Abstract

In this paper, we present an agentic AI application to do stock research is developed. This application is built in python using langgraph and langchain. It uses OpenAI LLM for decisioning. Current application recognizes only NSE stock symbols. It uses different web sources. It has 3 AI agents. First agent is for research which calls different tools to get stock information and sector information and news. Second agent does the analysis using the information collected by research agent. Third agent generates the report. Using langgraph, another alternative workflow is implemented where we have 7 specialized autonomous agents. That flow uses LLM as a tool. It is to optimize AI cost and rely more on code. It uses AI only for decision making.

Keywords: Agentic AI, langgraph, langchain

Introduction

As a part of Module 2 of Agentic AI Developer Certification, we need to build a Multi-Agent System. In this article, we will cover one of the multi-agent system built to solve a problem: stock research.

Problem Statement

Equity research report generation is a time-intensive process requiring expert analysts to manually gather data from multiple sources, perform complex financial calculations, analyze market trends, and synthesize findings into comprehensive reports. This process can take days or weeks, limiting the ability to respond quickly to market changes and making comprehensive analysis prohibitively expensive for many investors and institutions.

Additionally, the multi-faceted nature of stock analysis—encompassing financial metrics, management quality, technical indicators, valuation models, and sector dynamics—requires specialized expertise across various domains. Integrating insights from these diverse perspectives into a coherent investment thesis presents significant challenges.

Can we use AI to make the life of equity researchers easy?

Solution Overview

If we look traditional approach, first we collect the data about stock, sectors, economy and other domain specific knowledge as determined by the researcher. Then, collected data is analyzed using different analysis methods and tools. Once analysis is done, report is generated. As simple it seems, the process is iterative.

Here, we are trying to solve using an agentic AI system.

The Stock Report Generator addresses these challenges by automating the entire research and report generation pipeline through an autonomous multi-agent system. Each agent specializes in a particular aspect of analysis while maintaining the ability to collaborate and adapt based on emerging insights. The system combines:

• Autonomous Agent Coordination: Multiple specialized agents working in parallel and sequentially, each with distinct expertise
• Dynamic Tool Selection: Agents autonomously choose from multiple tools including stock data retrieval, web search, financial calculations, and report generation
• Real-time Data Integration: Live market data from Yahoo Finance and NSE validation
• Intelligent Content Generation: LLM-powered analysis, summarization, and professional report formatting
• Flexible Workflow Modes: Support for both iterative AI-driven research and structured traditional workflows

Important Disclaimer

This project is for learning purpose. Do not treat the output generated by the tool as stock recommendation. Please consult your investment adviser or research analysts. Any stock example shown is merely for education purpose, not investment recommendation.
This project uses yfinance python project which is intended for personal purpose only. Hence, do not use this project in production or commercial applications.

Methodology

Here, we discuss the solution being implemented to solve the problem.

System Architecture.

The Stock Report Generator tool employs a LangGraph-based orchestration framework that manages workflow state, agent coordination, and parallel execution. This generator works under two modes: Agentic AI Mode and Mixed Mode. This mode is controlled by the command line argument (-s or --skip-ai flag). By default, tool runs in Agentic AI Mode. By specifying option, it runs in Mixed Mode.

It works as follows:

  Stock symbol input --> Symbol validation --> Langgraph based Orchestration Execution --> Result

Currently, only NSE symbols are supported.

Agent Design Philosophy

Each agent in the system follows the BaseAgent architecture, which provides:

1. Autonomous Tool Selection: Agents analyze their tasks and autonomously select appropriate tools from available toolkits
2. State Management: Agents maintain and update shared state through the MultiAgentState Pydantic model
3. Error Handling: Robust error propagation and recovery mechanisms
4. Modular Design: Agents can be swapped, extended, or replaced without affecting the overall system

Agentic AI Mode

Under Agentic AI Mode, it has following langgraph structure:

In this mode, agents use iterative LLM-based reasoning to dynamically determine their research and analysis strategy:

Phase 1: AI Research (AIResearchAgent)

• Collects the data needed for analysis
• Iteratively decides which tools to call based on current information state
• Dynamically adjusts research depth based on data availability and quality
• Retrieves: stock metrics, company information, sector news, peer comparisons
• Adapts strategy as new information emerges

Phase 2: AI Analysis (AIAnalysisAgent)

• Performs comprehensive analysis across all dimensions (financial, management, technical, valuation)
• Uses iterative decision-making to determine analysis depth for each area
• Synthesizes insights across multiple analysis types
• Dynamically adjusts focus based on data completeness

Phase 3: AI Report Generation (AIReportAgent)

• LLM creates report content with professional formatting
• Tools handle PDF generation and styling
• Ensures all required sections are included
• Validates report completeness and coherence

Mixed Mode

Mixed mode uses AI but it does tool execution on its own. It relies on its own framework to execute the tools. It uses LLM only when needed. It still uses langgraph based workflow as shown below.

This mode follows a traditional research pipeline with ordered execution:

Phase 1: Research Planning (ResearchPlannerAgent)

• Analyzes research requirements
• Creates structured plan with ordered tool call sequence
• Outputs executable research plan

Phase 2: Data Gathering (ResearchAgent)

• Executes research plan sequentially
• Gathers company information, sector data, peer comparisons
• Retrieves financial metrics and market data

Phase 3: Parallel Analysis (4 Agents run concurrently)

• It runs analysis on the collected data.
• FinancialAnalysisAgent: Financial statement analysis, ratio computation, health assessment
• ManagementAnalysisAgent: Management effectiveness, governance, strategic initiatives
• TechnicalAnalysisAgent: Technical indicators, trend analysis, support/resistance levels
• ValuationAnalysisAgent: Valuation metrics, target price calculation, relative valuation

Phase 4: Report Synthesis (ReportAgent)

• Receives all analysis results
• Synthesizes comprehensive report
• Formats professional markdown
• Generates styled PDF output

This mode is introduced to save the LLM token.

