Major approach change. Rewrite app.py, ajust requirements and update .gitignore.

.gitignore

@@ -1 +1 @@
-testing_files/
+data_source/

app.py

@@ -1,91 +1,80 @@
 import pandas as pd
 import gradio as gr
-from langchain.agents import initialize_agent, Tool
-from langchain_google_genai import ChatGoogleGenerativeAI  # requires GOOGLE_API_KEY set as env var
-
-# Store the uploaded dataframe in memory
-dataframes = {}
-
-# -----------------------------
-# Tools (Excel Q&A Functions)
-# -----------------------------
-def get_otb_revenue(month: str) -> str:
-    if "df" not in dataframes:
-        return "Please upload a file first."
-    df = dataframes["df"]
-    try:
-        otb = df.loc[df['Month'].str.lower() == month.lower(), 'OTB Revenue'].values[0]
-        stly = df.loc[df['Month'].str.lower() == month.lower(), 'STLY Revenue'].values[0]
-        return f"OTB Revenue for {month}: {otb}, STLY Revenue: {stly}"
-    except Exception:
-        return f"Could not find OTB/ STLY Revenue for {month}"
-
-def check_occupancy(month: str) -> str:
-    if "df" not in dataframes:
-        return "Please upload a file first."
-    df = dataframes["df"]
-    try:
-        occ = df.loc[df['Month'].str.lower() == month.lower(), 'Occupancy %'].values[0]
-        target = df.loc[df['Month'].str.lower() == month.lower(), 'Budgeted Occupancy %'].values[0]
-        if occ >= target:
-            return f"Occupancy for {month} is {occ}% vs target {target}% ✅ On track!"
-        else:
-            return f"Occupancy for {month} is {occ}% vs target {target}% ⚠️ Might not meet target."
-    except Exception:
-        return f"Could not find occupancy data for {month}"
-
-# -----------------------------
-# Upload Excel
-# -----------------------------
-def upload_excel(file):
-    if file is None:
-        return "Please upload an Excel file."
-    dfs = pd.read_excel(file.name, sheet_name=None)
-    dataframes.clear()
-    dataframes.update(dfs)
-    sheet_info = []
-    for name, df in dfs.items():
-        sheet_info.append(f"'{name}': {len(df)} rows, {len(df.columns)} cols")
-    return f"✅ Loaded sheets: {', '.join(sheet_info)}"
-
-# -----------------------------
-# LangChain Agent Setup
-# -----------------------------
-tools = [
-    Tool(
-        name="Get OTB Revenue",
-        func=get_otb_revenue,
-        description="Get OTB Revenue and STLY Revenue for a given month (e.g. 'August')"
-    ),
-    Tool(
-        name="Check Occupancy",
-        func=check_occupancy,
-        description="Check occupancy vs target for a given month (e.g. 'August')"
-    )
-]
-
-llm = ChatGoogleGenerativeAI(model="gemini-2.5-pro", temperature=0)  # Requires GOOGLE_API_KEY in environment
-agent = initialize_agent(tools, llm, agent="zero-shot-react-description", verbose=True)
-
-def chat_agent(message, history):
-    if "df" not in dataframes:
-        return "Please upload a file first."
+import os
+import google.generativeai as genai  # requires GOOGLE_API_KEY set as env var
+
+# 1. Configure Gemini
+genai.configure(api_key=os.getenv("GEMINI_API_KEY"))
+model = genai.GenerativeModel("gemini-1.5-pro")  # or "gemini-pro"
+
+# 2. Load Excel data
+df1 = pd.read_excel("report1.xlsx")
+df2 = pd.read_excel("report2.xlsx")
+
+# Build schema info for prompts
+def get_schema_info():
+    schema1 = f"Report1 columns: {list(df1.columns)}"
+    schema2 = f"Report2 columns: {list(df2.columns)}"
+    return schema1 + "\n" + schema2
+
+schema_info = get_schema_info()
+
+# 3. Core function
+def answer_question(history, message):
+    """
+    history: chat history (list of [user, assistant] pairs)
+    message: latest user message (string)
+    """
+    # Build prompt for Gemini
+    prompt = f"""
+You are a data analysis assistant. 
+You can ONLY answer questions using the two Excel reports provided. 
+Do not hallucinate or use external knowledge.
+If the question is irrelevant, respond with:
+"I can only answer questions about the provided Excel reports."
+
+The reports have the following schema:
+{schema_info}
+
+The user asked:
+{message}
+
+Generate Python pandas code that uses df1 and/or df2 to answer.
+Return ONLY code, nothing else.
+    """
+
     try:
-        return agent.run(message)
+        # Call Gemini
+        response = model.generate_content(prompt)
+        code = response.text.strip("```python").strip("```")
+
+        # 4. Execute code safely
+        local_vars = {"df1": df1, "df2": df2, "pd": pd}
+        try:
+            result = eval(code, {"__builtins__": {}}, local_vars)
+        except Exception as e:
+            exec(code, {"__builtins__": {}}, local_vars)
+            result = local_vars.get("result", "No result variable found")
+
+        return str(result)
+
     except Exception as e:
-        return f"⚠️ Agent error: {e}"
+        return f"Error: {str(e)}"
 
-# -----------------------------
-# Gradio UI
-# -----------------------------
+
+# 5. Gradio UI
 with gr.Blocks() as demo:
-    gr.Markdown("# 🤖 Excel Agent (LangChain + Gemini)")
-    with gr.Row():
-        with gr.Column():
-            file_input = gr.File(label="Upload Excel file", type="filepath")
-            upload_output = gr.Textbox(label="Upload status")
-        file_input.upload(upload_excel, file_input, upload_output)
+    chatbot = gr.Chatbot()
+    msg = gr.Textbox(placeholder="Ask me a question about the reports...")
+    clear = gr.ClearButton([msg, chatbot])
+
+    def respond(message, chat_history):
+        answer = answer_question(chat_history, message)
+        chat_history.append((message, answer))
+        return "", chat_history
 
-    gr.ChatInterface(fn=chat_agent, title="Excel Chat Agent")
+    msg.submit(respond, [msg, chatbot], [msg, chatbot])
 
-demo.launch()
+# 6. Run locally (Spaces will call demo.launch() automatically)
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -1,5 +1,4 @@
 gradio==4.44.1
 pandas
 openpyxl
-langchain
-langchain-google-genai
+google-generativeai