next commit

logs/access.log

@@ -1640,3 +1640,28 @@
 2025-11-12 12:20:31,658 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/openapi.json
 2025-11-12 12:20:31,659 | INFO | access_logger | app.py:20 | Response status: 200
 2025-11-12 12:21:21,920 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/api/analytics-chatbot?msg=I%20want%20to%20view%20the%20overall%20audience%20analytics%20of%20divya
+2025-11-12 16:40:57,113 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/
+2025-11-12 16:40:57,114 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:41:00,503 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/docs
+2025-11-12 16:41:00,503 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:41:00,667 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/openapi.json
+2025-11-12 16:41:00,674 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:41:37,091 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/api/analytics-chatbot?msg=I%20want%20to%20view%20the%20emoji%20counts%20of%20divya%20dhakal
+2025-11-12 16:45:48,973 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/
+2025-11-12 16:45:48,974 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:45:53,820 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/docs
+2025-11-12 16:45:53,821 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:45:53,865 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/openapi.json
+2025-11-12 16:45:53,871 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:46:21,569 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/api/analytics-chatbot?msg=I%20want%20to%20view%20the%20emoji%20count%20of%20divya%20dhakal
+2025-11-12 16:53:03,322 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/docs
+2025-11-12 16:53:03,323 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:53:03,371 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/openapi.json
+2025-11-12 16:53:03,372 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:53:04,862 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/
+2025-11-12 16:53:04,863 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:53:07,974 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/docs
+2025-11-12 16:53:07,975 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:53:08,019 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/openapi.json
+2025-11-12 16:53:08,021 | INFO | access_logger | app.py:20 | Response status: 200
+2025-11-12 16:53:38,058 | INFO | access_logger | app.py:18 | Request: GET http://127.0.0.1:8000/api/analytics-chatbot?msg=I%20want%20to%20view%20the%20emoji%20count%20of%20divya%20dhakal

src/genai/analytics_chatbot/handlers/bot_and_diversity.py

@@ -0,0 +1,16 @@
+
+import requests
+from langchain_core.messages import SystemMessage,HumanMessage
+from ..utils.schemas import BotAndDiversityFormatter
+from ..utils.prompts import bot_and_diversity_prompt
+from ..utils.utils import process_query
+
+
+def get_bot_and_diversity(state,llm_gpt,url):
+    messages = [SystemMessage(content=bot_and_diversity_prompt), 
+    HumanMessage(content=str(state['messages']))]
+    parameters=llm_gpt.with_structured_output(BotAndDiversityFormatter , method='function_calling').invoke(messages)
+    print(parameters)
+    response = requests.get(url, params={'top_n': parameters.top_n, 'start_date': None , 'end_date':None,'influencer_username':parameters.influencer_name})
+    return response
+

src/genai/analytics_chatbot/handlers/comment_quality.py

@@ -0,0 +1,17 @@
+
+
+import requests
+from langchain_core.messages import SystemMessage,HumanMessage
+from ..utils.schemas import CommentQualityFormatter
+from ..utils.prompts import comment_quality_prompt
+from ..utils.utils import process_query
+
+
+def get_comment_quality(state,llm_gpt,url):
+    messages = [SystemMessage(content=comment_quality_prompt), 
+    HumanMessage(content=str(state['messages']))]
+    parameters=llm_gpt.with_structured_output(CommentQualityFormatter , method='function_calling').invoke(messages)
+    print(parameters)
+    response = requests.get(url, params={ 'start_date': parameters.start_date , 'end_date':parameters.end_date,'influencer_username':process_query(parameters.influencer_name)})
+    return response.json()
+

src/genai/analytics_chatbot/handlers/compare.py

@@ -2,11 +2,11 @@
 import requests
 from langchain_core.messages import SystemMessage,HumanMessage
 from ..utils.schemas import CompareBodyFormatter
-from ..utils.prompts import get_body_prompt
+from ..utils.prompts import compare_prompt
 from ..utils.utils import process_query
 
 def compare(state,llm_gpt,url):
-    messages = [SystemMessage(content=get_body_prompt()), 
+    messages = [SystemMessage(content=compare_prompt()), 
     HumanMessage(content=str(state['messages']))]
     response=llm_gpt.with_structured_output(CompareBodyFormatter , method='function_calling').invoke(messages)
     print('INF names response:', response)