State Management

The system uses a MultiAgentState (Pydantic model) that tracks:

• Stock information (symbol, company name, sector)
• Current agent execution context
• Research results from each phase
• Analysis results from each agent
• Final report content and PDF path
• Error logs and timestamps
• Workflow status

LangGraph's StateGraph ensures consistent state updates across concurrent and sequential agent executions, with automatic state merging and conflict resolution.

Tool Integration

Tools are implemented as LangChain-compatible callables that agents can autonomously invoke. Tools are categorized as:

1. Stock Data Tools: yfinance integration, NSE validation, company information retrieval
2. Web Search Tools: DuckDuckGo integration for news and market intelligence
3. Analysis Tools: Financial calculations, technical indicators, ratio computations
4. Text Processing Tools: Summarization, insight extraction, PDF parsing
5. Report Generation Tools: Markdown formatting, PDF generation, professional styling

Error Handling and Quality Assurance

The system implements multiple layers of error resilience:

1. Graceful Degradation: System continues with partial data when some sources fail
2. Automatic Retries: Transient failures trigger retry mechanisms
3. Fallback Analysis: AI-generated content when data sources are unavailable
4. Quality Validation: Consistency checking across agent outputs
5. Confidence Scoring: Agents rate the reliability of their analyses
6. Error Propagation: Errors are collected, deduplicated, and reported without stopping workflow

Parallel Execution

The orchestrator leverages LangGraph's parallel execution capabilities to run analysis agents concurrently, significantly reducing total execution time. State updates are safely merged using reducer functions that handle list concatenation and conflict resolution.

Installation & Setup

Prerequisites

Before installing the Stock Report Generator, ensure you have:

1. Python 3.10 or higher (Python 3.11+ recommended for best performance)
2. pip package manager (usually bundled with Python)
3. OpenAI API key (Get one here)
4. Git (for cloning the repository)
5. Optional: NVIDIA GPU with CUDA support for accelerated AI processing

System Requirements

• Operating System: macOS, Linux, or Windows
• RAM: Minimum 4GB (8GB+ recommended)
• Disk Space: At least 1GB free space
• Network: Internet connection for API calls and data retrieval

Installation Methods

Method 1: Pip Install (Recommended for Development)

This method installs the package in editable mode, allowing code modifications to be immediately reflected:

# Clone the repository
git clone https://github.com/devendermishra/stock-report-generator # or use ssh url.
cd stock-report-generator

# Create and activate virtual environment (highly recommended)
python3 -m venv venv
source venv/bin/activate  # On macOS/Linux
# venv\Scripts\activate  # On Windows

# Install in editable mode
pip install -e .

Method 2: Standard Installation

For standard installation without editable mode:

# Clone the repository
git clone https://github.com/devendermishra/stock-report-generator # or use ssh url
cd stock-report-generator

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On macOS/Linux
# venv\Scripts\activate  # On Windows

# Upgrade pip
pip install --upgrade pip

# Install dependencies
pip install -r requirements.txt

Environment Configuration

1. Copy the example environment file:

cp env.example .env

2. Edit the .env file with your API keys:

# On macOS/Linux:
nano .env
# or
vim .env

# On Windows:
notepad .env

3. Add your OpenAI API key:

OPENAI_API_KEY=sk-your-actual-api-key-here

4. Optional: Configure model settings:

DEFAULT_MODEL=gpt-4o-mini
MAX_TOKENS=4000
TEMPERATURE=0.1
OUTPUT_DIR=reports
TEMP_DIR=temp
MAX_REQUESTS_PER_MINUTE=60
REQUEST_DELAY=1.0

Post-Installation Setup

1. Create necessary directories:

mkdir -p reports temp data/inputs data/outputs

2. Verify installation:

# Check CLI help
python3 src/main.py --help

3. Run a test report:

python3 src/main.py RELIANCE

CLI Usage

After installation, you can use the CLI command:

# Generate report
stock-report RELIANCE

# Or use Python directly
python src/main.py RELIANCE

# With all parameters
python src/main.py RELIANCE "Reliance Industries Limited" "Oil & Gas"

# Using structured workflow (skip AI agents)
python src/main.py RELIANCE --skip-ai

Installation and usage issues

For any issue encountered during installation and issues, please raise using github issue using link

Experiments

We generated multiple stock reports (10 reports) using this tool.

