Update app.py

app.py

@@ -18,6 +18,8 @@ from google.analytics.data_v1beta.types import (
     Filter
 )
 
+from openai import OpenAI
+
 warnings.filterwarnings("ignore", category=FutureWarning)
 
 # =====================================================
@@ -25,7 +27,10 @@ warnings.filterwarnings("ignore", category=FutureWarning)
 # =====================================================
 GA4_PROPERTY_ID = os.environ["GA4_PROPERTY_ID"]
 SERVICE_ACCOUNT_JSON = json.loads(os.environ["SERVICE_ACCOUNT_JSON"])
-SITE_DOMAIN = "rishabhsoft.com"   # change if required
+OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
+SITE_DOMAIN = "rishabhsoft.com"
+
+client = OpenAI(api_key=OPENAI_API_KEY)
 
 # =====================================================
 # GA CLIENT
@@ -55,7 +60,7 @@ def classify_source(src):
     return "External"
 
 # =====================================================
-# CACHED TOP PAGES
+# TOP PAGES (FAST)
 # =====================================================
 @lru_cache(maxsize=3)
 def fetch_top_pages(start, end):
@@ -99,7 +104,7 @@ def fetch_journey(start, end, landing_page):
     return pd.DataFrame(rows)
 
 # =====================================================
-# FORM SUBMISSIONS (GA4-SAFE PROXY)
+# FORM SUBMISSIONS (GA4-SAFE)
 # =====================================================
 def get_form_submissions(start, end, landing_page):
     req = RunReportRequest(
@@ -134,21 +139,74 @@ def get_form_submissions(start, end, landing_page):
     res = ga.run_report(req)
     return sum(int(r.metric_values[0].value) for r in res.rows)
 
+# =====================================================
+# LLM INTENT CLASSIFIER (CACHED)
+# =====================================================
+@lru_cache(maxsize=100)
+def classify_intent_llm(start_page, pages, converted):
+    prompt = f"""
+You are a digital analytics expert.
+
+Classify the user intent into ONE of:
+- Conversion
+- Evaluation
+- Research
+- Navigation
+- Low Intent
+
+Rules:
+- Use only the data given
+- Do NOT invent behavior
+- Be concise
+
+Journey Data:
+Start Page: {start_page}
+Pages Visited: {", ".join(pages[:5])}
+Form Submitted: {"Yes" if converted else "No"}
+
+Return format:
+Intent: <Intent>
+Reason: <One sentence explanation>
+"""
+
+    resp = client.chat.completions.create(
+        model="gpt-4o-mini",
+        messages=[{"role": "user", "content": prompt}],
+        temperature=0
+    )
+
+    txt = resp.choices[0].message.content.strip()
+    intent = "Unknown"
+    reason = ""
+
+    for line in txt.splitlines():
+        if line.lower().startswith("intent"):
+            intent = line.split(":", 1)[1].strip()
+        if line.lower().startswith("reason"):
+            reason = line.split(":", 1)[1].strip()
+
+    return intent, reason
+
 # =====================================================
 # MAIN LOGIC
 # =====================================================
 def run_journey(start, end, landing_page):
     df = fetch_journey(start, end, landing_page)
     if df.empty:
-        return None, pd.DataFrame(), "No data found."
+        return None, pd.DataFrame(), "No journey data found."
 
     total_users = df["Views"].sum()
     form_yes = get_form_submissions(start, end, landing_page)
+    conv_rate = round((form_yes / total_users) * 100, 2) if total_users else 0.0
     journey_converted = "Yes" if form_yes > 0 else "No"
 
-    # ===========================
-    # EXIT TABLE
-    # ===========================
+    top_pages = df.sort_values("Views", ascending=False)["To"].unique().tolist()[:5]
+    intent, intent_reason = classify_intent_llm(
+        clean_path(landing_page),
+        [clean_path(p) for p in top_pages],
+        form_yes > 0
+    )
+
     exit_df = (
         df.groupby("From")["Views"]
         .sum()
@@ -161,72 +219,32 @@ def run_journey(start, end, landing_page):
     exit_df["Source Type"] = exit_df["Exit Source / Page"].apply(classify_source)
     exit_df["Exit Source / Page"] = exit_df["Exit Source / Page"].apply(clean_path)
 
