Refined the files

.gitignore

@@ -8,4 +8,5 @@ my_agent/utils/check.py
 /pages
 streamlit_app.py
 data
-extracted_data.csv
+extracted_data.csv
+Research

brainstroming_agent/utils/nodes.py

@@ -2,7 +2,7 @@ import pandas as pd
 import ast
 from .state import State
 from .tools import retrieve_tool
-from langchain_core.messages import SystemMessage
+from langchain_core.messages import SystemMessage ,HumanMessage
 from utils.models_loader import llm , ST
 from utils.data_loader import load_influencer_data
 from groq import Groq
@@ -37,6 +37,8 @@ def caption_image(state: State) -> State:
                   }
               ],
               model="meta-llama/llama-4-scout-17b-16e-instruct",
+              max_completion_tokens=50,
+              temperature = 1
           )
           response=chat_completion.choices[0].message.content
           state.image_captions.append(response)
@@ -59,7 +61,7 @@ def retrieve(state: State) -> State:
           scores, retrieved_examples = data.get_nearest_examples("embeddings", embedded_query, k=1)
 
           # Construct a list of dictionaries for this topic
-          result = [{user: story} for user, story in zip(retrieved_examples['username'], retrieved_examples['agentic_story'])]
+          result = [{user: story} for user, story in zip(retrieved_examples['username'], retrieved_examples['story'])]
           retrievals.append(result)
           print('Retrieval process completed......')
       state.retrievals.append(retrievals)
@@ -73,14 +75,11 @@ def retrieve(state: State) -> State:
           scores, retrieved_examples = data.get_nearest_examples("embeddings", embedded_query, k=1)
 
           # Construct a list of dictionaries for this topic
-          result = [{user: story} for user, story in zip(retrieved_examples['username'], retrieved_examples['agentic_story'])]
+          result = [{user: story} for user, story in zip(retrieved_examples['username'], retrieved_examples['story'])]
           retrievals.append(result)
           print('Retrieval process completed for preferred_topics......')
       state.latest_preferred_topics=[]
       state.retrievals.append(retrievals)
-
-#   print('The retrieval is:\n',state.retrievals )
-  # return State(messages="Retrieved",topic=state.topic,retrievals=state.retrievals)
   return state
 
 def generate_story(state:State)-> State:
@@ -98,7 +97,8 @@ def generate_story(state:State)-> State:
 
 # and {state.image_captions[-1]}
 
-    messages = [SystemMessage(content=template)]
+    messages = [SystemMessage(content=template),
+                HumanMessage(content=f'''The topic of the video is:\n{state.topic}\n''')]
 
     response = react_agent.invoke({'messages':messages})
     response = response['messages'][-1].content

brainstroming_agent/utils/prompts.py

@@ -11,43 +11,94 @@ Desired output:
 '''
 
 image_captioning_prompt = '''
-Analyze the image deeply and give the story of the image what it wants to say?
+Analyze the image deeply and give the story of the image what it potraits in just 50 words.
 '''
 
 
 def initial_story_prompt(state):
-    return (
-        f'''
-I want to create a detailed storyline for a video in any domain. You have to provide me that storyline what to include in the video.
-    Now, i am giving you the topic of the video. But the need is to generate the story focusing on the response of the tool provided to you..
-    
-    \n\n Now let's start creating the storyline for my topic. The idea of the video is: \n\n{state.topic}\n\n and the business details are:\n {state.business_details}'''
-    )
+    return f"""
+You are an expert video content strategist who generates the storyline of a video in exactly 300 words..
+
+Your task is to generate a **detailed and creative storyline** for a promotional video on the **given topic**. The storyline should be structured, engaging, and highly relevant to the following **business details** and **inspiration from the image context if it contains some informations*.
+Another important thing is that you have to give the response  focusing on the response of the tool provided to you. The tool contains the video stories and contents created by the influencers.
+Use that responses of tool for your reference. You can use your creativity but inside the boundary of tool's response.
+
+---
+
+ **Video Topic**:  
+{state.topic}
+
+**Business Details**:  
+{state.business_details}
+
+
+**Inspiration from Image Context** (must be strongly incorporated):  
+{state.image_captions[-1]}
+
+---
+
+🎯 **Important Instructions**:
+- Creatively connect the image cues to make the storyline compelling. (If provided).
+- Structure the storyline logically (e.g., beginning, middle, end), showing what scenes or visuals to include.
+- Match the tone and depth to the business type (e.g., fun, luxurious, emotional, professional).
+
+Now, generate the final storyline for the video topic..
+"""
 