Sample runs

SBIN Report in agentic mode

 % python3 src/main.py SBIN 
2025-11-02 17:18:06,142 - graph.multi_agent_graph - INFO - Building graph with use_ai_research=True, use_ai_analysis=True
2025-11-02 17:18:06,142 - graph.multi_agent_graph - INFO - Adding AI Research Agent node
2025-11-02 17:18:06,142 - graph.multi_agent_graph - INFO - Adding AI Analysis Agent node
2025-11-02 17:18:06,148 - __main__ - INFO - Stock Report Generator initialized successfully (Research: AI Research Agent, Analysis: AI Analysis Agent)
2025-11-02 17:18:06,148 - __main__ - INFO - Using AI Research Agent (iterative LLM-based research)
2025-11-02 17:18:06,148 - __main__ - INFO - Using AI Analysis Agent (iterative LLM-based comprehensive analysis)
Generating report for SBIN (SBIN) in unknown sector...
This may take a few minutes as the AI agents gather and analyze data...
2025-11-02 17:18:06,148 - __main__ - INFO - Starting report generation for SBIN
2025-11-02 17:18:06,148 - tools.stock_data_tool - INFO - Validating symbol SBIN.NS
2025-11-02 17:18:07,282 - tools.stock_data_tool - INFO - Fetching company info for SBIN.NS
2025-11-02 17:18:07,378 - tools.stock_data_tool - INFO - Successfully retrieved company info for SBIN.NS
2025-11-02 17:18:07,378 - graph.multi_agent_graph - INFO - Starting multi-agent workflow for SBIN
2025-11-02 17:18:07,380 - graph.multi_agent_graph - INFO - Executing AI research agent for SBIN
2025-11-02 17:18:09,978 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:09,996 - agents.base_agent.ai_research_agent - INFO - Iteration 1: Executing 4 tool(s): get_stock_metrics, get_company_info, search_sector_news, search_company_news
2025-11-02 17:18:09,998 - tools.stock_data_tool - INFO - Fetching company info for SBIN.NS
2025-11-02 17:18:09,999 - tools.stock_data_tool - INFO - Fetching stock metrics for SBIN.NS
2025-11-02 17:18:09,999 - tools.web_search_tool - INFO - Searching for sector news: Financial Services sector news India NSE market trends last week recent analysis outlook
2025-11-02 17:18:10,002 - tools.web_search_tool - INFO - Searching for company news: State Bank of India SBIN news India NSE announcements last week recent
2025-11-02 17:18:10,003 - tools.web_search_tool - INFO - Search attempt 1/3 for query: Financial Services sector news India NSE market trends last week recent analysis outlook
2025-11-02 17:18:10,003 - tools.web_search_tool - INFO - Search attempt 1/3 for query: State Bank of India SBIN news India NSE announcements last week recent
2025-11-02 17:18:10,003 - tools.web_search_tool - INFO - Trying search engine: mullvad_google
2025-11-02 17:18:10,003 - tools.web_search_tool - INFO - Trying search engine: mullvad_google
2025-11-02 17:18:10,411 - tools.stock_data_tool - INFO - Successfully retrieved company info for SBIN.NS
2025-11-02 17:18:10,542 - tools.stock_data_tool - INFO - Successfully retrieved metrics for SBIN.NS
2025-11-02 17:18:10,969 - primp - INFO - response: https://leta.mullvad.net/search/__data.json?q=Financial+Services+sector+news+India+NSE+market+trends+last+week+recent+analysis+outlook&engine=google&x-sveltekit-invalidated=001&country=us&lang=en 200
2025-11-02 17:18:10,972 - tools.web_search_tool - INFO - Successfully got 10 results using mullvad_google
2025-11-02 17:18:10,973 - tools.web_search_tool - INFO - Search completed successfully
2025-11-02 17:18:11,314 - primp - INFO - response: https://leta.mullvad.net/search/__data.json?q=State+Bank+of+India+SBIN+news+India+NSE+announcements+last+week+recent&engine=google&x-sveltekit-invalidated=001&country=us&lang=en 200
2025-11-02 17:18:11,316 - tools.web_search_tool - INFO - Successfully got 10 results using mullvad_google
2025-11-02 17:18:11,316 - tools.web_search_tool - INFO - Search completed successfully
2025-11-02 17:18:21,155 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:21,164 - agents.base_agent.ai_research_agent - INFO - Agent decided to finish after 2 iterations
2025-11-02 17:18:21,164 - agents.base_agent.ai_research_agent - INFO - AI Research Agent completed research for SBIN using 4 tools in 4 iterations
2025-11-02 17:18:21,165 - graph.multi_agent_graph - INFO - AI research agent completed for SBIN
2025-11-02 17:18:21,183 - graph.multi_agent_graph - INFO - Executing AI analysis agent for SBIN
2025-11-02 17:18:21,184 - graph.multi_agent_graph - INFO - Including research results for AI analysis agent
2025-11-02 17:18:21,184 - agents.base_agent.ai_analysis_agent - INFO - Extracted data from research: ['get_stock_metrics', 'get_company_info', 'search_company_news', 'search_market_trends']
2025-11-02 17:18:21,184 - agents.base_agent.ai_analysis_agent - INFO - Stock metrics available: True
2025-11-02 17:18:21,184 - agents.base_agent.ai_analysis_agent - INFO - Company info available: True
2025-11-02 17:18:31,845 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:31,850 - agents.base_agent.ai_analysis_agent - INFO - Agent decided to finish after 1 iterations
2025-11-02 17:18:34,303 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:36,924 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:39,358 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:39,362 - agents.base_agent.ai_analysis_agent - INFO - AI Analysis Agent completed analysis for SBIN using 0 tools in 0 iterations
2025-11-02 17:18:39,365 - graph.multi_agent_graph - INFO - Executing report agent for SBIN
2025-11-02 17:18:39,365 - graph.multi_agent_graph - INFO - Report agent received results from 4/4 analysis agents
2025-11-02 17:18:57,506 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:18:57,527 - tools.pdf_generator_tool - INFO - Starting PDF generation...
2025-11-02 17:18:57,539 - tools.pdf_generator_tool - INFO - Generating PDF: reports/stock_report_SBIN_20251102_171857.pdf
2025-11-02 17:18:57,656 - tools.pdf_generator_tool - INFO - PDF generated successfully: reports/stock_report_SBIN_20251102_171857.pdf
2025-11-02 17:18:57,658 - agents.base_agent.ai_report_agent - INFO - AI Report Agent completed report generation for SBIN
2025-11-02 17:18:57,659 - graph.multi_agent_graph - INFO - Report agent completed for SBIN
2025-11-02 17:18:57,756 - graph.multi_agent_graph - INFO - Completed multi-agent workflow for SBIN in 50.28s
2025-11-02 17:18:57,758 - __main__ - INFO - Successfully generated report for SBIN
2025-11-02 17:18:57,759 - __main__ - INFO - PDF report saved to: reports/stock_report_SBIN_20251102_171857.pdf

