: st.session_state.F = 0
|
| if "G" not in st.session_state: st.session_state.G = 0
|
| if "testing" not in st.session_state: st.session_state.testing = 0
|
| if "dev" not in st.session_state: st.session_state.dev = 0
|
| if "nb_projects_pod" not in st.session_state: st.session_state.nb_projects_pod = 1
|
| if "needs_web_interface" not in st.session_state: st.session_state.needs_web_interface = "No"
|
| if "business_impact" not in st.session_state: st.session_state.business_impact = ""
|
| if "alternative_tool" not in st.session_state: st.session_state.alternative_tool = ""
|
| if "expertise_risk" not in st.session_state: st.session_state.expertise_risk = ""
|
| if "training_required" not in st.session_state: st.session_state.training_required = ""
|
| if "Tform" not in st.session_state: st.session_state.Tform = 0.0
|
| if "Treel" not in st.session_state: st.session_state.Treel = 0.0
|
| if "Tia" not in st.session_state: st.session_state.Tia = 0.0
|
| if "Tvv" not in st.session_state: st.session_state.Tvv = 0.0
|
| if "Tvvia" not in st.session_state: st.session_state.Tvvia = 0.0
|
| if "data_classification" not in st.session_state: st.session_state.data_classification = ""
|
| if "nogo" not in st.session_state: st.session_state.nogo = False
|
| if "ai_emission" not in st.session_state: st.session_state.ai_emission = 0.0
|
| if "price_per_year" not in st.session_state: st.session_state.price_per_year = 0.0
|
|
|
|
|
| def next_step():
|
| st.session_state.step += 1
|
|
|
|
|
| def get_emission_visual(emission_g):
|
| emission_kg = emission_g / 1000.0
|
|
|
| if emission_kg < 1:
|
| text = "**Less than 1 kg CO₂eq**: Equivalent to the emissions of one passenger flying ~10 km in an A320neo (40 seconds)."
|
| elif emission_kg < 5:
|
| text = "**1 to 5 kg CO₂eq**: Equivalent to a standard car commute from Toulouse city center to the Airbus Blagnac campus."
|
| elif emission_kg < 10:
|
| text = "**5 to 10 kg CO₂eq**: Equivalent to the emissions of one passenger flying a very short route (~100 km)."
|
| elif emission_kg < 15:
|
| text = "**10 to 15 kg CO₂eq**: Equivalent to the manufacturing of a standard A320 passenger cabin window."
|
| elif emission_kg < 20:
|
| text = "**15 to 20 kg CO₂eq**: Equivalent to the combustion of about 5 to 6 liters of aviation fuel (Jet A-1)."
|
| elif emission_kg < 25:
|
| text = "**20 to 25 kg CO₂eq**: Equivalent to running an A320's APU on the ground for about 15 minutes."
|
| elif emission_kg < 30:
|
| text = "**25 to 30 kg CO₂eq**: Equivalent to the emissions of one passenger on a short regional flight (Toulouse to Lyon)."
|
| elif emission_kg < 35:
|
| text = "**30 to 35 kg CO₂eq**: Equivalent to the emissions of an A320neo operating at full takeoff thrust for just 5 seconds."
|
| elif emission_kg < 40:
|
| text = "**35 to 40 kg CO₂eq**: Equivalent to the fuel emissions of an Airbus Beluga taxiing on the tarmac for a few minutes."
|
| elif emission_kg < 45:
|
| text = "**40 to 45 kg CO₂eq**: Equivalent to the emissions of an A320 family aircraft running its engines at idle for about 90 seconds."
|
| elif emission_kg <= 50:
|
| text = "**45 to 50 kg CO₂eq**: Equivalent to the emissions of one passenger flying from Toulouse to Paris (entire flight)."
|
| else:
|
| text = "🛑 **More than 50 kg CO₂eq**: Exceeds a one-way flight from Toulouse to Paris per passenger. This is a massive impact and potentially a NO-GO for a single background IT tool."
