Sample data improved. Minor prompts improvements

app/ads1/budget_optimizer.py

@@ -41,7 +41,7 @@ def build_budget_optimizer_prompt(context: dict) -> str:
     name = context.get("campaign_name", "this account")
 
     return (
-        f"Write 3 to 5 bullet points of actionable Budget optimization insights for {name}.\n"
+        f"Write 3 to 5 bullet points of actionable Budget optimization methods for {name}.\n"
         "Use simple language. One insight per bullet. Start each line with '- '. No intro sentence.\n\n"
         f"Data (JSON):\n{payload}"
     )

app/ads1/keyword_inspector.py

@@ -27,13 +27,12 @@ def build_keyword_features(df: pd.DataFrame) -> pd.DataFrame:
 def build_keyword_prompt(context: dict) -> str:
     payload = json.dumps(context, indent=2, default=str)
 
-    return f"""
-        Write 3 to 5 bullet points of actionable keyword performance insights.\n"
-        "Analyze individual keywords from the data and classify them as winning, 
-        wasted spend, or scaling opportunities using CTR, cost, and conversions. Add reasoning.
+    return (
+        "Write 3 to 5 bullet points of actionable keyword performance insights.\n"
+        "Classify individual keywords as winning, wasted spend, or scaling opportunities using CTR, cost, and conversions.\n"
         "Use simple language. One insight per bullet. Start each line with '- '. No intro sentence.\n\n"
         f"Data (JSON):\n{payload}"
-        """
+    )
 # -------------------------
 # Main runner
 # -------------------------
@@ -70,4 +69,4 @@ def run_keyword_inspector(dfs: dict, campaign_name: str | None = None) -> str:
         )
 
     print("📤 [keyword_inspector] result received", flush=True)
-    return result
+    return result

app/ads1/merge.py

@@ -8,7 +8,7 @@ def merge_dfs(real_dfs: dict, sample_dfs: dict) -> dict:
 
     merged = {}
 
-    for key in real_dfs.keys():
+    for key in sorted(set(real_dfs.keys()) | set(sample_dfs.keys())):
         real_df = real_dfs.get(key)
         sample_df = sample_dfs.get(key)
 
@@ -44,4 +44,4 @@ def merge_dfs(real_dfs: dict, sample_dfs: dict) -> dict:
             sample_df["source"] = "sample"
             merged[key] = sample_df
 
-    return merged
+    return merged

app/ads1/sample_data.py

@@ -1,14 +1,13 @@
-import numpy as np
 import pandas as pd
 
-def _rand(low, high, size):
-    return np.random.randint(low, high, size)
 
-def _money(low, high, size):
-    return np.round(np.random.uniform(low, high, size), 2)
+CAMPAIGNS = [
+    "Nursery Admissions 2026",
+    "Playgroup Enrollment Campaign",
+    "Summer Camp 2026",
+    "School Tour Booking Campaign",
+]
 
-def _pick(arr, size):
-    return np.random.choice(arr, size)
 
 def generate_sample_campaigns():
     campaigns = [
@@ -16,196 +15,197 @@ def generate_sample_campaigns():
             "id": 1001,
             "name": "Nursery Admissions 2026",
             "status": "ENABLED",
-            "impressions": 120000,
-            "clicks": 8500,
-            "cost": 12000.0,
-            "ctr": 7.08,
+            "impressions": 118000,
+            "clicks": 8600,
+            "cost": 10450.0,
+            "ctr": 7.29,
             "conversions": 420,
         },
         {
             "id": 1002,
             "name": "Playgroup Enrollment Campaign",
             "status": "ENABLED",
-            "impressions": 90000,
-            "clicks": 6200,
-            "cost": 8000.0,
-            "ctr": 6.88,
-            "conversions": 310,
+            "impressions": 76000,
+            "clicks": 4300,
+            "cost": 9800.0,
+            "ctr": 5.66,
+            "conversions": 62,
         },
         {
             "id": 1003,
             "name": "Summer Camp 2026",
             "status": "ENABLED",
-            "impressions": 150000,
-            "clicks": 4000,
-            "cost": 7000.0,
-            "ctr": 2.66,
-            "conversions": 180,
+            "impressions": 132000,
+            "clicks": 5100,
+            "cost": 8600.0,
+            "ctr": 3.86,
+            "conversions": 105,
         },
         {
             "id": 1004,
             "name": "School Tour Booking Campaign",
             "status": "ENABLED",
-            "impressions": 60000,
-            "clicks": 2200,
-            "cost": 3500.0,
-            "ctr": 3.66,
-            "conversions": 140,
+            "impressions": 54000,
+            "clicks": 2300,
+            "cost": 3200.0,
+            "ctr": 4.26,
+            "conversions": 145,
         },
     ]
-
     return pd.DataFrame(campaigns)
 