============================================================
REPORT GENERATION COMPLETED
============================================================
✅ Successfully generated report for SBIN
📊 Company: State Bank of India
🏢 Sector: Financial Services
⏱️  Duration: 50.28 seconds
📄 PDF Report: reports/stock_report_SBIN_20251102_171857.pdf

📋 Report Preview (first 500 characters):
----------------------------------------
# State Bank of India (SBIN) Stock Research Report
**Generated on:** 2025-11-02  
**Report Type:** Comprehensive Equity Research Analysis  
**Sector:** Financial Services  

## Executive Summary
State Bank of India (SBIN) is a leading player in the Indian banking sector, providing a wide range of financial services both domestically and internationally. The bank has demonstrated resilience and growth potential, supported by a strong market position and diversified offerings. 

### Key Highlights...
----------------------------------------

============================================================

When the symbol is invalid:

 % python3 src/main.py SBI
2025-11-02 17:16:38,747 - graph.multi_agent_graph - INFO - Building graph with use_ai_research=True, use_ai_analysis=True
2025-11-02 17:16:38,748 - graph.multi_agent_graph - INFO - Adding AI Research Agent node
2025-11-02 17:16:38,748 - graph.multi_agent_graph - INFO - Adding AI Analysis Agent node
2025-11-02 17:16:38,753 - __main__ - INFO - Stock Report Generator initialized successfully (Research: AI Research Agent, Analysis: AI Analysis Agent)
2025-11-02 17:16:38,753 - __main__ - INFO - Using AI Research Agent (iterative LLM-based research)
2025-11-02 17:16:38,753 - __main__ - INFO - Using AI Analysis Agent (iterative LLM-based comprehensive analysis)
Generating report for SBI (SBI) in unknown sector...
This may take a few minutes as the AI agents gather and analyze data...
2025-11-02 17:16:38,753 - __main__ - INFO - Starting report generation for SBI
2025-11-02 17:16:38,753 - tools.stock_data_tool - INFO - Validating symbol SBI.NS
2025-11-02 17:16:40,065 - yfinance - ERROR - $SBI.NS: possibly delisted; no price data found  (period=1d)
2025-11-02 17:16:40,065 - __main__ - ERROR - Report generation failed for SBI: No price data available

============================================================
REPORT GENERATION COMPLETED
============================================================
❌ Report generation failed for SBI
Errors encountered:
  - No price data available

============================================================

Using mixed for IRCTC, we got the following output:

