Upload folder using huggingface_hub

common/aagents/google_agent.py

@@ -135,5 +135,6 @@ google_agent = Agent(
         - Respect timeouts and handle errors gracefully
         """,
 )
+google_agent.description = "A Google search agent that finds accurate, up-to-date information and recent news using Google Search."
 
 __all__ = ["google_agent", "google_search", "google_search_recent", "duckduckgo_search", "fetch_page_content", "current_datetime"]

common/aagents/healthcare_agent.py

@@ -6,7 +6,7 @@ from openai import AsyncOpenAI
 
 # Import tools
 from mcp.tools.rag_tool import rag_search, UserContext
-from mcp.tools.search_tools import duckduckgo_search
+from mcp.tools.search_tools import duckduckgo_search, fetch_page_content
 from mcp.tools.time_tools import current_datetime
 
 
@@ -34,7 +34,7 @@ groq_model = OpenAIChatCompletionsModel(model="groq/compound", openai_client=gro
 healthcare_agent = Agent[UserContext](
     name="HealthcareRAGAgent",
     model=gemini_model,
-    tools=[rag_search, duckduckgo_search],
+    tools=[rag_search, duckduckgo_search, fetch_page_content],
     instructions="""
         You are a healthcare information retrieval agent. You retrieve information from tools and synthesize it into well-formatted markdown responses.
         
@@ -96,5 +96,6 @@ healthcare_agent = Agent[UserContext](
         - Accept useless RAG results without calling web search
         """,
 )
+healthcare_agent.description = "A healthcare agent that combines RAG (Retrieval Augmented Generation) with web search to answer medical questions."
 
 __all__ = ["healthcare_agent"]

common/aagents/news_agent.py

@@ -3,7 +3,6 @@ import os
 from agents import Agent, OpenAIChatCompletionsModel
 from dotenv import load_dotenv
 from mcp.tools.news_tools import get_top_headlines, search_news, get_news_by_category
-from mcp.tools.search_tools import duckduckgo_search
 from mcp.tools.time_tools import current_datetime
 from openai import AsyncOpenAI
 
@@ -25,7 +24,7 @@ groq_model = OpenAIChatCompletionsModel(model="groq/compound", openai_client=gro
 news_agent = Agent(
     name="NewsAgent",
     model=gemini_model,
-    tools=[current_datetime, get_top_headlines, search_news, get_news_by_category, duckduckgo_search],
+    tools=[current_datetime, get_top_headlines, search_news, get_news_by_category],
     instructions="""
         You are a NewsAgent specialized in fetching and analyzing recent news articles and headlines.
         Your role is to provide users with up-to-date, relevant news information from reliable sources.
@@ -46,13 +45,8 @@ news_agent = Agent(
            - Categories: "business", "entertainment", "general", "health", "science", "sports", "technology"
            - Input: { "category": "business", "country": "us", "num_results": 5 }
 
-        **FALLBACK TOOL (DuckDuckGo Search):**
-        4. 'duckduckgo_search': Use ONLY when NewsAPI tools fail or API key is missing
-           - Set search_type to "news" for news-specific results
-           - Input: { "query": "topic", "max_results": 5, "search_type": "news", "timelimit": "d" }
-
         **TIME CONTEXT:**
-        5. 'current_datetime': Use to provide current date/time context in your responses
+        4. 'current_datetime': Use to provide current date/time context in your responses
            - Input: { "format": "natural" }
 
         ## Workflow
@@ -62,14 +56,11 @@ news_agent = Agent(
            - Topic-specific → use search_news
            - Category-specific → use get_news_by_category
         
-        2. **Try Primary Tools First**: Always attempt NewsAPI tools before fallback
-        
-        3. **Fallback if Needed**: If NewsAPI returns an error (missing API key, no results), 
-           use duckduckgo_search with search_type="news"
+        2. **Execute Search**: Use the appropriate NewsAPI tool.
         
-        4. **Include Time Context**: Use current_datetime to provide temporal context
+        3. **Include Time Context**: Use current_datetime to provide temporal context.
         
-        5. **Format Response**: Present news in a clear, organized format with:
+        4. **Format Response**: Present news in a clear, organized format with:
            - Headlines/titles
            - Sources
            - Publication dates
@@ -95,12 +86,12 @@ news_agent = Agent(
 
         - Always cite sources and include publication dates
         - Prioritize recent news (within last 7 days unless specified otherwise)
-        - If API key is missing, inform the user and use the fallback tool
         - Never fabricate news or sources
         - Present news objectively without bias
         - Include URLs so users can read full articles
         - Use current_datetime to ensure temporal accuracy
         """,
 )
+news_agent.description = "A news agent that fetches top headlines and searches for news articles by category or topic."
 