|
|
|
| pos = min((emission_kg / 50.0) * 100, 100)
|
|
|
| html = f"""
|
| <div style="width: 100%; margin-top: 10px; margin-bottom: 20px; padding: 20px; background-color: #f0f4f8; border-radius: 10px; border-left: 5px solid #00205b;">
|
| <p style="font-weight: 600; margin-bottom: 15px; font-size: 1.1rem; color: #00205b;">Environmental Impact Scale (0 to 50+ kg) :</p>
|
| <div style="position: relative; width: 100%; height: 20px; background: linear-gradient(to right, #4CAF50 0%, #8BC34A 20%, #FFEB3B 40%, #FF9800 60%, #F44336 80%, #b71c1c 100%); border-radius: 10px; box-shadow: inset 0 1px 3px rgba(0,0,0,0.2);">
|
| </div>
|
| <div style="position: relative; width: 100%; height: 25px;">
|
| <div style="position: absolute; left: {pos}%; top: -25px; transform: translateX(-50%); font-size: 24px; text-shadow: 0px 2px 4px rgba(0,0,0,0.3);">
|
| 📍
|
| </div>
|
| </div>
|
| <p style="margin-top: 10px; font-size: 1.05rem; color: #333; line-height: 1.5;">
|
| {text}
|
| </p>
|
| </div>
|
| """
|
| return html
|
|
|
| if st.session_state.step > 0:
|
| st.title("AI Tool Integration Evaluator")
|
|
|
| if st.session_state.step == 0:
|
| st.title("Welcome to the AI Tool Integration Evaluator")
|
| st.markdown(
|
| "### This tool aims to evaluate the environmental, financial, and operational impacts of an AI-driven tool.")
|
| st.write(
|
| "While Artificial Intelligence is a highly effective technology capable of driving major innovations, "
|
| "it also represents a significant environmental footprint and operational cost. "
|
| "Before investing resources into development, it is crucial to evaluate whether its integration is truly indispensable."
|
| )
|
| st.write(
|
| "This questionnaire will guide you step-by-step to assess the viability and true cost of your AI use case.")
|
| st.divider()
|
| if st.button("Start Questionnaire"):
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 1:
|
| st.subheader("Question 1")
|
| st.write("**How many users will use the tool?**")
|
| with st.form(key="form_q1"):
|
| user_input = st.number_input("Number of users", min_value=1, value=1, step=1, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.N = user_input
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 2:
|
| st.subheader("Question 2")
|
| st.write("**What will be the use frequency?**")
|
| freq_values = {"Daily": 200, "Weekly": 50, "Monthly": 10, "Annual": 2}
|
| with st.form(key="form_q2"):
|
| choice = st.radio("Select frequency:", options=list(freq_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.U = freq_values[choice]
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 3:
|
| st.subheader("Question 3")
|
| st.write("**Which AI model will be integrated?**")
|
| model_values = {
|
| "Light model - SLM (Llama3, Phi-3)": {"FEin": 0.005 / 1000, "FEout": 0.015 / 1000},
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": {"FEin": 0.03 / 1000, "FEout": 0.08 / 1000},
|
| "Large model - LLM (GPT4, Gemini Pro)": {"FEin": 0.15 / 1000, "FEout": 0.45 / 1000}
|
| }
|
| with st.form(key="form_q3"):
|
| choice = st.radio("Select model:", options=list(model_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.model_choice = choice
|
| st.session_state.FEin = model_values[choice]["FEin"]
|
| st.session_state.FEout = model_values[choice]["FEout"]
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 4:
|
| st.subheader("Question 4")
|
| st.write("**What type of data will be provided to the AI?**")
|
| tin_values = {
|
| "Question (sentence)": 50, "Picture": 765, "Document (less than 15 pages)": 2000,
|
| "Document (more than 15 pages)": 15000, "Video": 30000
|
| }
|
| with st.form(key="form_q4"):
|
| choice = st.radio("Select data type:", options=list(tin_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tin = tin_values[choice]
|
| st.session_state.Tin_type = choice
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 5:
|
| st.subheader("Question 5")
|
| st.write("**What type of data will be generated by the AI?**")
|
| tout_values = {
|
| "Sentence (10-20 words)": 50, "text": 350, "function (code)": 600,
|
| "document (less than 15 pages)": 1500, "document (more than 15 pages)": 20000,
|
| "picture": 4000, "video": 50000
|
| }
|
| with st.form(key="form_q5"):
|
| choice = st.radio("Select generated data type:", options=list(tout_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tout = tout_values[choice]