src/genai/analytics_chatbot/handlers/emoji_count.py

@@ -0,0 +1,18 @@
+
+import requests
+from langchain_core.messages import SystemMessage,HumanMessage
+from ..utils.schemas import EmojiCountFormater
+from ..utils.prompts import emoji_count_prompt
+from ..utils.utils import process_query
+
+
+
+def get_emoji_count(state,llm_gpt , url):
+    messages = [SystemMessage(content=emoji_count_prompt), 
+    HumanMessage(content=str(state['messages']))]
+    parameters=llm_gpt.with_structured_output(EmojiCountFormater , method='function_calling').invoke(messages)
+    print(parameters)
+    response = requests.get(url, params={'top_n': parameters.top_n, 'influencer_username':process_query(parameters.influencer_name)})
+    print('The response is:', response)
+    return response
+

src/genai/analytics_chatbot/handlers/peak_comment_hour.py

@@ -0,0 +1,13 @@
+import requests
+from langchain_core.messages import SystemMessage,HumanMessage
+from ..utils.schemas import PeakCommentHourFormatter
+from ..utils.prompts import peak_comment_hour_prompt
+from ..utils.utils import process_query
+
+def get_peak_comment_hour(state,llm_gpt,url):
+    messages = [SystemMessage(content=peak_comment_hour_prompt), 
+    HumanMessage(content=str(state['messages']))]
+    parameters=llm_gpt.with_structured_output(PeakCommentHourFormatter , method='function_calling').invoke(messages)
+    response = requests.get(url, params={'start_date': parameters.start_date , 'end_date':parameters.end_date, 'influencer_username':process_query(parameters.influencer_name)})
+    return response.json()
+

src/genai/analytics_chatbot/handlers/posting_time.py

@@ -0,0 +1,18 @@
+import requests
+from langchain_core.messages import SystemMessage,HumanMessage
+from ..utils.schemas import PostingTimeFormatter
+from ..utils.prompts import posting_time_analysis_prompt
+from ..utils.utils import process_query
+
+    
+
+def get_posting_time(state,llm_gpt,url):
+    messages = [SystemMessage(content=posting_time_analysis_prompt), 
+    HumanMessage(content=str(state['messages']))]
+    parameters=llm_gpt.with_structured_output(PostingTimeFormatter , method='function_calling').invoke(messages)
+    response = requests.get(url, params={'start_date': parameters.start_date , 'end_date':parameters.end_date, 'influencer_username':process_query(parameters.influencer_name)})
+    return response
+
+
+
+

src/genai/analytics_chatbot/utils/nodes.py

@@ -2,14 +2,19 @@ import requests
 from langchain_core.messages import SystemMessage , HumanMessage , FunctionMessage
 from .state import State
 from .tools import RetrieverBackup
-from .schemas import CompareBodyFormatter, ParameterFormatter, EndpointFormatter
-from .prompts import query_check_prompt , get_body_prompt , fetch_last_message_prompt , fetch_parameters_prompt, fetch_endpoint_prompt, backup_retrieval_prompt
+from .schemas import ParameterFormatter, EndpointFormatter
+from .prompts import query_check_prompt, fetch_last_message_prompt , fetch_parameters_prompt, fetch_endpoint_prompt
 from .utils import  process_query, get_endpoint_info
 from src.genai.utils.models_loader import llm_gpt
 import numpy as np
 from src.genai.utils.data_loader import api_knowledge_df, api_index, caption_df , caption_index
 from src.genai.utils.models_loader import embedding_model
 from ..handlers.compare import compare
+from ..handlers.posting_time import get_posting_time
+from ..handlers.peak_comment_hour import get_peak_comment_hour
+from ..handlers.emoji_count import get_emoji_count
+from ..handlers.comment_quality import get_comment_quality
+
 
 
 class FetchLastMessage:
@@ -33,16 +38,17 @@ class RetrievePossibleEndpoints:
     def __init__(self):
         self.df = api_knowledge_df
         self.index = api_index
-        self.results = []
+        # self.results = []
+        self.results = ['/api/v1/compare/', '/api/v1/engagement/basic-metrics', '/api/v1/content/hashtags-analysis', '/api/v1/audience/emoji-count', '/api/v1/engagement/temporal_analysis']
 
     def run(self,state:State):
         print('Gone to retrieve possible endpoints')
-        query_embedding = np.array(embedding_model.embed_query(state['latest_message'])).reshape(1, -1).astype('float32')
-        distances, indices = self.index.search(query_embedding, 5)
-        for idx in indices[0]:
-            row = self.df.iloc[idx]
-            print('Endpoint:',row['endpoint'])
-            self.results.append(row['endpoint'])
+        # query_embedding = np.array(embedding_model.embed_query(state['latest_message'])).reshape(1, -1).astype('float32')
+        # distances, indices = self.index.search(query_embedding, 5)
+        # for idx in indices[0]:
+        #     row = self.df.iloc[idx]
+        #     print('Endpoint:',row['endpoint'])
+        #     self.results.append(row['endpoint'])
         print('The possible endpoints are:', self.results)
         return {
             "possible_endpoints": self.results,
@@ -91,13 +97,13 @@ class QueryCheckNode:
 class FetchParametersNode:
     def __init__(self):
         self.llm = llm_gpt
-    
+        self.complex_endpoints=['/api/v1/compare/','/api/v1/engagement/posting-time-analysis','/api/v1/audience/peak-comment-hour','/api/v1/audience/emoji-count','/api/v1/audience/comment-quality']
+
     def run(self , state:State):
         try:
             print('Entered to fetch parameters')
-            print(state['method'])
 