-__all__ = ["news_agent", "get_top_headlines", "search_news", "get_news_by_category", "duckduckgo_search", "current_datetime"]
+__all__ = ["news_agent", "get_top_headlines", "search_news", "get_news_by_category", "current_datetime"]

common/aagents/search_agent.py

@@ -0,0 +1,51 @@
+"""Search agent module for comprehensive web searches."""
+import os
+from agents import Agent, OpenAIChatCompletionsModel
+from openai import AsyncOpenAI
+from dotenv import load_dotenv
+from mcp.tools.search_tools import duckduckgo_search, fetch_page_content
+from mcp.tools.time_tools import current_datetime
+
+# ---------------------------------------------------------
+# Load environment variables
+# ---------------------------------------------------------
+load_dotenv()
+
+GEMINI_BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
+google_api_key = os.getenv('GOOGLE_API_KEY')
+gemini_client = AsyncOpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)
+gemini_model = OpenAIChatCompletionsModel(model="gemini-2.0-flash-exp", openai_client=gemini_client)
+
+search_agent = Agent(
+    name="Web Search Agent",
+    model=gemini_model,
+    tools=[current_datetime, duckduckgo_search, fetch_page_content],
+    instructions="""
+        You are a highly efficient and specialized **Web Search Agent** 🌐. Your sole function is to retrieve and analyze information from the internet using the **duckduckgo_search** and **fetch_page_content** functions. You must act as a digital librarian and researcher, providing synthesized, cited, and up-to-date answers.
+
+        ## Core Directives & Priorities
+        1.  **Time Awareness First:** ALWAYS invoke **current_datetime** at the very beginning of your execution to establish the current temporal context. This is crucial for answering questions about "today", "yesterday", or recent events.
+        2.  **Search Strategy:**
+            *   Analyze the user's request and construct 1-3 targeted search queries.
+            *   Use **duckduckgo_search** to find relevant information. Use the 'news' type for current events.
+            *   **Mandatory Deep Dive:** You MUST select the **top 3** most relevant search results and use **fetch_page_content** to retrieve their full text. *Do not rely solely on the short search snippets.*
+        3.  **Synthesis & Answer Construction:**
+            *   Read the fetched content thoroughly.
+            *   Synthesize the information into a coherent answer.
+            *   **Conflict Resolution:** If sources disagree, note the discrepancy and favor the most recent or authoritative source.
+            *   **Citations:** You **must** cite your sources. At the end of your response, list the *Title* and *URL* of the pages you used.
+        4.  **Clarity:** Use professional, plain language. Use headings and bullet points for readability.
+        5.  **Data Gaps:** If you cannot find a conclusive answer after searching and fetching, state: **"A conclusive answer could not be verified by current web search results."**
+
+        ## Workflow Example
+        1.  Call `current_datetime()`.
+        2.  Call `duckduckgo_search(query="...")`.
+        3.  Loop through top 3 results: `fetch_page_content(url=...)`.
+        4.  Synthesize findings into final answer.
+
+        **Crucially, never fabricate information. Your answer must be grounded in the text you have fetched.**
+    """,
+)
+search_agent.description = "A web search agent that retrieves information using DuckDuckGo and fetches page content for detailed answers."
+
+__all__ = ["search_agent"]

common/aagents/weather_agent.py

@@ -32,7 +32,6 @@ groq_model = OpenAIChatCompletionsModel(model="groq/compound", openai_client=gro
 weather_agent = Agent(
     name="WeatherAgent",
     model=gemini_model, #"gpt-4o-mini",
-    # description="An agent that can perform web searches using DuckDuckGo.",
     tools=[current_datetime, get_weather_forecast, search_weather_fallback_ddgs, search_weather_fallback_bs],
     instructions="""
         You are a Weather Forecast agent who forecasts weather information ONLY.
@@ -61,9 +60,7 @@ weather_agent = Agent(
             }.
         ]
         """,
-    # output_type=AgentOutputSchema(list[searchResult], strict_json_schema=False),
-    # output_type=list[dict],  # safer than list[searchResult],    
-    # output_type=list[searchResult],
 )
+weather_agent.description = "A weather agent that provides current and forecasted weather information for specific cities."
 