|
| st.session_state.Tout_type = choice
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 6:
|
| st.subheader("Question 6")
|
| st.write("**How many AI calls will be made each time the tool is used?**")
|
| with st.form(key="form_q6"):
|
| user_input = st.number_input("Number of AI calls", min_value=1, value=1, step=1, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.R = user_input
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 7:
|
| st.subheader("Question 7")
|
| st.write("**How will the AI call be triggered?**")
|
| f_values = {"User on-demand": 1, "Automatic / Background": 5}
|
| with st.form(key="form_q7"):
|
| choice = st.radio("Select trigger type:", options=list(f_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.F = f_values[choice]
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 8:
|
| st.subheader("Question 8")
|
| st.write("**How will the tool be coded?**")
|
| g_values = {"Self-competencies or GenAI documentation": 1, "Gemini support": 10}
|
| with st.form(key="form_q8"):
|
| choice = st.radio("Select coding method:", options=list(g_values.keys()), label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.G = g_values[choice]
|
| next_step()
|
| st.rerun()
|
|
|
| elif st.session_state.step == 9:
|
| st.subheader("Question 9")
|
| st.write("**How will you resolve code failures?**")
|
| with st.form(key="form_q9"):
|
| choice = st.radio("Select resolution method:", options=["Other tools", "With AI"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| if choice == "Other tools":
|
| st.session_state.testing = 0
|
| st.session_state.dev = 0
|
| st.session_state.step = 12
|
| else:
|
| st.session_state.step = 10
|
| st.rerun()
|
|
|
| elif st.session_state.step == 10:
|
| st.subheader("Question 10")
|
| st.write("**Is AI generation required for every test of the tool?**")
|
| with st.form(key="form_q10"):
|
| choice = st.radio("Select:", options=["NO", "YES"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| if choice == "NO":
|
| st.session_state.testing = 0
|
| st.session_state.dev = 0
|
| st.session_state.step = 12
|
| else:
|
| st.session_state.step = 11
|
| st.rerun()
|
|
|
| elif st.session_state.step == 11:
|
| st.subheader("Question 11")
|
| st.write("**What is the forecasted volume of requests per year (for development and testing phases)?**")
|
| with st.form(key="form_q11"):
|
| user_input = st.number_input("Number of requests", min_value=0, value=0, step=1, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.testing = user_input
|
| st.session_state.dev = user_input
|
| st.session_state.step = 12
|
| st.rerun()
|
|
|
| elif st.session_state.step == 12:
|
| st.subheader("Question 12")
|
| st.write("**Are there other projects within the Google Cloud POD that will be used?**")
|
| with st.form(key="form_q12"):
|
| choice = st.radio("Select:", options=["No", "Yes"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| if choice == "No":
|
| st.session_state.nb_projects_pod = 1
|
| st.session_state.step = 14
|
| else:
|
| st.session_state.step = 13
|
| st.rerun()
|
|
|
| elif st.session_state.step == 13:
|
| st.subheader("Question 13")
|
| st.write("**How many projects are there within the POD?**")
|
| with st.form(key="form_q13"):
|
| user_input = st.number_input("Total projects", min_value=1, value=2, step=1, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.nb_projects_pod = user_input
|
| st.session_state.step = 14
|
| st.rerun()
|
|
|
| elif st.session_state.step == 14:
|
| st.subheader("Question 14")
|
| st.write("**Will the tool have a dedicated web interface that runs continuously?**")
|
| st.write("*(e.g., A custom dashboard or web app deployed on Cloud Run).*")
|
| with st.form(key="form_q14"):
|
| choice = st.radio("Select:", options=["Yes", "No"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.needs_web_interface = choice
|
| st.session_state.step = 15
|
| st.rerun()
|
|
|
| elif st.session_state.step == 15:
|
| st.subheader("Question 15")
|
| st.write("**What is the business impact if this tool is not developed?**")
|
| options_impact = [
|
| "Slight (AI used for Document formatting, translate, generation of data not linked with the business)",
|
| "Significant (Major improvements, KPIs enhancement, time saving...)"