 def refined_story_prompt(state):
-    return(
-        f'''
-I want to create a detailed storyline for a video in the given topic. You have to provide me that storyline what to include in the video.
-    Now, i am giving you the topic of the video. But the need is to generate the story focusing on the response of the tool provided to you..
-  Now, i am giving you the topic of the video. 
-  \n\n Now let's start creating the storyline for my topic. The topic of the video is: \n\n{state.topic}\n\n and the business details are:\n {state.business_details}
+    return f"""
+You are an expert video content strategist who generates the storyline of a video in exactly 300 words..
 
-  **Final and must Reminder**: Please You have to strongly focus on these topics while creating the storyline: {state.preferred_topics[-1]} and {state.image_captions[-1]}'''
-    )
+Your task is to generate a **detailed and creative storyline** for a promotional video on the **given topic**. The storyline should be structured, engaging, and highly relevant to the following **business details** , **preferred focus areas** and **inspiration from the image context(if any)**.
+Another important thing is that you have to give the response only focusing on the response of the tool provided to you. The tool contains the video stories and contents created by the influencers.
+Use that responses of tool for your reference. You can use your creativity but inside the boundary of tool's response.
+
+---
+
+ **Video Topic**:  
+{state.topic}
+
+**Business Details**:  
+{state.business_details}
+
+**Preferred Focus Areas** *(must be strongly incorporated)*:  
+{state.preferred_topics[-1]}
+
+**Inspiration from Image Context** (must be strongly incorporated):  
+{state.image_captions[-1]}
+
+---
+
+🎯 **Important Instructions**:
+- Creatively connect the preferred topics and image cues to make the storyline compelling.
+- Structure the storyline logically (e.g., beginning, middle, end), showing what scenes or visuals to include.
+- Match the tone and depth to the business type (e.g., fun, luxurious, emotional, professional).
+
+Now, generate the final storyline for the video topic..
+"""
 
 def brainstroming_prompt(state):
-    return(
-        f'''
-I want to brainstorm ways to diversify or improve a storyline in exactly 4 sentences.
-The goal is to generate creative and actionable ideas that are not on the storyline on how the storyline can be expanded or modified for better engagement.
-For example: If the storyline is about creating a promotional video for a restaurant, the new suggestions might include:
-- I want to showcase the chef preparing a signature dish.
-- I want to add a sequence of customers sharing their experiences at the restaurant.
-- I want to highlight the farm-to-table sourcing of ingredients with a short segment showing local farms.
-- I want to include a time-lapse of the restaurant transforming from day to night, capturing its unique ambiance.
-- I want to feature a quick interview with the owner sharing the story behind the restaurant.
-Now, I will provide you with the storyline. The storyline is:\n{state.stories[-1]}'''
-    )
+    return f'''
+You are a creative consultant tasked with enhancing a storyline.
+
+Suggest **4 unique and creative enhancements** to the following storyline. Each suggestion must:
+- Be just 1 sentence long”
+- Add something **new or different** (not rephrasing what’s already in the story)
+- Improve the storyline’s **engagement, emotional depth, or audience appeal**
+
+Output format: a numbered list from 1 to 4.
+
+Here’s an example for a restaurant promo storyline:
+1. Showcase the chef preparing a signature dish.
+2. Add a sequence of customers sharing their experiences at the restaurant.
+3. Highlight the farm-to-table sourcing of ingredients with a short segment showing local farms.
+4. Include a time-lapse of the restaurant transforming from day to night, capturing its unique ambiance.
+
+Now, here is the current storyline you need to enhance:
+
+"{state.stories[-1]}"
+'''
+
 