 __all__ = ["weather_agent", "get_weather_forecast", "search_weather_fallback_ddgs", "search_weather_fallback_bs"]

common/aagents/web_research_agent.py

@@ -29,9 +29,7 @@ groq_client = AsyncOpenAI(base_url=GROQ_BASE_URL, api_key=groq_api_key)
 groq_model = OpenAIChatCompletionsModel(model="groq/compound", openai_client=groq_client)
 
 web_research_agent = Agent(
-    name="WebResearchAgent",
     model="gpt-4o-mini",
-    # description="An agent that can perform web searches using DuckDuckGo.",
     tools=[duckduckgo_search, fetch_page_content],
     instructions="""
 You are WebResearchAgent — an advanced internet research assistant with two core abilities:
@@ -79,5 +77,6 @@ IMPORTANT RULES
 """
     ,
 )
+web_research_agent.description = "A deep research agent that performs extensive web searches and content fetching for complex research queries."
 
 __all__ = ["web_research_agent", "duckduckgo_search", "fetch_page_content", "searchQuery", "searchResult"]

common/aagents/yf_agent.py

@@ -2,7 +2,7 @@
 import os
 from agents import Agent, OpenAIChatCompletionsModel
 from dotenv import load_dotenv
-from mcp.tools.yf_tools import get_summary, get_market_sentiment, get_history
+from mcp.tools.yf_tools import get_summary, get_market_sentiment, get_history, get_analyst_recommendations, get_earnings_calendar
 from mcp.tools.time_tools import current_datetime
 from openai import AsyncOpenAI
 
@@ -24,7 +24,7 @@ groq_model = OpenAIChatCompletionsModel(model="groq/compound", openai_client=gro
 yf_agent = Agent(
     name="YahooFinanceAgent",
     model=gemini_model,
-    tools=[current_datetime, get_summary, get_market_sentiment, get_history],
+    tools=[current_datetime, get_summary, get_market_sentiment, get_history, get_analyst_recommendations, get_earnings_calendar],
     instructions="""
         You are a specialized **Financial Analysis Agent** 💰, expert in market research, financial data retrieval, and market analysis. 
         Your primary role is to provide *actionable*, *data-driven*, and *concise* financial reports based on the available tools.
@@ -35,14 +35,16 @@ yf_agent = Agent(
            Financial data is extremely time-sensitive.
         
         2. **Financial Data Integrity:** Use the Yahoo Finance tools for specific stock/index data:
-           - 'get_summary': Get latest summary information and intraday price data for a ticker
-           - 'get_market_sentiment': Analyze recent price changes and provide market sentiment (Bullish/Bearish/Neutral)
-           - 'get_history': Fetch historical price data for a given ticker
+           - 'get_summary': Get latest summary information and intraday price data for a ticker.
+           - 'get_market_sentiment': Analyze recent price changes and provide market sentiment (Bullish/Bearish/Neutral).
+           - 'get_history': Fetch historical price data for a given ticker.
+           - 'get_analyst_recommendations': Fetch latest analyst ratings (Buy/Sell/Hold) for a symbol to provide **trading recommendations**.
+           - 'get_earnings_calendar': Fetch upcoming earnings dates for a symbol.
            
            Be precise about the date range and data source.
         
-        3. **Synthesis and Analysis:** Do not just list data. You must **synthesize** financial data (prices, volume, sentiment) 
-           to provide a complete analytical perspective (e.g., "Stock X is up 5% today driven by strong market momentum").
+        3. **Synthesis and Analysis:** Do not just list data. You must **synthesize** financial data (prices, volume, sentiment, recommendations) 
+           to provide a complete analytical perspective (e.g., "Stock X is up 5% today driven by strong market momentum and a generic 'Buy' rating from analysts").
         
         4. **Professional Clarity:** Present information in a clear, professional, and structured format. 
            Use numerical data and financial terminology correctly.
@@ -62,9 +64,12 @@ yf_agent = Agent(
         
         Tool: get_market_sentiment
         Input: { "symbol": "AAPL", "period": "1mo" }
-        
-        Tool: get_history
-        Input: { "symbol": "AAPL", "period": "1mo" }
+
+        Tool: get_analyst_recommendations
+        Input: { "symbol": "AAPL" }
+
+        Tool: get_earnings_calendar
+        Input: { "symbol": "AAPL" }
 
         ## Output Format Guidelines
         
@@ -74,5 +79,6 @@ yf_agent = Agent(
         * Always include a disclaimer: "This analysis is for informational purposes only and is not financial advice."
         """,
 )
+yf_agent.description = "A financial analysis agent that provides stock summaries, market sentiment, and historical data using Yahoo Finance."
 
