Multi-agent business intelligence orchestration system powered by GPT-5/Deepseek v3.2exp, LangGraph, and research-augmented generation.

LangGraph BI Agent Orchestrator

Author: Raja Hussain
Built with: Python, LangGraph, FastAPI, Docker, DeepSeek, GPT-5
Status: MVP-Ready
Last Updated: December 6, 2025

TLDR

• Coordinates 5 specialized AI agents with LangGraph to answer business questions
• Retrieves academic papers from Semantic Scholar and arXiv with proper citations
• Routes queries between DeepSeek and GPT-5 to cut costs by 86%
• Auto-generates PowerPoint decks and Excel workbooks from single queries
• 2.1x speedup with parallel execution, 138x speedup on cache hits

Table of Contents

• Overview
• Quick Start
• System Architecture
• Key Features
• Performance Metrics
• Technology Stack
• Project Structure
• Configuration
• API Usage
• Testing
• Deployment
• Cost Analysis
• Troubleshooting
• Development
• Roadmap

Overview

This system coordinates multiple specialized AI agents to provide comprehensive business analysis backed by academic research. It automatically generates three deliverables from a single query:

1. Structured JSON - Machine-readable data for API integrations
2. PowerPoint Presentation - Professional executive summary decks (10-12 slides)
3. Excel Workbook - Analysis spreadsheets with