 def final_story_prompt(final_state):
     return(

brainstroming_agent/utils/state.py

@@ -25,6 +25,6 @@ class BrainstromTopicFormatter(BaseModel):
     topic4:str=Field(description="Fourth brainstorming topic of the story")
 
 class QueryFormatter(BaseModel):
-    messages: str = Field(description="The message that user passes to the agent")
-    business_details: dict = Field (description="The business details that user passes to the agent")
+    video_topic: str = Field(description="The video topic that user passes to the agent")
+    # business_details: dict = Field (description="The business details that user passes to the agent")
 

brainstroming_agent/utils/tools.py

@@ -7,7 +7,6 @@ import numpy as np
 from langchain_core.tools import tool
 from utils.data_loader import load_influencer_data
 from utils.models_loader import  ST , llm
-from sklearn.metrics.pairwise import cosine_similarity
 import numpy as np
 from langchain_core.messages import SystemMessage
 import re
@@ -18,10 +17,10 @@ from .state import QueryFormatter
 
 os.environ['GROQ_API_KEY']=os.getenv('GROQ_API_KEY')
 @tool("influencer's data-retrieval-tool", args_schema=QueryFormatter, return_direct=False,description="Retrieve influencer-related data for a given query.")
-def retrieve_tool(messages, business_details):
+def retrieve_tool(video_topic):
     '''
     Always invoke this tool.
-    Retrieve influencer's data by semantic search of **user messages** and the **business details**.
+    Retrieve influencer's data by semantic search of **video topic**.
     '''
     # === Load CSV ===
     csv_path = 'extracted_data.csv'
@@ -39,8 +38,8 @@ def retrieve_tool(messages, business_details):
     # === Load SentenceTransformer model ===
 
     # === Encode the query and search ===
-    query_embedding = ST.encode(str(messages)+str(business_details)).reshape(1, -1).astype('float32')
-    top_k=10
+    query_embedding = ST.encode(str(video_topic)).reshape(1, -1).astype('float32')
+    top_k=7
     distances, indices = index.search(query_embedding, top_k)
 
 

business_interaction_agent/utils/prompts.py

@@ -38,7 +38,7 @@ Rules:
 - Do NOT invent or assume any information beyond what is explicitly provided in the retrievals and business details.
 
 Keep your response:
-- Context-aware
+- Context-aware, direct and short as possible
 - Grounded only in the given data
 - Helpful, concise, and user-friendly
 '''.strip()

main.py

@@ -9,7 +9,8 @@ from business_interaction_agent.agent import BusinessInteractionChatbot
 from context_analysis_agent.utils.utils import save_to_db
 import ast
 from brainstroming_agent.utils.utils import encode_image_to_base64 , generate_final_story, generate_image
-
+from fastapi import Body, Query
+from ideation_agent.agent import ideation_graph
 
 import json
 
@@ -18,11 +19,12 @@ import json
 app = FastAPI()
 context_analysis_graph = IntroductionChatbot()
 business_interaction_graph = BusinessInteractionChatbot()
+idea_graph = ideation_graph()
 brainstrom_graph = brainstroming_graph()
 
 stored_data={}
-stored_data['business_details']={"business_type": "restaurant", "platform": "instagram", "target_audience": "youths", "business_goals": "to go global", "offerings": "nepali foods", "Challenges_faced": "finding new customers, attracting large customers"}
-
+stored_data['business_details']={"business_type": "cosmetic", "platform": "instagram", "target_audience": "youths", "business_goals": "to go global", "offerings": "nepali skin care products", "Challenges_faced": "finding new customers, attracting large customers"}
+# stored_data['business_details']={}
 class UserMessage(BaseModel):
     message: str
 @app.post("/context-analysis")
@@ -49,32 +51,44 @@ def business_interaction(interaction: str):
     stored_data['business_details']=business_details
     return {'response': response}
 