-# 2. SEARCH TERMS
-def generate_sample_search_terms():
-    positive_intent = [
-        "nursery admission near me",
-        "best preschool in Yelahanka",
-        "play school admission 2026",
-        "kg admission Bangalore",
-        "preschool fees near me",
-        "daycare near Yelahanka",
-    ]
-
-    negative_intent = [
-        "free babysitting jobs",
-        "toy store near me",
-        "kids games online",
-        "montessori certification course",
-        "child psychology course",
-        "teaching jobs preschool",
-    ]
 
-    campaigns = [
-        "Nursery Admissions 2026",
-        "Playgroup Enrollment Campaign",
-        "Summer Camp 2026",
-        "School Tour Booking Campaign",
+def _search_rows_for_campaign(campaign_name, multiplier=1.0):
+    base_rows = [
+        # Clear high-intent / scale candidates
+        ("nursery admission near me", 180, 4200, 260.0, 22),
+        ("preschool fees near me", 140, 3500, 190.0, 18),
+        ("best preschool in Yelahanka", 125, 2800, 210.0, 16),
+        ("school tour booking near me", 90, 1800, 135.0, 14),
+        ("kg admission Bangalore", 155, 3900, 245.0, 15),
+        # Investigate: engagement but weak conversion
+        ("daycare near Yelahanka", 170, 5200, 380.0, 2),
+        ("play school admission 2026", 210, 6100, 430.0, 3),
+        # Clear negative/waste terms
+        ("free babysitting jobs", 190, 5000, 620.0, 0),
+        ("toy store near me", 160, 4300, 510.0, 0),
+        ("kids games online", 135, 3600, 420.0, 0),
+        ("montessori certification course", 155, 4100, 540.0, 0),
+        ("teaching jobs preschool", 145, 3700, 460.0, 0),
     ]
 
     rows = []
-
-    for campaign in campaigns:
-        for term in positive_intent + negative_intent:
-
-            clicks = np.random.randint(5, 250)
-            impressions = clicks * np.random.randint(5, 60)
-            cost = round(clicks * np.random.uniform(0.5, 4.0), 2)
-
-            # realistic conversion logic
-            if term in positive_intent:
-                conversions = np.random.randint(1, 25)
-            else:
-                conversions = np.random.choice([0, 0, 0, 1])
-
-            rows.append({
+    for idx, (term, clicks, impressions, cost, conversions) in enumerate(base_rows, start=1):
+        rows.append(
+            {
                 "search_term": term,
-                "campaign_name": campaign,
-                "clicks": clicks,
-                "impressions": impressions,
-                "cost": cost,
-                "conversions": conversions,
-            })
-
-    return pd.DataFrame(rows)
-
-# 3. KEYWORDS
-def generate_sample_keywords():
-    keywords = [
-        "preschool admission",
-        "nursery school near me",
-        "play school Bangalore",
-        "kg admission",
-        "daycare Yelahanka",
-        "best preschool",
-    ]
+                "campaign_name": campaign_name,
+                "clicks": int(round(clicks * multiplier)),
+                "impressions": int(round(impressions * multiplier)),
+                "cost": round(cost * multiplier, 2),
+                "conversions": int(round(conversions * multiplier)),
+            }
+        )
+    return rows
 
