slab_sandbox_app_noKey.py

import streamlit as st
import pandas as pd
import openai

st.set_page_config(layout="wide")

if "df_slab" not in st.session_state:
    file_path = "SLAB_Data.xlsx"
    df_slab = pd.read_excel(file_path)

    # Clean and standardize the data
    df_slab["Year"] = df_slab["Year"].astype(str).str.replace(".0", "", regex=False)
    df_slab = df_slab.applymap(lambda x: "-" if str(x).strip().lower() == "missing" else x)

    df_slab.insert(0, "Row Number", range(1, len(df_slab) + 1))
    df_slab.insert(1, "Select", False)

    df_slab.rename(columns={"Description": "Description - loaded from file"}, inplace=True)

    st.session_state.df_slab = df_slab

df_slab = st.session_state.df_slab

# Dropdown Options
grading_companies = ["", "PSA", "BGS", "SGC", "CSG", "HGA", "GMA", "ISA", "KSA", "RAW"]
grades = [""] + [str(x) for x in [1.0, 1.5] + [i for i in range(2, 11)]]
rookie_options = ["", "Yes", "No"]
card_sets = [
    "Court Kings", "Crusade", "Elite", "Flawless", "Gold Standard", "Hoops",
    "Immaculate Collection", "Innovation", "Intrigue", "National Treasures",
    "Pinnacle", "Preferred", "Prestige", "Prizm", "Select", "Signatures",
    "Spectra", "Timeless Treasures", "Upper Deck Black", "Titanium", "Totally Certified"
]
parallels = [
    "Purple Prizm", "Gold", "Prizm", "Silver Prizm", "Red", "Silver", "Blue",
    "Red Blue White Prizm", "Orange Die-Cut", "Blue Green", "Blue Prizm",
    "Purple", "Purple Pulsar", "Red Blue White Pulsar Prizm", "Die-Cut",
    "Black", "BLACK", "White", "Green", "Gold White", "Gold Prizm", "Green Black",
    "Red Blue White Mosaic Prizm", "Green Wave", "Blue White Prizm",
    "Red Blue Prizm", "Green Prizm", "Orange Prizm", "Red Prizm",
    "Silver Prizm Refractor", "Wave", "Blue Die-Cut Prizm", "Cracked Ice Prizm",
    "Purple Die-Cut Prizm", "Silver Green Prizm", "Gold Black", "Orange",
    "Gold Green", "Purple White", "Red Purple Prizm", "Gold Die-Cut", "Lustrous Rookies Signature",
    "Red Green Prizm", "Blue Gold Prizm", "Red Gold Prizm", "Silver White Prizm",
    "Red Silver Prizm", "Green Prizm Refractor", "Prizm Refractor", "Holo",
    "Gold Orange Prizm", "Silver Prizm Refractor Holo", "Purple White Prizm",
    "Purple Orange Prizm", "Atomic", "Green White Prizm", "Red Green",
    "Cracked Ice", "Blue Prizm Refractor", "Red Blue Mosaic Prizm", "Red Die-Cut",
    "2-on-2 Quad Memorabilia Platinum", "2-on-2 Quad Prime Memorabilia",
    "Art Nouveau Jersey Nameplates", "Art Nouveau Jersey Patches",
    "Art Nouveau Jerseys", "Art Nouveau Prime Jerseys", "Base Set Gold", "NBA Game Gear Duals", "Hall Monitors", "Brilliance", "Memorabilia Red",
    "Totally Red", "Hall of Fame Heroes", "Red Back", "Base", "Choice Award",
    "Rookie", "Purple Select Swatches", "Status", "Portraits", "Dominance", "Signatures"
]

st.title("SLAB Protocol™ AI Builder – Sandbox Demo")
st.markdown("#### A trading card standard powered by AI")

tabs = st.tabs(["🔄 Load Data File/Run Through Model", "📝 Edit AI Generated Fields", "📊 Preview SLAB Generated Data", "📄 Preview SLAB Generated XML", "🔍 SLAB Insights"])

with tabs[0]:
    st.markdown("### Processing Selected SLAB_Data.xlsx through Engine")
    st.dataframe(df_slab[["Description - loaded from file"]], use_container_width=True)

with tabs[1]:
    st.markdown(
        """
        ### 📝 Edit SLAB Fields Workspace
        This workspace allows you to **review and refine the fields generated by the AI model** from the card description.  
        **Every change you make contributes to improving model accuracy over time.**
        """
    )

    search_query = st.text_input("Search Description")

    st.markdown("#### Bulk Update Selected Rows")
    col1, col2 = st.columns([1, 3])

    with col1:
        field_to_update = st.selectbox(
            "Field Drop Down",
            df_slab.columns[3:],
            key="bulk_field_select"
        )

    with col2:
        bulk_value = st.text_input("Value for Bulk Update", key="bulk_value_generic")

    master_checkbox = st.checkbox("Select All Visible Rows")

    if search_query:
        filtered_df = df_slab[df_slab["Description - loaded from file"].str.contains(search_query, case=False, na=False)]
    else:
        filtered_df = df_slab.copy()