+class IdeationRequest(BaseModel):
+    topic : List[str]
+    business_details: dict
+
+@app.post("/ideation")
+def ideation_endpoint(request:IdeationRequest):
+    result = idea_graph.invoke({
+        'topic': request.topic,
+        'business_details': stored_data['business_details']
+    })
+    return {'response':result}
+
+
+
+class BrainstormRequest(BaseModel):
+    query: List[str]
+    preferred_topics: Optional[List] = []
+    image_base64_list: Optional[List] = []
+    thread_id: Optional[str]="default-session"
 
 @app.post("/brainstrom")
 def brainstroming_endpoint(
-    query: List[str],  # sent as JSON body
-    preferred_topics: Optional[list] = [],
-    images: Optional[List[UploadFile]] = [],  # ✅ Optional UploadFile list
-    thread_id: Optional[str] = "default-session",
+    request: BrainstormRequest,  # 🔥 Full JSON body here
+    # thread_id: Optional[str] = Query("default-session"),  # Separate query param
 ):
-    # Convert uploaded images to base64
-    image_base64_list = [encode_image_to_base64(img) for img in images]
-
-    # Invoke LangGraph
     result = brainstrom_graph.invoke({
-        'topic': query,
-        'images': image_base64_list,
-        'latest_preferred_topics':preferred_topics,
+        'topic': request.query,
+        'images': request.image_base64_list,
+        'latest_preferred_topics': request.preferred_topics,
         'business_details': (lambda d: d['business_details'] if 'business_details' in d else {})(stored_data)
-
     },
-    config={"configurable": {"thread_id": thread_id}})
-    stored_data['brainstroming_response']=result
-    # brainstorm_store[thread_id] = result
+    config={"configurable": {"thread_id": request.thread_id}})
+    
+    stored_data['brainstroming_response'] = result
+
+    return {'response': result}
+
 
-    return {
-        'response': result,
-    }
 
 @app.post("/generate-final-story")
 def generate_final_story_endpoint():

utils/models_loader.py

@@ -14,19 +14,19 @@ login(os.environ['HUGGINGFACEHUB_ACCESS_TOKEN'])
 os.environ['GROQ_API_KEY']=os.getenv('GROQ_API_KEY')
 
 
-llm = ChatGroq(
-    model="llama-3.3-70b-versatile",
+# llm = ChatGroq(
+#     model="llama3-8b-8192",
+#     temperature=0.3,
+#     max_tokens=500,
+
+# )
+
+llm = ChatOpenAI(
+    model="gpt-4o-mini",
     temperature=0.3,
     max_tokens=500,
-
 )
 
-# llm = ChatOpenAI(
-#     model="gpt-4o-mini",
-#     temperature=0,
-#     max_tokens=500,
-# )
-
 
 class HFEmbeddingAPI:
     def __init__(self, api_url, token):
@@ -52,3 +52,42 @@ ST = HFEmbeddingAPI(
     api_url="https://router.huggingface.co/hf-inference/models/mixedbread-ai/mxbai-embed-large-v1/pipeline/feature-extraction",
     token=os.environ.get('HUGGINGFACEHUB_ACCESS_TOKEN')
 )
+
+# Initiated the models for ideation
+
+ideator_llm = ChatGroq(
+    model="llama3-8b-8192",
+    temperature=0.3,
+    max_tokens=500,
+
+)
+
+critic_llm = ChatGroq(
+    model="gemma2-9b-it",
+    temperature=0.3,
+    max_tokens=500,
+
+)
+
+# improver_llm = ChatOpenAI(
+#     model="gpt-4o-mini",
+#     temperature=0.3,
+#     max_tokens=500,
+# )
+
+improver_llm = ChatGroq(
+    model="meta-llama/llama-4-scout-17b-16e-instruct",
+    temperature=0.3,
+    max_tokens=500,
+
+)
+
+
+validator_llm = ChatGroq(
+    model="gemma2-9b-it",
+    temperature=0.3,
+    max_tokens=500,
+
+)
+
+