-            if state['method']=='GET':
+            if state['endpoint'] not in self.complex_endpoints:
                 template = fetch_parameters_prompt
                 messages=[SystemMessage(content=template),HumanMessage(content=f'''The query is: {state['latest_message']}\n. The needed parameters: {str(state['needed_parameters'])}''')]
                 # print('messages:', messages)
@@ -132,32 +138,30 @@ class FetchDataNode:
 
     def run(self, state:State):
         try:
+
             print('Entered to fetch data')
             url = f'''{self.base_url}{state['endpoint']}'''
 
             if state['endpoint'] == '/api/v1/compare/':
                 response=compare(state,llm_gpt,url)
                 return {'response': response.json()}
-                # print('Condition satisfied')
-                # messages = [SystemMessage(content=get_body_prompt()), 
-                # HumanMessage(content=str(state['messages']))]
-                # response=llm_gpt.with_structured_output(CompareBodyFormatter , method='function_calling').invoke(messages)
-                # print('INF names response:', response)
-                # payload = {
-                # "usernames": list(map(process_query,response.names)),
-                # "freq": response.frequency          
-                # }
-                
-                # print('The payload is:',payload)
-                
-                # headers = {
-                # "Content-Type": "application/json"
-                # }
-
-                # response = requests.post(url, json=payload, headers=headers)
-                # print('Data from api:', response)
-                # return {'response': response.json()}
             
+            elif state['endpoint'] == '/api/v1/engagement/posting-time-analysis':
+                response = get_posting_time(state, llm_gpt,url)
+                return {'response': response.json()}
+            
+            elif state['endpoint']=='/api/v1/audience/peak-comment-hour':
+                response = get_peak_comment_hour(state,llm_gpt,url)
+                return {'response':response.json()}
+            
+            elif state['endpoint']== '/api/v1/audience/emoji-count':
+                response = get_emoji_count(state,llm_gpt,url)
+                return {'response:',response.json()}
+            
+            elif state['endpoint']== '/api/v1/audience/comment-quality':
+                response = get_comment_quality(state,llm_gpt,url)
+                return {'response:',response.json()}
+
             elif 'single_influencer_query' in  state['query_type']:
                 response = requests.get(url, params=state['parameters_values'],headers=self.headers)
                 print('Data from api:', response)

src/genai/analytics_chatbot/utils/prompts.py

@@ -1,47 +1,5 @@
-def chatbot_prompt():
-    return f"""
-You are an intelligent assistant whose task is to route user queries to the correct API endpoint.
-You have access to the API knowledge base, which contains information about each endpoint:
-- The endpoint path
--The method 'GET' or 'POST'
-- Its required parameters
-- A description of what the endpoint does
-
-Your job is to:
-1. Read the user's natural language query.
-2. Analyze the API knowledge base.
-3. Identify the **most appropriate endpoint** that can satisfy the user's request.
-4. Determine the required parameters for that endpoint and fill in their values based on the user's query.
-5. Return the result in a **strict JSON format** exactly like this:
-
-  "endpoint": "<chosen endpoint path>",
-  "method": GET or POST
-  "parameters":
-    "<param1>": "<value1>",
-    "<param2>": "<value2>"
-
-Important instructions:
-- Only return endpoints that exist in the API knowledge base.
-- Include all required parameters for the endpoint.
-- If the parameter or method  is not specified in the user's query, return it as null.
-- Do not add any extra explanation or text; return **only the JSON**.
-- The API knowledge base will be provided as a separate function message.
 
-Example:
-User query: "Give me the buzz trend of influencer John for last month"
-API knowledge: contains endpoint "/overview/buzz_trend" with parameters ["period", "influencer_username"]
-Expected output:
-
-  "endpoint": "/api/v1/overview/buzz_trend",
-  "method": GET
-  "parameters": 
-    "period": "monthly",
-    "influencer_username": "John"
-
-Your response must always follow this exact JSON format.
-"""
-
-def get_body_prompt():
+def compare_prompt():
     return '''You are given a user query  for comparing influencers.
 
