File size: 3,193 Bytes
c3bf538 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 |
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
from typing import Dict, Any
import json
def generate_investment_thesis(full_job_result: Dict[str, Any]) -> str:
"""
Uses the Gemini 1.5 Flash model to generate a qualitative investment thesis.
"""
print("Generating investment thesis with Gemini 1.5 Flash...")
# Initialize the LLM
llm = ChatGoogleGenerativeAI(model="gemini-1.5-flash")
# Create a simplified summary of the data to pass to the LLM
# This prevents sending thousands of characters of raw data
fundamentals_summary = (
f"Company: {full_job_result.get('company_name', 'N/A')}\n"
f"Current Price: {full_job_result.get('current_price', 'N/A')}\n"
f"Market Cap: {full_job_result.get('market_cap', 'N/A')}\n"
f"P/E Ratio: {full_job_result.get('pe_ratio', 'N/A'):.2f}\n"
f"Sector: {full_job_result.get('sector', 'N/A')}"
)
prediction_summary = full_job_result.get('prediction_analysis', {}).get('summary', 'No prediction summary available.')
# We need to handle the case where intelligence gathering might have failed
intelligence_briefing = full_job_result.get('intelligence_briefing', {})
if intelligence_briefing and intelligence_briefing.get('news'):
news_summary = ", ".join([f"'{article['title']}' ({article['sentiment']})" for article in intelligence_briefing['news'][:2]])
else:
news_summary = "No news articles found."
# Define the prompt template
prompt = PromptTemplate(
input_variables=["fundamentals", "prediction", "news"],
template="""
You are a sharp, concise senior financial analyst for an Indian market-focused fund.
Your task is to provide a clear investment thesis based on the data provided.
Do not offer financial advice. Analyze the data objectively.
**Data Overview:**
- **Fundamentals:** {fundamentals}
- **Quantitative Forecast:** {prediction}
- **Recent News Headlines & Sentiment:** {news}
**Your Analysis (in Markdown format):**
**1. Executive Summary:** A 2-sentence summary of the company's current situation based on the data.
**2. Bull Case:** 2-3 bullet points on the positive signals from the data.
**3. Bear Case:** 2-3 bullet points on the primary risks or negative signals.
**4. Final Recommendation:** State ONE of the following: 'Strong Buy', 'Buy', 'Hold', 'Sell', or 'Strong Sell' and provide a brief 1-sentence justification based purely on the provided data mix.
"""
)
# Create the LangChain chain
chain = LLMChain(llm=llm, prompt=prompt)
# Run the chain with our summarized data
try:
response = chain.run(
fundamentals=fundamentals_summary,
prediction=prediction_summary,
news=news_summary
)
print("Successfully generated thesis from Gemini.")
return response
except Exception as e:
print(f"Error calling Gemini API: {e}")
return "Error: Could not generate the advisor summary." |