|
| ]
|
| with st.form(key="form_q15"):
|
| choice = st.radio("Select business impact:", options=options_impact, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.business_impact = choice
|
| if choice == options_impact[0]:
|
| st.session_state.step = 17
|
| else:
|
| st.session_state.step = 16
|
| st.rerun()
|
|
|
| elif st.session_state.step == 16:
|
| st.subheader("Question 16")
|
| st.write("**Is there a tool other than AI capable of performing the task?**")
|
| with st.form(key="form_q16"):
|
| choice = st.radio("Select:", options=["Yes", "No"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.alternative_tool = choice
|
| if choice == "Yes":
|
| st.session_state.step = 17
|
| else:
|
| st.session_state.step = 18
|
| st.rerun()
|
|
|
| elif st.session_state.step == 17:
|
| st.header("🛑 NO-GO AI (Info)")
|
| st.info(
|
| "Based on your answers, AI might not be the recommended solution for this specific step. You will skip the remaining feasibility questions.")
|
| st.divider()
|
| if st.button("Continue (Skip next questions)"):
|
| st.session_state.nogo = True
|
| st.session_state.step = 27
|
| st.rerun()
|
|
|
| elif st.session_state.step == 18:
|
| st.subheader("Question 17")
|
| st.write("**Does automation risk causing teams to lose their manual expertise?**")
|
| with st.form(key="form_q18"):
|
| choice = st.radio("Select:", options=["Yes", "No"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.expertise_risk = choice
|
| if choice == "Yes":
|
| st.session_state.step = 17
|
| else:
|
| st.session_state.step = 19
|
| st.rerun()
|
|
|
| elif st.session_state.step == 19:
|
| st.subheader("Question 18")
|
| st.write("**Is training required before using the tool?**")
|
| with st.form(key="form_q19"):
|
| choice = st.radio("Select:", options=["Yes", "No"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.training_required = choice
|
| if choice == "No":
|
| st.session_state.Tform = 0.0
|
| st.session_state.step = 21
|
| else:
|
| st.session_state.step = 20
|
| st.rerun()
|
|
|
| elif st.session_state.step == 20:
|
| st.subheader("Question 19")
|
| st.write("**What is the expected training duration for an individual (in hours)?**")
|
| with st.form(key="form_q20"):
|
| user_input = st.number_input("Training duration (hours)", min_value=0.0, value=1.0, step=0.5,
|
| label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tform = user_input
|
| st.session_state.step = 21
|
| st.rerun()
|
|
|
| elif st.session_state.step == 21:
|
| st.subheader("Question 20")
|
| st.write("**How many hours does it currently take to perform this task manually (without AI)?**")
|
| with st.form(key="form_q21"):
|
| user_input = st.number_input("Current time (hours)", min_value=0.0, value=1.0, step=0.5,
|
| label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Treel = user_input
|
| st.session_state.step = 22
|
| st.rerun()
|
|
|
| elif st.session_state.step == 22:
|
| st.subheader("Question 21")
|
| st.write("**What is the estimated time (in hours) to complete this task using the AI tool?**")
|
| with st.form(key="form_q22"):
|
| user_input = st.number_input("Time with AI (hours)", min_value=0.0, value=0.5, step=0.5,
|
| label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tia = user_input
|
| st.session_state.step = 23
|
| st.rerun()
|
|
|
| elif st.session_state.step == 23:
|
| st.subheader("Question 22")
|
| st.write("**Are the AI-generated data intended to be certified and thus go through a V&V process?**")