    if master_checkbox:
        filtered_df.loc[:, "Select"] = True
        df_slab.loc[filtered_df.index, "Select"] = True

    column_config = {col: st.column_config.TextColumn() for col in df_slab.columns[3:]}

    edited_df = st.data_editor(
        filtered_df,
        use_container_width=True,
        num_rows="fixed",
        hide_index=True,
        column_config=column_config
    )

    for idx, row in edited_df.iterrows():
        st.session_state.df_slab.loc[idx] = row

    if st.button("Apply Bulk Update"):
        selected_rows = df_slab[df_slab["Select"].astype(bool)].index.tolist()

        if bulk_value and selected_rows:
            for idx in selected_rows:
                df_slab.at[idx, field_to_update] = bulk_value

        df_slab.loc[selected_rows, "Select"] = False
        st.session_state.df_slab = df_slab.copy()
        st.rerun()

with tabs[2]:
    st.markdown("### 📊 Preview SLAB Generated Data")
    st.dataframe(
        st.session_state.df_slab.drop(columns=["Select"]),
        use_container_width=True
    )
with tabs[3]:
    for _, row in st.session_state.df_slab.iterrows():
        st.code(f"""
<Card>
  <Year>{row['Year']}</Year>
  <PlayerName>{row['Player Name']}</PlayerName>
  <CardNumber>{row['Card Number']}</CardNumber>
  <Set>{row['Card Set']}</Set>
  <Parallel>{row['Parallel']}</Parallel>
  <GradingCompany>{row['Grading Company']}</GradingCompany>
  <Grade>{row['Grade']}</Grade>
  <Limited>{row['Limited']}</Limited>
  <ShortPrint>{row['Short Print']}</ShortPrint>
  <Auto>{row['Auto']}</Auto>
  <Rookie>{row['Rookie']}</Rookie>
</Card>
""", language="xml")

with tabs[4]:
    st.markdown("### 🤖 SLAB Assistant – Ask Anything About SLAB Protocol™")
    user_query = st.text_area("Paste one or more card descriptions for AI breakdown:")

    if st.button("Submit Query"):
        if user_query.strip():
            try:
                import json

                openai.api_key = st.secrets["OPENAI_API_KEY"]

                client = openai.Client(api_key=openai.api_key)

                response = client.chat.completions.create(
                    model="gpt-4",  # If you don't have GPT-4, change this to "gpt-3.5-turbo"
                    messages=[
                        {
                            "role": "system",
                            "content": """
                            You are an expert in the SLAB Protocol™ trading card standard.
                            Break down card descriptions into the following fields:
                            - Year (e.g., 2024)
                            - Player Name (e.g., Shohei Ohtani)
                            - Card Number (e.g., #S-2 or S-2)
                            - Card Set (e.g., Topps Chrome, Panini Prizm)
                            - Parallel (e.g., Silver Prizm, Rainbow Foil)
                            - Grading Company (e.g., PSA, BGS, RAW)
                            - Grade (e.g., 10, 9.5, or empty if not graded)
                            - Limited (e.g., /99, #/99, or empty if not limited)
                            - Short Print (e.g., SP, SSP, or empty if not short print)
                            - Auto (e.g., Yes if autograph, empty otherwise)
                            - Rookie (e.g., Yes if rookie card, empty otherwise)

                            Return only a valid JSON array of objects. Each object must have:
                            {"Year": "", "Player Name": "", "Card Number": "", "Card Set": "", "Parallel": "", "Grading Company": "", "Grade": "", "Limited": "", "Short Print": "", "Auto": "", "Rookie": ""}

                            If any field is unknown or not present, leave it as an empty string "".
                            DO NOT return explanations or additional text.
                            """
                        },
                        {"role": "user", "content": user_query}
                    ],
                )

                gpt_response = response.choices[0].message.content.strip()

                # Handle GPT sometimes wrapping in ```json ... ```
                if gpt_response.startswith("```json"):
                    gpt_response = gpt_response[7:-3].strip()

                try:
                    parsed_response = json.loads(gpt_response)

                    if isinstance(parsed_response, dict):
                        parsed_response = [parsed_response]

                    slab_columns = [
                        "Year", "Player Name", "Card Number", "Card Set", "Parallel",
                        "Grading Company", "Grade", "Limited", "Short Print", "Auto", "Rookie"
                    ]
                    normalized_data = [{col: item.get(col, "") for col in slab_columns} for item in parsed_response]

                    result_df = pd.DataFrame(normalized_data)
                    st.markdown("**Extracted SLAB Fields:**")
                    st.dataframe(result_df, use_container_width=True)

                except json.JSONDecodeError:
                    st.error("❌ Failed to parse AI response as JSON. GPT may have included text or formatting issues.")
                    st.markdown("Here’s the raw response:")
                    st.text(gpt_response)

            except Exception as e:
                st.error(f"❌ Error: {e}")
        else:
            st.warning("Please enter card descriptions before submitting.")