 % python3 src/main.py IRCTC -s
2025-11-02 17:12:41,471 - graph.multi_agent_graph - INFO - Building graph with use_ai_research=False, use_ai_analysis=False
2025-11-02 17:12:41,471 - graph.multi_agent_graph - INFO - Adding Research Planner and Research Agent nodes
2025-11-02 17:12:41,471 - graph.multi_agent_graph - INFO - Adding separate analysis agent nodes
2025-11-02 17:12:41,481 - __main__ - INFO - Stock Report Generator initialized successfully (Research: Research Planner + Research Agent, Analysis: Separate Analysis Agents)
2025-11-02 17:12:41,481 - __main__ - INFO - Using Research Planner + Research Agent (structured workflow)
2025-11-02 17:12:41,481 - __main__ - INFO - Using separate Financial, Management, Technical, Valuation Analysis Agents
Generating report for IRCTC (IRCTC) in unknown sector...
This may take a few minutes as the AI agents gather and analyze data...
2025-11-02 17:12:41,481 - __main__ - INFO - Starting report generation for IRCTC
2025-11-02 17:12:41,482 - tools.stock_data_tool - INFO - Validating symbol IRCTC.NS
2025-11-02 17:12:42,579 - tools.stock_data_tool - INFO - Fetching company info for IRCTC.NS
2025-11-02 17:12:42,689 - tools.stock_data_tool - INFO - Successfully retrieved company info for IRCTC.NS
2025-11-02 17:12:42,689 - graph.multi_agent_graph - INFO - Starting multi-agent workflow for IRCTC
2025-11-02 17:12:42,691 - graph.multi_agent_graph - INFO - Executing research planner agent for IRCTC
2025-11-02 17:12:42,691 - agents.base_agent.research_planner_agent - INFO - Starting research planning for Indian Railway Catering & Tourism Corporation Limited (IRCTC)
2025-11-02 17:12:47,307 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Generated research plan with 6 tool calls
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 1: get_company_info (order: 1)
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 2: get_stock_metrics (order: 2)
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 3: search_sector_news (order: 3)
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 4: search_company_news (order: 4)
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 5: search_market_trends (order: 5)
2025-11-02 17:12:47,319 - agents.base_agent.research_planner_agent - INFO - Step 6: generic_web_search (order: 6)
2025-11-02 17:12:47,320 - agents.base_agent.research_planner_agent - INFO - Agent research_planner_agent completed task 'research_planning' for IRCTC in 4.63s (confidence: 0.80, tools: 0)
2025-11-02 17:12:47,320 - graph.multi_agent_graph - INFO - Research planner agent completed for IRCTC
2025-11-02 17:12:47,325 - graph.multi_agent_graph - INFO - Executing research agent for IRCTC
2025-11-02 17:12:47,325 - graph.multi_agent_graph - INFO - Including research plan for research agent
2025-11-02 17:12:47,325 - agents.base_agent.research_agent - INFO - Starting research for IRCTC (Indian Railway Catering & Tourism Corporation Limited) using research plan with 6 tools
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Executing research plan with 6 tool calls
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 1: get_company_info with parameters {'symbol': 'IRCTC'}
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 2: get_stock_metrics with parameters {'symbol': 'IRCTC'}
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 3: search_sector_news with parameters {'sector': 'Industrials', 'days_back': 7, 'include_analysis': True}
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 4: search_company_news with parameters {'company_name': 'Indian Railway Catering & Tourism Corporation Limited', 'stock_symbol': 'IRCTC', 'days_back': 7}
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 5: search_market_trends with parameters {'query': 'Indian Industrials sector trends', 'max_results': 5}
2025-11-02 17:12:47,326 - agents.base_agent.research_agent - INFO - Queuing tool 6: generic_web_search with parameters {'query': 'IRCTC peer comparison', 'search_type': 'web', 'max_results': 5}
2025-11-02 17:12:47,327 - agents.base_agent.research_agent - INFO - Executing planned tool [1]: get_company_info
2025-11-02 17:12:47,328 - tools.stock_data_tool - INFO - Fetching company info for IRCTC.NS
2025-11-02 17:12:47,537 - tools.stock_data_tool - INFO - Successfully retrieved company info for IRCTC.NS
2025-11-02 17:12:47,537 - agents.base_agent.research_agent - INFO - Executing planned tool [2]: get_stock_metrics
2025-11-02 17:12:47,538 - tools.stock_data_tool - INFO - Fetching stock metrics for IRCTC.NS
2025-11-02 17:12:47,720 - tools.stock_data_tool - INFO - Successfully retrieved metrics for IRCTC.NS
2025-11-02 17:12:47,721 - agents.base_agent.research_agent - INFO - Executing planned tool [3]: search_sector_news
2025-11-02 17:12:47,721 - tools.web_search_tool - INFO - Searching for sector news: Industrials sector news India NSE market trends last week recent analysis outlook
2025-11-02 17:12:47,722 - tools.web_search_tool - INFO - Search attempt 1/3 for query: Industrials sector news India NSE market trends last week recent analysis outlook
2025-11-02 17:12:47,722 - tools.web_search_tool - INFO - Trying search engine: mullvad_google
2025-11-02 17:12:48,405 - primp - INFO - response: https://leta.mullvad.net/search/__data.json?q=Industrials+sector+news+India+NSE+market+trends+last+week+recent+analysis+outlook&engine=google&x-sveltekit-invalidated=001&country=us&lang=en 200
2025-11-02 17:12:48,406 - tools.web_search_tool - INFO - Successfully got 10 results using mullvad_google
2025-11-02 17:12:48,407 - tools.web_search_tool - INFO - Search completed successfully
2025-11-02 17:12:48,407 - agents.base_agent.research_agent - INFO - Executing planned tool [4]: search_company_news
2025-11-02 17:12:48,407 - tools.web_search_tool - INFO - Searching for company news: Indian Railway Catering & Tourism Corporation Limited IRCTC news India NSE announcements last week recent
2025-11-02 17:12:48,407 - tools.web_search_tool - INFO - Search attempt 1/3 for query: Indian Railway Catering & Tourism Corporation Limited IRCTC news India NSE announcements last week recent
2025-11-02 17:12:48,407 - tools.web_search_tool - INFO - Trying search engine: mullvad_google
2025-11-02 17:12:48,784 - primp - INFO - response: https://leta.mullvad.net/search/__data.json?q=Indian+Railway+Catering+%26+Tourism+Corporation+Limited+IRCTC+news+India+NSE+announcements+last+week+recent&engine=google&x-sveltekit-invalidated=001&country=us&lang=en 200
2025-11-02 17:12:48,785 - tools.web_search_tool - INFO - Successfully got 10 results using mullvad_google
2025-11-02 17:12:48,785 - tools.web_search_tool - INFO - Search completed successfully
2025-11-02 17:12:48,785 - agents.base_agent.research_agent - INFO - Executing planned tool [5]: search_market_trends
2025-11-02 17:12:48,785 - tools.web_search_tool - INFO - Searching for market trends: Indian Industrials sector trends India NSE market analysis trends
2025-11-02 17:12:48,785 - tools.web_search_tool - INFO - Search attempt 1/3 for query: Indian Industrials sector trends India NSE market analysis trends
2025-11-02 17:12:48,786 - tools.web_search_tool - INFO - Trying search engine: mullvad_google
2025-11-02 17:12:48,955 - primp - INFO - response: https://leta.mullvad.net/search/__data.json?q=Indian+Industrials+sector+trends+India+NSE+market+analysis+trends&engine=google&x-sveltekit-invalidated=001&country=us&lang=en 200