|
| with st.form(key="form_q23"):
|
| choice = st.radio("Select:", options=["Yes", "No"], label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| if choice == "Yes":
|
| st.session_state.step = 24
|
| else:
|
| st.session_state.Tvv = 0.0
|
| st.session_state.Tvvia = 0.0
|
| st.session_state.step = 26
|
| st.rerun()
|
|
|
| elif st.session_state.step == 24:
|
| st.subheader("Question 23")
|
| st.write("**What is the current time required for the V&V process (in hours)?**")
|
| with st.form(key="form_q24"):
|
| user_input = st.number_input("Current V&V time (hours)", min_value=0.0, value=1.0, step=0.5,
|
| label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tvv = user_input
|
| st.session_state.step = 25
|
| st.rerun()
|
|
|
| elif st.session_state.step == 25:
|
| st.subheader("Question 24")
|
| st.write("**What is the expected time for the V&V process for AI-generated data (in hours)?**")
|
| with st.form(key="form_q25"):
|
| user_input = st.number_input("V&V time for AI data (hours)", min_value=0.0, value=1.0, step=0.5,
|
| label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.Tvvia = user_input
|
| st.session_state.step = 26
|
| st.rerun()
|
|
|
| elif st.session_state.step == 26:
|
| st.subheader("Question 25")
|
| st.write("**What is the classification of the input and output data?**")
|
| options_class = ["Airbus Amber / without any classification", "Airbus RED"]
|
| with st.form(key="form_q26"):
|
| choice = st.radio("Select data classification:", options=options_class, label_visibility="collapsed")
|
| if st.form_submit_button("Next"):
|
| st.session_state.data_classification = choice
|
| if choice == "Airbus RED":
|
| st.session_state.step = 17
|
| else:
|
| st.session_state.step = 27
|
| st.rerun()
|
|
|
| elif st.session_state.step == 27:
|
|
|
| model = st.session_state.model_choice
|
| Tin = st.session_state.Tin
|
| Tout = st.session_state.Tout
|
| Tin_type = st.session_state.Tin_type
|
| Tout_type = st.session_state.Tout_type
|
| N = st.session_state.N
|
| U = st.session_state.U
|
| R = st.session_state.R
|
| F = st.session_state.F
|
| dev = st.session_state.dev
|
| testing = st.session_state.testing
|
| FEin = st.session_state.FEin
|
| FEout = st.session_state.FEout
|
| G = st.session_state.G
|
| nb_projects_pod = max(st.session_state.nb_projects_pod, 1)
|
|
|
| Pin = 0.0
|
| Pout = 0.0
|
| if model == "Light model - SLM (Llama3, Phi-3)":
|
| Pin = 0.3 * Tin / 1_000_000
|
| Pout = 0.7 * Tout / 1_000_000
|
| elif model == "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)":
|
| Pin = 1.5 * Tin / 1_000_000
|
| Pout = 7.5 * Tout / 1_000_000
|
| elif model == "Large model - LLM (GPT4, Gemini Pro)":
|
| Pin = 2.0 * Tin / 1_000_000 if Tin <= 200000 else 4.0 * Tin / 1_000_000
|
| Pout = 12.0 * Tout / 1_000_000 if Tout <= 200000 else 18.0 * Tout / 1_000_000
|
|
|
| execution_times = {
|
| "Sentence (10-20 words)": {"Light model - SLM (Llama3, Phi-3)": 0.80,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 0.5,
|
| "Large model - LLM (GPT4, Gemini Pro)": 2},
|
| "text": {"Light model - SLM (Llama3, Phi-3)": 2.50,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 1.5,
|
| "Large model - LLM (GPT4, Gemini Pro)": 5},
|
| "function (code)": {"Light model - SLM (Llama3, Phi-3)": 5.50,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 3.0,
|
| "Large model - LLM (GPT4, Gemini Pro)": 11},
|