 Your task:
@@ -110,11 +68,6 @@ endpoint: /api/v1/analytics/engagement
 
 '''
 
-backup_retrieval_prompt = '''
-You are provided with the retrieved data as a function message and the user query.
-Respond to the user query only through the context of retrieved data. Don't give hallucinated responses.
-'''
-
 query_check_prompt = '''
 You are an intent classification assistant.  
 Given a user query about influencer analytics, classify it as one of the following types:
@@ -123,4 +76,36 @@ Given a user query about influencer analytics, classify it as one of the followi
 2. aggregate_query — if the query involves comparing multiple influencers, rankings, or overall statistics (e.g., "Who has the highest engagement?").  
 
 Return only one label: "single_influencer_query" or "aggregate_query".
+'''
+
+
+posting_time_analysis_prompt = '''
+You are perfect parameters extractor for posting time analysis of the influencer.
+Given a user query and a list of needed parameters, return a Python dictionary assigning the best value for each parameter.  
+You have to return a dictionary containing influencer_name , start_date and end_date. If there is no any mention of the dates, keep the dates as None.
+'''
+
+peak_comment_hour_prompt = '''
+You are perfect parameters extractor for analysis of peak comment hour the influencer.
+Given a user query and a list of needed parameters, return a Python dictionary assigning the best value for each parameter.  
+You have to return a dictionary containing influencer_name , start_date and end_date. If there is no any mention of the dates, keep the dates as None.
+'''
+
+emoji_count_prompt = '''
+You are perfect parameters extractor for analysis of emoji count of the influencer.
+Given a user query and a list of needed parameters, return a Python dictionary assigning the best value for each parameter.  
+You have to return a dictionary containing influencer_name , and the number of emoji (top_n) by understanding the user query. If there is no any mention of the number of emoji, then keep it 15 as default.
+'''
+
+comment_quality_prompt = '''
+You are perfect parameters extractor for analysis of comment quality of the influencer.
+Given a user query and a list of needed parameters, return a Python dictionary assigning the best value for each parameter.  
+You have to return a dictionary containing influencer_name , start_date and end_date. If there is no any mention of the dates, keep the dates as None.
+'''
+
+bot_and_diversity_prompt = '''
+You are perfect parameters extractor for analysis of bot and comment diversity of the influencer.
+Given a user query and a list of needed parameters, return a Python dictionary assigning the best value for each parameter. 
+You have to return a dictionary containing influencer_name , number of commentors (top_n), start_date and  end_date from the user query.
+If there is no any specific mention of dates, you can return None for dates. In the case of number of commentors, return a default value of 10 if the number is not passed from the user.
 '''

src/genai/analytics_chatbot/utils/schemas.py

@@ -1,5 +1,6 @@
 from pydantic import BaseModel, Field
 from typing import Optional , Dict , Any
+from datetime import date
 
 class ResponseFormatter(BaseModel):
     endpoint: str = Field(description='Return the exact endpoint from the knowledge base of endpoints.')
@@ -21,4 +22,29 @@ class ParameterFormatter(BaseModel):
 
 class EndpointFormatter(BaseModel):
     endpoint: str
+
+class PostingTimeFormatter(BaseModel):
+    start_date: Optional[date] = None
+    end_date: Optional[date] = None
+    influencer_name: str
+
+class PeakCommentHourFormatter(BaseModel):
+    start_date: Optional[date] = None
+    end_date: Optional[date] = None
+    influencer_name: str
+
+class EmojiCountFormater(BaseModel):
+    top_n: int
+    influencer_name: str
+
+class CommentQualityFormatter(BaseModel):
+    start_date: Optional[date] = None
+    end_date: Optional[date] = None
+    influencer_name: str
+
+class BotAndDiversityFormatter(BaseModel):
+    start_date: Optional[date] = None
+    end_date: Optional[date] = None
+    influencer_name: str
+    top_n: int
     

src/genai/utils/models_loader.py

@@ -23,8 +23,8 @@ llm_groq = ChatGroq(model="llama-3.3-70b-versatile",temperature=0)
 llm_gpt_small = ChatOpenAI(model="gpt-3.5-turbo",temperature=0.3)
 llm_gpt = ChatOpenAI(model="gpt-3.5-turbo",temperature=0.3)
 llm_gpt_high = ChatOpenAI(model="gpt-5-nano",temperature=0.5)
-encoding_model = tiktoken.encoding_for_model('gpt-4o-mini')
-# encoding_model = 'encoding_model'
+# encoding_model = tiktoken.encoding_for_model('gpt-4o-mini')
+encoding_model = 'encoding_model'
 
 
 captioning_model = "meta-llama/llama-4-scout-17b-16e-instruct"