-__all__ = ["yf_agent", "get_summary", "get_market_sentiment", "get_history", "current_datetime"]
+__all__ = ["yf_agent", "get_summary", "get_market_sentiment", "get_history", "get_analyst_recommendations", "get_earnings_calendar", "current_datetime"]

common/mcp/tools/time_tools.py

@@ -18,6 +18,7 @@ def current_datetime(format: str = "natural") -> str:
     Returns:
         str: Current date and time in the specified format
     """
+    print(f"[DEBUG] current_datetime called with format='{format}'")
     now = datetime.now()
     
     # Natural format options

common/mcp/tools/yf_tools.py

@@ -37,6 +37,7 @@ def get_summary(symbol: str, period: str = "1d", interval: str = "1h") -> str:
         - Volume
         - Period and interval used
     """
+    print(f"[DEBUG] get_summary called for symbol='{symbol}', period='{period}', interval='{interval}'")
     try:
         ticker = yf.Ticker(symbol)
 
@@ -109,6 +110,7 @@ def get_market_sentiment(symbol: str, period: str = "1mo") -> str:
     str
         A human-readable sentiment string including percentage change.
     """
+    print(f"[DEBUG] get_market_sentiment called for symbol='{symbol}', period='{period}'")
     try:
         ticker = yf.Ticker(symbol)
 
@@ -164,6 +166,7 @@ def get_history(symbol: str, period: str = "1mo") -> str:
     str
         A formatted string showing the last 5 rows of historical prices (Open, High, Low, Close, Volume).
     """
+    print(f"[DEBUG] get_history called for symbol='{symbol}', period='{period}'")
     try:
         ticker = yf.Ticker(symbol)
 
@@ -190,3 +193,65 @@ def get_history(symbol: str, period: str = "1mo") -> str:
 
     except Exception as e:
         return f"Error fetching historical data for '{symbol}': {e}"
+
+@function_tool
+def get_analyst_recommendations(symbol: str) -> str:
+    """
+    Fetch analyst recommendations for a given ticker.
+    
+    Parameters:
+    -----------
+    symbol : str
+        The ticker symbol.
+
+    Returns:
+    --------
+    str
+        Formatted string string of analyst recommendations.
+    """
+    print(f"[DEBUG] get_analyst_recommendations called for symbol='{symbol}'")
+    try:
+        ticker = yf.Ticker(symbol)
+        recs = ticker.recommendations
+        if recs is None or recs.empty:
+             return f"No analyst recommendations found for {symbol}."
+        
+        # Format the last few recommendations
+        latest = recs.tail(5)
+        return f"Analyst Recommendations for {symbol}:\n{latest.to_string()}"
+    except Exception as e:
+         return f"Error fetching recommendations for '{symbol}': {e}"
+
+@function_tool
+def get_earnings_calendar(symbol: str) -> str:
+    """
+    Fetch the next earnings date for a ticker.
+
+    Parameters:
+    -----------
+    symbol : str
+        The ticker symbol.
+
+    Returns:
+    --------
+    str
+        Next earnings date info.
+    """
+    print(f"[DEBUG] get_earnings_calendar called for symbol='{symbol}'")
+    try:
+        ticker = yf.Ticker(symbol)
+        calendar = ticker.calendar
+        if calendar is None:
+            return f"No earnings calendar found for {symbol}."
+        
+        # Handle dict (new yfinance) or DataFrame (old yfinance)
+        if isinstance(calendar, dict):
+             if not calendar:
+                 return f"No earnings calendar found for {symbol}."
+        elif hasattr(calendar, 'empty') and calendar.empty:
+             return f"No earnings calendar found for {symbol}."
+            
+        return f"Earnings Calendar for {symbol}:\n{calendar}"
+    except Exception as e:
+         return f"Error fetching earnings calendar for '{symbol}': {e}"
+