2025-11-02 17:12:48,958 - tools.web_search_tool - INFO - Successfully got 10 results using mullvad_google
2025-11-02 17:12:48,958 - tools.web_search_tool - INFO - Search completed successfully
2025-11-02 17:12:48,958 - agents.base_agent.research_agent - INFO - Executing planned tool [6]: generic_web_search
2025-11-02 17:12:48,958 - agents.base_agent.research_agent - INFO - Peer data missing or empty in planned results, initializing structure
2025-11-02 17:12:48,959 - agents.base_agent.research_agent - INFO - Peer data missing after planned research, gathering peer data as fallback
2025-11-02 17:12:48,959 - agents.base_agent.research_agent - INFO - Gathering peer data for IRCTC in Industrials
2025-11-02 17:12:48,959 - agents.base_agent.research_agent - WARNING - No peer symbols found for sector: Industrials
2025-11-02 17:12:48,959 - agents.base_agent.research_agent - WARNING - No peer symbols found for sector Industrials
2025-11-02 17:12:48,959 - agents.base_agent.research_agent - INFO - Agent research_agent completed task 'planned_research' for IRCTC in 1.63s (confidence: 1.00, tools: 6)
2025-11-02 17:12:48,959 - graph.multi_agent_graph - INFO - Research agent completed for IRCTC
2025-11-02 17:12:48,961 - graph.multi_agent_graph - INFO - Executing financial analysis agent for IRCTC
2025-11-02 17:12:48,961 - agents.base_agent.financial_analysis_agent - INFO - Starting financial analysis for IRCTC (Indian Railway Catering & Tourism Corporation Limited)
2025-11-02 17:12:48,961 - agents.base_agent.financial_analysis_agent - INFO - Performing financial analysis for IRCTC
2025-11-02 17:12:48,961 - agents.base_agent.financial_analysis_agent - INFO - Reusing stock_metrics from ResearchAgent
2025-11-02 17:12:48,962 - graph.multi_agent_graph - INFO - Executing management analysis agent for IRCTC
2025-11-02 17:12:48,962 - agents.base_agent.management_analysis_agent - INFO - Starting management analysis for IRCTC (Indian Railway Catering & Tourism Corporation Limited)
2025-11-02 17:12:48,962 - agents.base_agent.management_analysis_agent - INFO - Performing management analysis for IRCTC
2025-11-02 17:12:48,962 - agents.base_agent.management_analysis_agent - INFO - Reusing company_info from ResearchAgent
2025-11-02 17:12:48,962 - agents.base_agent.management_analysis_agent - INFO - Reusing company_news from ResearchAgent
2025-11-02 17:12:48,962 - agents.base_agent.management_analysis_agent - INFO - Agent management_analysis_agent completed task 'management_analysis' for IRCTC in 0.00s (confidence: 0.70, tools: 0)
2025-11-02 17:12:48,962 - graph.multi_agent_graph - INFO - Management analysis agent completed for IRCTC
2025-11-02 17:12:48,962 - graph.multi_agent_graph - INFO - Executing technical analysis agent for IRCTC
2025-11-02 17:12:48,962 - agents.base_agent.technical_analysis_agent - INFO - Starting technical analysis for IRCTC (Indian Railway Catering & Tourism Corporation Limited)
2025-11-02 17:12:48,962 - agents.base_agent.technical_analysis_agent - INFO - Performing technical analysis for IRCTC
2025-11-02 17:12:48,962 - agents.base_agent.technical_analysis_agent - INFO - Reusing stock_metrics from ResearchAgent for technical analysis
2025-11-02 17:12:48,962 - agents.base_agent.technical_analysis_agent - INFO - Agent technical_analysis_agent completed task 'technical_analysis' for IRCTC in 0.00s (confidence: 0.70, tools: 0)
2025-11-02 17:12:48,962 - graph.multi_agent_graph - INFO - Technical analysis agent completed for IRCTC
2025-11-02 17:12:48,962 - graph.multi_agent_graph - INFO - Executing valuation analysis agent for IRCTC
2025-11-02 17:12:48,962 - agents.base_agent.valuation_analysis_agent - INFO - Starting valuation analysis for IRCTC (Indian Railway Catering & Tourism Corporation Limited)
2025-11-02 17:12:48,962 - agents.base_agent.valuation_analysis_agent - INFO - Performing valuation analysis for IRCTC
2025-11-02 17:12:48,963 - agents.base_agent.valuation_analysis_agent - INFO - Reusing stock_metrics from ResearchAgent for valuation analysis
2025-11-02 17:12:48,963 - agents.base_agent.valuation_analysis_agent - INFO - Reusing market_trends from ResearchAgent
2025-11-02 17:12:48,963 - agents.base_agent.valuation_analysis_agent - INFO - Agent valuation_analysis_agent completed task 'valuation_analysis' for IRCTC in 0.00s (confidence: 0.70, tools: 0)
2025-11-02 17:12:48,963 - graph.multi_agent_graph - INFO - Valuation analysis agent completed for IRCTC
2025-11-02 17:12:51,780 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:12:51,783 - agents.base_agent.financial_analysis_agent - INFO - Agent financial_analysis_agent completed task 'financial_analysis' for IRCTC in 2.82s (confidence: 0.70, tools: 0)
2025-11-02 17:12:51,783 - graph.multi_agent_graph - INFO - Financial analysis agent completed for IRCTC
2025-11-02 17:12:51,784 - graph.multi_agent_graph - INFO - Executing report agent for IRCTC
2025-11-02 17:12:51,784 - graph.multi_agent_graph - INFO - Report agent received results from 4/4 analysis agents
2025-11-02 17:12:51,784 - agents.base_agent.report_agent - INFO - Starting report generation for IRCTC (Indian Railway Catering & Tourism Corporation Limited)
2025-11-02 17:12:51,784 - agents.base_agent.report_agent - INFO - Generating executive summary for IRCTC
2025-11-02 17:12:51,784 - agents.base_agent.report_agent - INFO - Generating stock details for IRCTC
2025-11-02 17:12:51,784 - agents.base_agent.report_agent - INFO - Generating analysis sections
2025-11-02 17:12:51,784 - agents.base_agent.report_agent - INFO - Generating sector outlook section for sector: Industrials
2025-11-02 17:12:51,785 - agents.base_agent.report_agent - INFO - Generating peer analysis section for IRCTC
2025-11-02 17:12:51,785 - agents.base_agent.report_agent - WARNING - No peer data found for IRCTC
2025-11-02 17:12:51,785 - agents.base_agent.report_agent - INFO - Generating recommendations section
2025-11-02 17:12:51,785 - agents.base_agent.report_agent - INFO - Generating technical analysis section
2025-11-02 17:12:57,604 - httpx - INFO - HTTP Request: POST https://api.openai.com/v1/chat/completions "HTTP/1.1 200 OK"
2025-11-02 17:12:57,608 - agents.base_agent.report_agent - INFO - Generating final report for IRCTC
2025-11-02 17:12:57,609 - agents.base_agent.report_agent - INFO - Generating PDF report for IRCTC
2025-11-02 17:12:57,610 - tools.pdf_generator_tool - INFO - Starting PDF generation...
2025-11-02 17:12:57,616 - tools.pdf_generator_tool - INFO - Generating PDF: reports/stock_report_IRCTC_20251102_171257.pdf
2025-11-02 17:12:57,660 - tools.pdf_generator_tool - INFO - PDF generated successfully: reports/stock_report_IRCTC_20251102_171257.pdf
2025-11-02 17:12:57,660 - agents.base_agent.report_agent - INFO - Agent report_agent completed task 'comprehensive_report_generation' for IRCTC in 5.88s (confidence: 0.70, tools: 0)
2025-11-02 17:12:57,660 - graph.multi_agent_graph - INFO - Report agent completed for IRCTC
2025-11-02 17:12:57,661 - graph.multi_agent_graph - INFO - Completed multi-agent workflow for IRCTC in 14.97s
2025-11-02 17:12:57,661 - __main__ - INFO - Successfully generated report for IRCTC
2025-11-02 17:12:57,662 - __main__ - INFO - PDF report saved to: reports/stock_report_IRCTC_20251102_171257.pdf