-    campaigns = [
-        "Nursery Admissions 2026",
-        "Playgroup Enrollment Campaign",
-        "Summer Camp 2026",
-        "School Tour Booking Campaign",
-    ]
 
+def generate_sample_search_terms():
     rows = []
+    multipliers = {
+        "Nursery Admissions 2026": 1.15,
+        "Playgroup Enrollment Campaign": 0.85,
+        "Summer Camp 2026": 0.75,
+        "School Tour Booking Campaign": 1.0,
+    }
+    for campaign in CAMPAIGNS:
+        rows.extend(_search_rows_for_campaign(campaign, multipliers[campaign]))
+    return pd.DataFrame(rows)
 
-    for campaign in campaigns:
-        for kw in keywords:
-
-            clicks = np.random.randint(50, 500)
-            impressions = clicks * np.random.randint(10, 80)
-            cost = round(clicks * np.random.uniform(0.8, 5.0), 2)
 
-            # realistic conversion behavior
-            if "admission" in kw or "preschool" in kw:
-                conversions = np.random.randint(5, 40)
-            else:
-                conversions = np.random.choice([0, 0, 1, 2, 3])
+def _keyword_rows_for_campaign(campaign_name, profile):
+    if profile == "winner":
+        return [
+            ("preschool admission", 720, 14800, 820.0, 54),
+            ("nursery school near me", 510, 9400, 610.0, 38),
+            ("best preschool", 430, 7600, 520.0, 34),
+            ("kg admission", 390, 7000, 480.0, 29),
+            ("daycare Yelahanka", 280, 8100, 560.0, 4),
+            ("play school Bangalore", 330, 9800, 690.0, 0),
+        ]
+    if profile == "drain":
+        return [
+            ("preschool admission", 620, 14500, 1850.0, 12),
+            ("nursery school near me", 540, 12800, 1760.0, 8),
+            ("play school Bangalore", 470, 11800, 1620.0, 0),
+            ("daycare Yelahanka", 430, 10100, 1490.0, 0),
+            ("best preschool", 390, 9200, 1380.0, 2),
+            ("kg admission", 250, 7400, 860.0, 1),
+        ]
+    if profile == "investigate":
+        return [
+            ("preschool admission", 360, 13200, 910.0, 14),
+            ("nursery school near me", 300, 10100, 780.0, 9),
+            ("summer camp for kids", 560, 22000, 1900.0, 11),
+            ("kids activity camp", 480, 21000, 1680.0, 3),
+            ("free summer activities", 410, 18000, 1250.0, 0),
+            ("art classes for kids", 310, 12000, 940.0, 1),
+        ]
+    return [
+        ("school tour booking", 310, 6200, 340.0, 28),
+        ("book school visit", 250, 5100, 290.0, 22),
+        ("preschool admission", 210, 4800, 260.0, 16),
+        ("nursery school near me", 180, 4200, 230.0, 13),
+        ("daycare Yelahanka", 160, 6100, 410.0, 2),
+        ("teaching jobs preschool", 190, 7800, 520.0, 0),
+    ]
 
-            ctr = round((clicks / impressions) * 100, 2)
 
-            rows.append({
-                "campaign_name": campaign,
-                "keyword": kw,
-                "clicks": clicks,
-                "impressions": impressions,
-                "cost": cost,
-                "conversions": conversions,
-                "ctr": ctr,
-            })
+def generate_sample_keywords():
+    profiles = {
+        "Nursery Admissions 2026": "winner",
+        "Playgroup Enrollment Campaign": "drain",
+        "Summer Camp 2026": "investigate",
+        "School Tour Booking Campaign": "scale",
+    }
 
