updates Mar6

app/routers/analytics.py

@@ -45,6 +45,10 @@ class ChatbotAnalyticsResponse(BaseModel):
     top_queries: List[TopQuery]
     languages_used: Dict[str, int]
     peak_hour: Optional[int]  # 0-23
+    unanswered_count: int = 0
+    unanswered_queries: List[TopQuery] = []
+    feedback_positive: int = 0
+    feedback_negative: int = 0
 
 
 class OverviewAnalyticsResponse(BaseModel):
@@ -212,6 +216,13 @@ async def get_analytics_overview(user=Depends(get_current_user)):
         if rating:
             all_ratings.append(rating)
 
+        # Feedback counts
+        fb_result = supabase.table("message_feedback").select("feedback", count="exact") \
+            .eq("chatbot_id", cid).execute()
+        total_fb = fb_result.count or 0
+        fb_pos = sum(1 for f in (fb_result.data or []) if f.get("feedback") == "positive")
+        fb_neg = total_fb - fb_pos
+
         # Average messages per conversation
         avg_msgs = round(msg_count / conv_count, 1) if conv_count > 0 else 0.0
 
@@ -223,7 +234,7 @@ async def get_analytics_overview(user=Depends(get_current_user)):
             total_messages=msg_count,
             average_messages_per_conversation=avg_msgs,
             average_rating=rating,
-            total_ratings=0,  # would need a ratings table for precise count
+            total_ratings=total_fb,
             conversations_today=today_count,
             conversations_this_week=week_count,
             conversations_this_month=month_count,
@@ -231,6 +242,8 @@ async def get_analytics_overview(user=Depends(get_current_user)):
             top_queries=top_queries,
             languages_used=lang_counts,
             peak_hour=peak,
+            feedback_positive=fb_pos,
+            feedback_negative=fb_neg,
         ))
 
     # Overall average rating
@@ -345,6 +358,48 @@ async def get_chatbot_analytics(chatbot_id: str, user=Depends(get_current_user))
 
     avg_msgs = round(msg_count / conv_count, 1) if conv_count > 0 else 0.0
 
+    # Feedback counts
+    fb_pos = 0
+    fb_neg = 0
+    if conv_ids and conv_ids != [""]:
+        feedback = supabase.table("message_feedback").select("feedback") \
+            .eq("chatbot_id", chatbot_id).execute()
+        for f in (feedback.data or []):
+            if f["feedback"] == "positive":
+                fb_pos += 1
+            else:
+                fb_neg += 1
+
+    # Unanswered queries (low confidence)
+    unanswered_queries: List[TopQuery] = []
+    unanswered_count = 0
+    if conv_ids and conv_ids != [""]:
+        try:
+            low_conf_msgs = supabase.table("messages").select("id, conversation_id, confidence_score") \
+                .in_("conversation_id", conv_ids[:100]) \
+                .eq("role", "assistant") \
+                .lt("confidence_score", 0.2) \
+                .limit(200).execute()
+            unanswered_count = len(low_conf_msgs.data or [])
+            # For each low-confidence assistant message, find the preceding user message
+            if low_conf_msgs.data:
+                unanswered_q_counts: Dict[str, int] = {}
+                for lm in low_conf_msgs.data[:20]:  # limit work
+                    prev_user = supabase.table("messages").select("content") \
+                        .eq("conversation_id", lm["conversation_id"]) \
+                        .eq("role", "user") \
+                        .lt("created_at", lm.get("created_at", "9999")) \
+                        .order("created_at", desc=True) \
+                        .limit(1).execute()
+                    if prev_user.data:
+                        q = (prev_user.data[0].get("content") or "")[:100].strip()
+                        if q:
+                            unanswered_q_counts[q] = unanswered_q_counts.get(q, 0) + 1
+                top_unanswered = sorted(unanswered_q_counts.items(), key=lambda x: -x[1])[:5]
+                unanswered_queries = [TopQuery(query=q, count=n) for q, n in top_unanswered]
+        except Exception:
+            pass  # unanswered queries is optional
+
     return ChatbotAnalyticsResponse(
         chatbot_id=chatbot_id,
         chatbot_name=cb.get("name", "Untitled"),
@@ -353,7 +408,7 @@ async def get_chatbot_analytics(chatbot_id: str, user=Depends(get_current_user))
         total_messages=msg_count,
         average_messages_per_conversation=avg_msgs,
         average_rating=cb.get("average_rating"),
-        total_ratings=0,
+        total_ratings=fb_pos + fb_neg,
         conversations_today=today_count,
         conversations_this_week=week_count,
         conversations_this_month=month_count,
@@ -361,4 +416,56 @@ async def get_chatbot_analytics(chatbot_id: str, user=Depends(get_current_user))
         top_queries=top_queries,
         languages_used=lang_counts,
         peak_hour=peak,
+        unanswered_count=unanswered_count,
+        unanswered_queries=unanswered_queries,
+        feedback_positive=fb_pos,
+        feedback_negative=fb_neg,
     )
+
+
+@router.get("/chatbot/{chatbot_id}/gaps", response_model=List[TopQuery])
+async def get_knowledge_gaps(chatbot_id: str, user=Depends(get_current_user)):
+    """Returns top queries where the bot had low confidence (knowledge gaps). Starter+ only."""
+    plan = _get_user_plan(user.id)
+    _check_analytics_access(plan)
+
+    supabase = get_supabase()
+    company = supabase.table("companies").select("id").eq("owner_id", user.id).execute()
+    if not company.data:
+        raise HTTPException(status_code=404, detail="Company not found")
+
+    chatbot = supabase.table("chatbots").select("id") \
+        .eq("id", chatbot_id).eq("company_id", company.data[0]["id"]).execute()
+    if not chatbot.data:
+        raise HTTPException(status_code=404, detail="Chatbot not found")
+
+    # Find conversations
+    convs = supabase.table("conversations").select("id").eq("chatbot_id", chatbot_id).execute()
+    conv_ids = [c["id"] for c in (convs.data or [])]
+    if not conv_ids:
+        return []
+
+    # Low confidence assistant messages
+    low_conf = supabase.table("messages").select("id, conversation_id, created_at") \
+        .in_("conversation_id", conv_ids[:100]) \
+        .eq("role", "assistant") \
+        .lt("confidence_score", 0.2) \
+        .limit(100).execute()
+
+    if not low_conf.data:
+        return []
+
+    q_counts: Dict[str, int] = {}
+    for msg in low_conf.data[:30]:
+        prev = supabase.table("messages").select("content") \
+            .eq("conversation_id", msg["conversation_id"]) \
+            .eq("role", "user") \
+            .order("created_at", desc=True) \
+            .limit(1).execute()
+        if prev.data:
+            content = (prev.data[0].get("content") or "")[:100].strip()
+            if content:
+                q_counts[content] = q_counts.get(content, 0) + 1
+
+    sorted_gaps = sorted(q_counts.items(), key=lambda x: -x[1])[:10]
+    return [TopQuery(query=q, count=n) for q, n in sorted_gaps]