Update app.py

app.py

@@ -1,7 +1,6 @@
 import os
 import nltk
 import requests
-import datetime
 
 # Use a directory within the user's home directory
 nltk_data_dir = os.path.expanduser("~/.nltk_data")
@@ -10,7 +9,6 @@ nltk.data.path.append(nltk_data_dir)
 
 # Download NLTK data
 nltk.download('punkt', download_dir=nltk_data_dir, quiet=True)
-
 import chainlit as cl
 from llama_index.core import VectorStoreIndex, Document
 from llama_index.embeddings.huggingface import HuggingFaceEmbedding
@@ -23,40 +21,45 @@ import pandas as pd
 
 load_dotenv()
 
-# We won't print your keys to avoid exposing them
+# Fetch the API keys from environment variables
 GROQ_API_KEY = os.getenv("GROQ_API_KEY")
 FMP_API_KEY = os.getenv("FMP_API_KEY")
 
+# Initialize models
 embed_model = HuggingFaceEmbedding(model_name="sentence-transformers/all-MiniLM-L6-v2")
 llm = Groq(model="llama3-70b-8192", api_key=GROQ_API_KEY)
 
+# Create service context
 service_context = ServiceContext.from_defaults(
     llm=llm,
     embed_model=embed_model,
     node_parser=SentenceSplitter(chunk_size=1000, chunk_overlap=200)
 )
 
-def fetch_annual_report_10k(symbol: str) -> str:
+def fetch_earnings_transcript(symbol: str) -> str:
     """
-    Fetches the latest available 10-K filing from FMP's SEC Filings endpoint.
-    Mirrors the approach in the working 'earnings transcript' code.
+    Fetch the latest transcript for a company's earnings call.
+    Args:
+    - symbol (str): The stock ticker symbol (e.g., 'AAPL').
+    Returns:
+    - str: The earnings call transcript or an error message.
     """
-    sec_url = (
-        f"https://financialmodelingprep.com/api/v4/sec_filings"
-        f"?symbol={symbol}&type=10-K&apikey={FMP_API_KEY}"
-    )
+    transcript_url = f"https://financialmodelingprep.com/api/v3/earning_call_transcript/{symbol}?apikey={FMP_API_KEY}"
+    
     try:
-        response = requests.get(sec_url, timeout=10)
+        response = requests.get(transcript_url, timeout=10)
         response.raise_for_status()
-        filings_data = response.json()
-
-        if not filings_data:
-            return f"No 10-K data available for {symbol}."
-
-        # Like your earnings code, pick the first record
-        # Return it as a string for indexing
-        first_filing = filings_data[0]
-        return str(first_filing)
+        transcript_data = response.json()
+        
+        if not transcript_data:
+            return f"No transcript available for {symbol}."
+        
+        # Extract the first available transcript
+        latest_transcript = transcript_data[0].get("content", "")
+        if not latest_transcript:
+            return f"No transcript content found for {symbol}."
+        
+        return latest_transcript
 
     except requests.exceptions.HTTPError as http_err:
         return f"HTTP error occurred: {http_err}"
@@ -65,84 +68,78 @@ def fetch_annual_report_10k(symbol: str) -> str:
     except Exception as err:
         return f"An unexpected error occurred: {err}"
 
-# Prompts remain unchanged
+# Prompts
 summary_prompt = (
-    "You are a world-class financial analyst with extensive experience analyzing annual reports. "
-    "Provide a comprehensive summary of the 10-K report. Focus on Strategic Insights, Key Financial Figures, and Risk Factors. "
-    "Answer in extensive bullet points, summarizing the company's performance, strengths, and weaknesses."
+    "You are a world-class financial analyst with extensive experience analyzing quarterly reports. "
+    "Give me a comprehensive summary of the earnings report. Focus on the Strategic Insights and Key Financial Figures. "
+    "Answer in extensive bullet points please."
 )
 
 question_prompt = (
-    "You are a financial analyst with extensive experience analyzing annual reports. "
-    "Read the 10-K report and generate 10 strategic questions focusing on the company's performance, risks, and financial figures. "
-    "Ask questions that provide strategic insights into the company's long-term goals, revenue trends, competitive position, and more. "
-    "Format the questions as a numbered list (e.g., '1. Question')."
+    "You are a financial analyst with extensive experience analyzing quarterly reports. "
+    "Read the earnings call transcript and earnings presentation report and generate 10 questions focusing on the strategic insights and financial figures. "
+    "Ask questions that require precise answers and provide strategic insight into the company's financial and strategic performance, such as revenue growth, market trends, profit margins, and more. "
+    "Only ask questions that can be answered using the provided document, without making any assumptions or inferences beyond the text. "
+    "Please format the questions as a list with a simple '1. Question 1', '2. Question 2', etc. structure. "
+    "Unless retrievable from the documents, don't ask questions which cannot be compared to previous periods."
 )
 