+    rows = []
+    for campaign in CAMPAIGNS:
+        for keyword, clicks, impressions, cost, conversions in _keyword_rows_for_campaign(campaign, profiles[campaign]):
+            rows.append(
+                {
+                    "campaign_name": campaign,
+                    "keyword": keyword,
+                    "clicks": clicks,
+                    "impressions": impressions,
+                    "cost": cost,
+                    "conversions": conversions,
+                    "ctr": round((clicks / impressions) * 100, 2) if impressions else 0,
+                }
+            )
     return pd.DataFrame(rows)
 
-# ==================================================
-# 4. HOURLY (OPTIONAL BUT MATCHES REAL STRUCTURE)
-# ==================================================
 
 def generate_sample_hourly():
-    hours = list(range(24))
     rows = []
-
-    for campaign in [
-        "Nursery Admissions 2026",
-        "Playgroup Enrollment Campaign",
-        "Summer Camp 2026",
-        "School Tour Booking Campaign",
-    ]:
-        for h in hours:
-            clicks = np.random.randint(0, 80)
-            impressions = clicks * np.random.randint(5, 30)
-            cost = round(clicks * np.random.uniform(0.2, 3.0), 2)
-
-            rows.append({
-                "date": "2026-06-01",
-                "hour": h,
-                "clicks": clicks,
-                "impressions": impressions,
-                "cost": cost,
-            })
-
+    dates = pd.date_range("2026-06-01", periods=14, freq="D")
+    campaign_click_bases = {
+        "Nursery Admissions 2026": 18,
+        "Playgroup Enrollment Campaign": 10,
+        "Summer Camp 2026": 13,
+        "School Tour Booking Campaign": 8,
+    }
+
+    for day_index, date in enumerate(dates):
+        trend_boost = day_index * 0.08
+        for campaign in CAMPAIGNS:
+            base = campaign_click_bases[campaign]
+            for hour in range(24):
+                daypart = 1.4 if 8 <= hour <= 12 or 18 <= hour <= 21 else 0.55
+                clicks = int(round(base * daypart * (1 + trend_boost)))
+                impressions = clicks * 28
+                cost = round(clicks * (1.45 if campaign != "Playgroup Enrollment Campaign" else 2.6), 2)
+                rows.append(
+                    {
+                        "date": date.strftime("%Y-%m-%d"),
+                        "hour": hour,
+                        "campaign_name": campaign,
+                        "clicks": clicks,
+                        "impressions": impressions,
+                        "cost": cost,
+                    }
+                )
     return pd.DataFrame(rows)
 
 
-# ==================================================
-# 5. GEO (OPTIONAL BUT MATCHES REAL STRUCTURE)
-# ==================================================
-
 def generate_sample_geo():
-    countries = ["IN", "AE", "US"]
-    rows = []
-
-    for c in countries:
-        rows.append({
-            "country_id": c,
-            "clicks": np.random.randint(500, 5000),
-            "impressions": np.random.randint(10000, 100000),
-            "cost": round(np.random.uniform(500, 5000), 2),
-        })
-
+    rows = [
+        {"country_id": "IN", "clicks": 6200, "impressions": 128000, "cost": 7600.0},
+        {"country_id": "AE", "clicks": 840, "impressions": 21000, "cost": 1380.0},
+        {"country_id": "US", "clicks": 360, "impressions": 11200, "cost": 980.0},
+    ]
     return pd.DataFrame(rows)
 
 
-# 4. MASTER GENERATOR
 def generate_sample_dfs():
     return {
         "campaigns": generate_sample_campaigns(),
@@ -213,6 +213,6 @@ def generate_sample_dfs():
         "keywords": generate_sample_keywords(),
         "hourly": generate_sample_hourly(),
         "geo": generate_sample_geo(),
-        "devices": pd.DataFrame(),            # optional placeholder
-        "recommendations": pd.DataFrame(),    # optional placeholder
-    }
+        "devices": pd.DataFrame(),
+        "recommendations": pd.DataFrame(),
+    }