| "document (less than 15 pages)": {"Light model - SLM (Llama3, Phi-3)": 80.0,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 45.0,
|
| "Large model - LLM (GPT4, Gemini Pro)": 160},
|
| "document (more than 15 pages)": {"Light model - SLM (Llama3, Phi-3)": 150.0,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 85.0,
|
| "Large model - LLM (GPT4, Gemini Pro)": 300},
|
| "picture": {"Light model - SLM (Llama3, Phi-3)": 2.0,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 10.0,
|
| "Large model - LLM (GPT4, Gemini Pro)": 15},
|
| "video": {"Light model - SLM (Llama3, Phi-3)": 120.0,
|
| "Medium model (Claude Sonnet, Gemini Flash, Mistral 8x7b)": 160.0,
|
| "Large model - LLM (GPT4, Gemini Pro)": 180}
|
| }
|
|
|
| if Tout_type and model:
|
| st.session_state.time_exec = execution_times[Tout_type][model]
|
| st.session_state.Prun = (0.0000265 * st.session_state.time_exec) + (0.40 / 1_000_000)
|
|
|
| total_calls_year = (N * U * R * F) + G + testing
|
|
|
| inference_emission = total_calls_year * ((Tin * FEin) + (Tout * FEout))
|
| inference_cost = total_calls_year * (st.session_state.Prun + Pin + Pout)
|
|
|
| data_weights = {
|
| "Question (sentence)": 0.001, "Sentence (10-20 words)": 0.001, "text": 0.01, "function (code)": 0.01,
|
| "Picture": 2.0, "picture": 2.0, "Document (less than 15 pages)": 1.0, "document (less than 15 pages)": 1.0,
|
| "Document (more than 15 pages)": 5.0, "document (more than 15 pages)": 5.0, "Video": 50.0, "video": 50.0
|
| }
|
| size_in_mb = data_weights.get(Tin_type, 0.1)
|
| size_out_mb = data_weights.get(Tout_type, 0.1)
|
|
|
| total_data_gb_year = (total_calls_year * (size_in_mb + size_out_mb)) / 1000.0
|
|
|
| network_cost = total_data_gb_year * 0.12
|
| network_emission = total_data_gb_year * 50.0
|
|
|
| infra_cost = 0.0
|
| infra_emission = 0.0
|
|
|
| infra_cost += (132.0 / nb_projects_pod)
|
| infra_emission += (5500.0 / nb_projects_pod)
|
|
|
| if st.session_state.needs_web_interface == "Yes":
|
| infra_cost += 350.0
|
| infra_emission += 12000.0
|
|
|
| st.session_state.ai_emission = inference_emission + network_emission + infra_emission
|
| st.session_state.price_per_year = inference_cost + network_cost + infra_cost
|
|
|
| annual_time_savings = ((st.session_state.Treel - st.session_state.Tia) + (
|
| st.session_state.Tvv - st.session_state.Tvvia) - st.session_state.Tform) * st.session_state.U
|
| tool_implementation_time = st.session_state.Tform + st.session_state.Tia + st.session_state.Tvvia
|
|
|
| st.header("Final Evaluation Results")
|
|
|
| if not st.session_state.nogo:
|
| st.success("** Go AI but keep in mind...**")
|
| else:
|
| st.warning(
|
| "**Note**: You bypassed some steps due to a NO-GO AI condition. Here are your partial/simulated metrics.")
|
|
|
| col1, col2 = st.columns(2)
|
|
|
| with col1:
|
| st.metric(label="Price per year", value=f"€ {st.session_state.price_per_year:,.2f}")
|
| st.metric(label="Annual Time Savings", value=f"{annual_time_savings:,.2f} h")
|
|
|
| with col2:
|
| st.metric(label="AI Emission", value=f"{st.session_state.ai_emission:,.2f} gCO₂eq")
|
| st.metric(label="Tool Implementation Time", value=f"{tool_implementation_time:,.2f} h")
|
|
|
| st.markdown(get_emission_visual(st.session_state.ai_emission), unsafe_allow_html=True)
|
|
|
| st.divider()
|
|
|
| if st.button("Start Over"):
|
| for key in list(st.session_state.keys()):
|
| del st.session_state[key]
|
| st.rerun() |