 @cl.on_chat_start
 async def on_chat_start():
-    # Ask the user for a ticker symbol
     ticker_response = await cl.AskUserMessage(
         content=(
-            "This tool is designed to analyze 10-K annual reports for publicly traded companies. "
-            "Provide the company's ticker symbol, and the tool will fetch the latest available 10-K report. "
-            "It generates summaries and strategic due diligence. Ask your own questions afterwards.\n\n"
+            "This tool is designed to analyze earnings call transcripts for publicly traded companies. "
+            "Provide the company's ticker symbol, and the tool will fetch the latest earnings call transcript. "
+            "It generates summaries and strategic due diligence. Ask your own questions afterwards. \n\n"
             "Please enter the ticker symbol for the company you want to analyze (e.g. MSFT):"
         )
     ).send()
 
-    if not ticker_response or 'content' not in ticker_response:
-        await cl.Message(content="No ticker symbol provided. Please enter a valid ticker symbol to proceed.").send()
-        return
-
     ticker_symbol = ticker_response['content'].upper()
-    msg = cl.Message(content=f"Retrieving the latest 10-K report for {ticker_symbol}...")
+
+    msg = cl.Message(content=f"Retrieving earnings call transcript for {ticker_symbol}...")
     await msg.send()
 
     try:
-        # Fetch the 10-K, just like your working approach for earnings
-        annual_report_text = fetch_annual_report_10k(ticker_symbol)
-
-        # Check if there's an immediate error
-        if annual_report_text.startswith("No 10-K data") or \
-           annual_report_text.startswith("HTTP error") or \
-           annual_report_text.startswith("Request error") or \
-           annual_report_text.startswith("An unexpected error occurred"):
-            await cl.Message(content=annual_report_text).send()
+        # Fetch the transcript using FMP API
+        transcript_text = fetch_earnings_transcript(ticker_symbol)
+
+        # Check if an error message was returned
+        if transcript_text.startswith("No transcript") or \
+           transcript_text.startswith("HTTP error") or \
+           transcript_text.startswith("Request error") or \
+           transcript_text.startswith("An unexpected error occurred"):
+            await cl.Message(content=transcript_text).send()
             return
 
-        # Create a Document from the raw JSON/string
-        document = Document(text=annual_report_text, metadata={"company": ticker_symbol})
+        # Create a Document object with the transcript text
+        document = Document(text=transcript_text, metadata={"company": ticker_symbol})
 
-        # Build the index
-        index = VectorStoreIndex.from_documents([document], service_context=service_context)
+        # Create index
+        index = VectorStoreIndex.from_documents(
+            [document], service_context=service_context
+        )
 
-        # Save index to user session
+        # Store the index in the user session
         cl.user_session.set("index", index)
 
-        # Summaries
+        # Generate summary
         query_engine = index.as_query_engine()
         summary_response = await cl.make_async(query_engine.query)(summary_prompt)
-        summary_text = str(summary_response)
-        await cl.Message(content=f"**Summary:**\n{summary_text}").send()
+        await cl.Message(content=f"**Summary:**\n{summary_response}").send()
 
-        # Generated questions
+        # Generate questions
         questions_response = await cl.make_async(query_engine.query)(question_prompt)
-        questions_text = str(questions_response)
-        questions_format = questions_text.split('\n')
-        relevant_questions = [
-            question.strip() for question in questions_format
-            if question.strip() and question.strip()[0].isdigit()
-        ]
-
-        await cl.Message(content="Generated strategic questions and answers:").send()
+        questions_format = str(questions_response).split('\n')
+        relevant_questions = [question.strip() for question in questions_format if question.strip() and question.strip()[0].isdigit()]
+
+        # Answer generated questions
+        await cl.Message(content="Generated questions and answers:").send()
         for question in relevant_questions:
-            await cl.Message(content=f"**{question}**").send()
-            answer_obj = await cl.make_async(query_engine.query)(question)
-            answer_text = str(answer_obj)
-            await cl.Message(content=f"**Answer:**\n{answer_text}").send()
+            response = await cl.make_async(query_engine.query)(question)
+            await cl.Message(content=f"**{question}**\n{response}").send()
 
-        msg.content = "Processing done. You can now ask more questions about the 10-K report!"
+        msg.content = "Processing done. You can now ask more questions about the earnings call transcript!"
         await msg.update()
 
     except Exception as e:
@@ -151,22 +148,17 @@ async def on_chat_start():
 @cl.on_message
 async def main(message: cl.Message):
     index = cl.user_session.get("index")
+    
     if index is None:
         await cl.Message(content="Please provide a ticker symbol first before asking questions.").send()
         return
 
     query_engine = index.as_query_engine()
-    user_query = message.content
-    try:
-        raw_response = await cl.make_async(query_engine.query)(user_query)
-        response_str = str(raw_response)
-
-        # Stream the response token-by-token
-        response_message = cl.Message(content="")
-        for token in response_str:
-            await response_message.stream_token(token=token)
-
-        await response_message.send()
-
-    except Exception as e:
-        await cl.Message(content=f"Error while processing your question: {str(e)}").send()
+    
+    response = await cl.make_async(query_engine.query)(message.content)
+    
+    response_message = cl.Message(content="")
+    for token in str(response):
+        await response_message.stream_token(token=token)
+    
+    await response_message.send()