src/chatbot/aagents/input_validation_agent.py

@@ -0,0 +1,65 @@
+
+import os
+import json
+from agents import Agent, OpenAIChatCompletionsModel, Runner, GuardrailFunctionOutput
+from pydantic import BaseModel
+from openai import AsyncOpenAI
+from dotenv import load_dotenv
+
+load_dotenv()
+
+class ValidatedOutput(BaseModel):
+    is_valid: bool
+    reasoning: str
+
+GEMINI_BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
+google_api_key = os.getenv('GOOGLE_API_KEY')
+gemini_client = AsyncOpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)
+gemini_model = OpenAIChatCompletionsModel(model="gemini-2.0-flash", openai_client=gemini_client)
+
+input_validation_agent = Agent(
+    name="Guardrail Input Validation Agent",
+    instructions="""
+        You are a highly efficient and specialized **Agent** 🌐. Your sole function is to validate the user inputs.
+        
+        ## Core Directives & Priorities
+        1. You should flag if the user uses unparaliamentary language ONLY.
+        2. You MUST give reasoning for the same.
+        
+        ## Rules
+        - If it contains any of these, mark `"is_valid": false` and explain **why** in `"reasoning"`.
+        - Otherwise, mark `"is_valid": true` with reasoning like "The input follows respectful communication guidelines."
+
+
+        ## Output Format (MANDATORY)
+        * Return a JSON object with the following structure:
+        {
+            "is_valid": <boolean>,
+            "reasoning": <string>
+        }
+    """,
+    model=gemini_model,
+    output_type=ValidatedOutput,
+)
+input_validation_agent.description = "A guardrail agent that validates user input for unparliamentary language."
+
+async def input_validation_guardrail(ctx, agent, input_data):
+    result = await Runner.run(input_validation_agent, input_data, context=ctx.context)
+    raw_output = result.final_output
+
+    # Handle different return shapes gracefully
+    if isinstance(raw_output, ValidatedOutput):
+        final_output = raw_output
+        print("Parsed ValidatedOutput:", final_output)
+    else:
+        final_output = ValidatedOutput(
+            is_valid=False,
+            reasoning=f"Unexpected output type: {type(raw_output)}"
+        )
+
+    return GuardrailFunctionOutput(
+        output_info=final_output,
+        tripwire_triggered=not final_output.is_valid,
+    )
+
+__all__ = ["input_validation_agent", "input_validation_guardrail", "ValidatedOutput"]