============================================================
REPORT GENERATION COMPLETED
============================================================
✅ Successfully generated report for IRCTC
📊 Company: Indian Railway Catering & Tourism Corporation Limited
🏢 Sector: Industrials
⏱️  Duration: 14.97 seconds
📄 PDF Report: reports/stock_report_IRCTC_20251102_171257.pdf

📋 Report Preview (first 500 characters):
----------------------------------------
# Stock Research Report: Indian Railway Catering & Tourism Corporation Limited (IRCTC)
**Generated on:** 2025-11-02 17:12:57
**Report Type:** Comprehensive Equity Research Analysis
---
## Executive Summary

**Indian Railway Catering & Tourism Corporation Limited (IRCTC)** is a Industrials sector company trading at ₹718.70 with a market capitalization of ₹575,399,985,152.

**Key Metrics:**
- Current Price: ₹718.70
- Market Cap: ₹575,399,985,152
- P/E Ratio: 42.91
- Target Price: ₹646.83
- Recomme...
----------------------------------------

============================================================

Some of the report samples are attached. However, please note that these are not actual investment analysis or any recommendation. These are only for educational and personal purpose, not for commercial purpose.

Fine-tuning and Other Parameters

Model Configuration

The system uses OpenAI's GPT models with configurable parameters that can be adjusted in src/config.py or through environment variables:

Default Model Settings

DEFAULT_MODEL = "gpt-4o-mini"  # Cost-effective, high-performance model
MAX_TOKENS = 4000               # Maximum tokens per LLM request
TEMPERATURE = 0.1               # Low temperature for consistent outputs

Model Selection Rationale

• GPT-4o-mini: Selected as the default model for optimal balance between cost and performance

  • Faster inference compared to GPT-4
  • Lower API costs for large-scale usage
  • Maintains high-quality outputs for financial analysis
  • Supports the token limits required for comprehensive reports

• Alternative Models: The system can be configured to use other OpenAI models:

  • gpt-4o: Higher quality but increased cost and latency
  • gpt-4-turbo: Enhanced reasoning for complex analysis
  • gpt-3.5-turbo: Lower cost option with reduced quality

Temperature Parameter

Current Setting: 0.1

Temperature controls the randomness of LLM outputs:

• 0.1 (Low): Ensures consistent, deterministic outputs suitable for financial analysis
• 0.3-0.5 (Medium): More creative but still focused
• 0.7-1.0 (High): Creative and diverse outputs (not recommended for financial reports)

Recommendation: Keep temperature at 0.1-0.2 for financial analysis to ensure:

• Consistent numerical reporting
• Reproducible analysis results
• Professional report formatting

Token Limits

Current Setting: MAX_TOKENS = 4000

This parameter controls the maximum response length:

• 4000 tokens: Sufficient for comprehensive analysis sections
• 8000 tokens: For very detailed reports or longer sections
• 16000 tokens: Maximum for GPT-4o-mini, useful for extensive analysis

Adjustment Guidelines:

• Increase if reports are being truncated
• Decrease if cost optimization is needed
• Monitor token usage in logs to optimize

Agent-Specific Parameters

Research Agents

ResearchPlannerAgent:

• Max iterations: Configurable research plan depth
• Tool selection strategy: Prioritizes stock data tools, then web search
• Timeout: 300 seconds per research task

AIResearchAgent (Iterative Mode):

• Max tool calls per iteration: 5-10 depending on context
• Iteration limit: 3-5 iterations before finalizing
• Context window: Up to 10,000 entries for state management

Results

In this project, we have implemented an agentic AI system consisting of multiple AI agents (3) and uses multiple tools (at least 5) to generate the stock research report. Using langgraph, clear communication and co-ordination between the agents is established.