-    # 🔑 Option C logic
     exit_df["Journey Converted?"] = journey_converted
-    exit_df["Form Submitted Before Exit?"] = "No"
-
-    exit_df["Insight"] = exit_df.apply(
-        lambda r:
-            "Converted journey – exit after intent met"
-            if r["Journey Converted?"] == "Yes"
-            else (
-                "High exit – landing intent mismatch"
-                if r["Source Type"] == "External"
-                else "Conversion leak – improve CTA"
-            ),
-        axis=1
-    )
+    exit_df["Journey Conversion Rate %"] = f"{conv_rate:.2f}%"
+    exit_df["Exit Severity Score"] = (
+        exit_df["Users Exited"] * (1 - conv_rate / 100)
+    ).round(1)
+
+    exit_df["User Intent"] = intent
+    exit_df["Intent Reason"] = intent_reason
 
-    exit_df = exit_df[
-        [
-            "Exit Source / Page",
-            "Source Type",
-            "Users Entered",
-            "Users Exited",
-            "Exit Rate %",
-            "Journey Converted?",
-            "Form Submitted Before Exit?",
-            "Insight"
-        ]
-    ].sort_values("Users Exited", ascending=False)
-
-    # ===========================
-    # SANKEY
-    # ===========================
-    labels, idx = [], {}
-
-    def add(n):
-        if n not in idx:
-            idx[n] = len(labels)
-            labels.append(n)
-        return idx[n]
-
-    s, t, v = [], [], []
-
-    for _, r in df.iterrows():
-        s.append(add(clean_path(r["From"])))
-        t.append(add(clean_path(r["To"])))
-        v.append(r["Views"])
-
-    p = add(clean_path(landing_page))
-    e = add("Exit")
-    s.append(p)
-    t.append(e)
-    v.append(total_users - form_yes)
-
-    if form_yes > 0:
-        f = add("Form Submitted")
-        s.append(p)
-        t.append(f)
-        v.append(form_yes)
+    exit_df = exit_df.sort_values("Exit Severity Score", ascending=False)
 
     fig = go.Figure(go.Sankey(
-        node=dict(label=labels, pad=22, thickness=18),
-        link=dict(source=s, target=t, value=v)
+        node=dict(
+            label=["Entry", clean_path(landing_page), "Exit", "Form Submitted"],
+            pad=20,
+            thickness=18
+        ),
+        link=dict(
+            source=[0, 1, 1],
+            target=[1, 2, 3],
+            value=[total_users, total_users - form_yes, form_yes]
+        )
     ))
 
     fig.update_layout(
-        title="Customer Journey Flow (Entry → Exit / Conversion)",
+        title="Customer Journey Flow (Business View)",
         height=650
     )
 
@@ -237,7 +255,9 @@ def run_journey(start, end, landing_page):
 
 - **Journey Converted:** {journey_converted}
 - **Form Submissions:** {form_yes}
-- **Exited Without Form:** {total_users - form_yes}
+- **Conversion Rate:** {conv_rate:.2f}%
+- **Primary User Intent:** **{intent}**
+- **Why:** {intent_reason}
 """
 
     return fig, exit_df, summary
@@ -246,7 +266,7 @@ def run_journey(start, end, landing_page):
 # UI
 # =====================================================
 with gr.Blocks() as app:
-    gr.Markdown("## 🚀 Customer Journey Explorer (GA4 – Business View)")
+    gr.Markdown("## 🚀 Customer Journey Intelligence (GA4 + AI)")
 
     with gr.Row():
         sd = gr.Textbox(value=str(date.today() - timedelta(days=30)), label="Start Date")