src/chatbot/aagents/orchestrator_agent.py

@@ -0,0 +1,126 @@
+
+import os
+import asyncio
+from common.aagents.search_agent import search_agent
+from common.aagents.news_agent import news_agent
+from common.aagents.yf_agent import yf_agent
+from aagents.input_validation_agent import input_validation_guardrail
+from agents import Agent, OpenAIChatCompletionsModel, Runner, function_tool
+from openai import AsyncOpenAI
+from dotenv import load_dotenv
+
+load_dotenv()
+
+# --- Model setup ---
+GEMINI_BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
+google_api_key = os.getenv("GOOGLE_API_KEY")
+gemini_client = AsyncOpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)
+gemini_model = OpenAIChatCompletionsModel(
+    model="gemini-2.0-flash",
+    openai_client=gemini_client
+)
+
+# ----------------------------------------------------------
+# PARALLEL EXECUTION TOOL
+# ----------------------------------------------------------
+@function_tool
+async def broadcast_research(query: str) -> str:
+    """
+    Broadcasts the search query to multiple specialized agents (Finance, News, Web Search) 
+    in parallel and aggregates their responses.
+
+    Args:
+        query: The user's question or topic to research.
+
+    Returns:
+        Combined reports from all agents.
+    """
+    print(f"[DEBUG] broadcast_research called with query='{query}'")
+    
+    # Define tasks for parallel execution
+    # We use a new Runner.run call for each agent. 
+    # Note: We are not passing a session to keep them stateless/independent for this query.
+    
+    task_yf = Runner.run(yf_agent, query)
+    task_news = Runner.run(news_agent, query)
+    task_search = Runner.run(search_agent, query)
+    
+    # Run all in parallel
+    results = await asyncio.gather(task_yf, task_news, task_search, return_exceptions=True)
+    
+    yf_res, news_res, search_res = results
+    
+    # Access .final_output safely (handling potential exceptions)
+    def extract_output(res, name):
+        if isinstance(res, Exception):
+            return f"❌ {name} Error: {str(res)}"
+        return f"✅ {name} Report:\n{res.final_output}"
+
+    out_yf = extract_output(yf_res, "YahooFinanceAgent")
+    out_news = extract_output(news_res, "NewsAgent")
+    out_search = extract_output(search_res, "WebSearchAgent")
+    
+    combined_response = f"""
+    --- START OF AGENT REPORTS ---
+    
+    {out_yf}
+    
+    -----------------------------------
+    
+    {out_news}
+    
+    -----------------------------------
+    
+    {out_search}
+    
+    --- END OF AGENT REPORTS ---
+    """
+    return combined_response
+
+orchestrator_agent = Agent(
+    name="AI Market Research Orchestrator",
+    tools=[broadcast_research],
+    instructions="""
+    You are the **AI Market Research Orchestrator**. 
+    Your goal is to provide a comprehensive, multi-perspective answer by synthesizing data from specialized sub-agents.
+
+    **Workflow**:
+    1.  **Analyze Request**: Understand the user's question.
+    2.  **Broadcast Query**: IMMEDIATELY call the `broadcast_research` tool with a relevant search query.
+        *   This tool runs the Finance, News, and Web Search agents in parallel.
+    3.  **Synthesize Results**: Read the returned "Agent Reports".
+        *   Combine the financial data (prices, sentiment), news headlines, and general search context.
+        *   Compare and contrast findings if necessary.
+        *   Resolve conflicts by prioritizing specific data (e.g., Yahoo Finance for prices) over general text.
+    4.  **Final Response**: Generate a clear, professional, and well-structured summary for the user. Do not simply paste the individual reports.
+
+    **Final Response Structure (MANDATORY)**:
+    You MUST structure your final response exactly as follows:
+    
+    # [Market Analysis Title]
+    
+    ## 📊 Financial Snapshot
+    *   **Price/Sentiment**: [Synthesized from Yahoo Finance]
+    *   **Analyst Rating**: [Buy/Sell/Hold consensus]
+    
+    ## 📰 Key Developments
+    *   [Headline 1] - [Source]
+    *   [Headline 2] - [Source]
+    
+    ## 🔍 Web Insights
+    *   [Key finding from general search, if any]
+    
+    ## ⚖️ Synthesis & Recommendation
+    *   [Your comprehensive summary merging all data points. Highlight any conflicts.]
+    
+    **Constraint**:
+    *   Do NOT try to answer based on your own knowledge if live data is needed.
+    *   ALWAYS use `broadcast_research` for queries requiring up-to-date information.
+    *   If agents return "No data", explicitly state that in the relevant section.
+    """,
+    model=gemini_model,
+)
+orchestrator_agent.description = "An intelligent orchestrator that queries Finance, News, and Search agents in parallel and synthesizes a comprehensive response."
+
+
+__all__ = ["orchestrator_agent"]

src/chatbot/app.py

@@ -4,11 +4,14 @@ import glob
 import asyncio
 import sys
 import uuid
-
+from pathlib import Path
 # Add project root
-sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), ".")))
+# sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), ".")))
+# Add common directory to path for imports
+project_root = Path(__file__).resolve().parent.parent.parent
+sys.path.insert(0, str(project_root))
 
-from appagents.OrchestratorAgent import OrchestratorAgent
+from aagents.orchestrator_agent import orchestrator_agent
 from agents import Runner, trace, SQLiteSession
 from agents.exceptions import InputGuardrailTripwireTriggered
 
@@ -187,7 +190,7 @@ st.markdown("""
 # -----------------------------
 async def get_ai_response(prompt: str) -> str:
     try:
-        agent = OrchestratorAgent.create()
+        agent = orchestrator_agent
         # Ensure session is valid
         current_session = st.session_state.ai_session
         with trace("Chatbot Agent Run"):
@@ -216,6 +219,11 @@ with st.sidebar:
     for idx, prompt_text in enumerate(prompts):
         label = prompt_labels[idx] if idx < len(prompt_labels) else f"Prompt {idx+1}"
         if st.button(label, key=f"sidebar_btn_{idx}", use_container_width=True):
+            # Reset conversation
+            st.session_state.messages = []
+            st.session_state.ai_session_id = str(uuid.uuid4())
+            # Recreate session object with new ID
+            st.session_state.ai_session = SQLiteSession(f"conversation_{st.session_state.ai_session_id}.db")
             selected_prompt = prompt_text
 
     st.markdown("---")

src/chatbot/prompts/upcoming_earnings.txt

@@ -1 +1 @@
-Provide me the upcoming critical earnings in next 2 weeks.
+Search for upcoming critical earnings in the stock market.