As the submission requirement, it meets every requirement.

✅ 1. Multi-Agent System

1.1 Multi-Agent Collaboration

• Total Agents: 3 Agentic AI
• Mixed Mode:
  • ResearchPlannerAgent
  • ResearchAgent
  • FinancialAnalysisAgent
  • ManagementAnalysisAgent
  • TechnicalAnalysisAgent
  • ValuationAnalysisAgent
  • ReportAgent
• Agentic AI Agents:
  • AIResearchAgent
  • AIAnalysisAgent
  • AIReportAgent
• Modes:
  • Agentic AI Mode: AIResearch → AIAnalysis → AIReport
  • Mixed Mode: Planner → Research → Analysis Agents (4) → Report

Status: ✅ Meets requirement

1.2 Agent Communication & Coordination

• Agents share data via MultiAgentState (Pydantic model)
• Communication handled by LangGraph state passing
• Parallel analysis supported; errors handled via reducers
• Evidence in multi_agent_graph.py (lines 43–601)

Status: ✅ Meets — Clear state-based coordination with structured data flow

---

1.3 Orchestration Framework

• Uses LangGraph for orchestration
• MultiAgentOrchestrator class handles dynamic workflows (AI vs Structured)
• Supports sequential & parallel execution, conditional routing, and error handling

Status: ✅ Meets — Advanced orchestration using LangGraph

✅ 2. Tool Integration

2.1 Tool Count and Variety

• Integrated Tools: 5+ total
  • Stock Data Tools:
    • get_stock_metrics
    • get_company_info
    • validate_symbol
  • Web Search Tools:
    • search_sector_news
    • search_company_news
    • search_market_trends
    • generic_web_search
    • search_web_generic
  • Analysis & Calculation Tools:
    • TechnicalAnalysisFormatter
    • StockDataCalculator
  • Report Generation Tools:
    • PDFGeneratorTool
    • ReportFormatterTool
  • Text Processing Tools:
    • SummarizerTool
    • extract_insights
    • summarize_text
  • Additional Tools:
    • PDFParserTool
    • ReportFetcherTool
    • report_formatter_helpers
    • report_formatter_utils
    • stock_data_models
    • stock_data_validator

Status: ✅ Meets — 5+ tools integrated (well beyond 3 minimum)

---

2.2 Tool Implementation Types

• Mix of LangChain built-in and custom tools
  • LangChain Tools: Defined via @tool decorator
  • Custom Tools: PDFGeneratorTool, ReportFormatterTool, SummarizerTool, TechnicalAnalysisFormatter, StockDataCalculator

Status: ✅ Meets — Hybrid approach (custom + built-in)

---

2.3 Tools Extend Beyond LLM Capabilities

All tools extend functionality beyond standard LLM responses:

• Stock Data Tools: Real-time API integration for live data
• Web Search Tools: Fetch internet-based news, trends, and reports
• PDF Tools: Generate and parse professional PDF reports
• Summarizer Tool: Long-text analysis, insight extraction, structured summaries
• Analysis Tools: Financial and technical computations (RSI, MACD, P/E ratios)
• Report Tools: Generate 8–12 page professional markdown/PDF reports

Status: ✅ Meets — Tools enhance LLM with live data, computation, and I/O operations

---

✅ Overall Verification

All system requirements are met.

Future works

• Add other exchanges (NYSE, NASDAQ etc.).
• Add other countries sectors
• Add geopolitical analysis of the company
• Add LangSmith for the monitoring
• Add evaluations
• Integrate MCP
• Add one-shot examples in the prompts
• Add detailed analysis of peer
• Add config support:
  • Model parameters (model name, token limit, temperature for each step)
• Add other models (Groq, Gemini)
• Add text prompt input instead of stock symbol
• Create a web interface

Conclusion

This document provides comprehensive coverage of the Stock Report Generator system, from high-level architecture to detailed parameter configuration. The system's flexibility allows for extensive customization while maintaining robust default settings optimized for most use cases. This tool is to make life of research analysts easy. Report generated by this tool can be used as a starting point. This tool can be modified to improve the quality of research report.

For additional information, refer to:

• README.md for quick start guide
• Source code documentation for API references

Availability

The complete system is available as open-source software at:

Repository: https://github.com/devendermishra/stock-report-generator
Documentation: Comprehensive README and API documentation
License: MIT License for academic and commercial use
Support: Active community support and issue tracking

Acknowledgments: We thank the LangChain community, the open-source contributors who have made this work possible. Special thanks to OpenAI for providing access to their LLM APIs for research and development.

Funding: This work was supported by the open-source community and individual contributions.

Conflicts of Interest: The authors declare no conflicts of interest.

Data Availability: All code and documentation are publicly available under the MIT License.