🧠 ContextLens Agentic Analyst

A LangGraph-based, agentic AI data analyst for Excel-driven exploratory analysis, reasoning, and reporting

---

🚀 Project Overview

ContextLens Agentic Analyst is a context-aware, agentic AI system that transforms raw Excel spreadsheets into meaningful insights.
Built using LangGraph, it models how a human data analyst thinks — understanding context, deciding what analysis to run, answering reasoning-based questions, and generating business-ready reports.

This is not just automation — it’s an agentic analytical workflow.

---

✨ Key Capabilities

• 📂 Excel Ingestion (single or multiple sheets)
• 🧭 Context Understanding (plain-English dataset description)
• 📊 Automated Exploratory Data Analysis (EDA)
  • distributions, boxplots, trends, correlations
• 💬 Reasoning-Based Q&A Chatbot
  • ask analytical questions in layman language
• 📄 PDF Report Generation
  • charts + concise explanations
• 🧠 Agentic Workflow
  • stateful, explainable, extensible

---

🤖 Why This Is an Agentic System

Unlike scripts or dashboards, this system:

• Maintains state across steps
• Interprets user intent from context and questions
• Selects analyses dynamically instead of running everything blindly
• Separates ingestion, analysis, reasoning, and reporting as agent steps

Powered by LangGraph, the workflow mirrors how a real analyst operates.

---

🧩 How It Works (High-Level)

1. Upload an Excel file
2. Describe the dataset context in simple terms
3. The agent:
   • profiles the data
   • identifies key parameters
   • performs baseline EDA
4. Ask reasoning-based questions via chatbot
5. Generate a structured PDF report

Architecture Overview

---

📁 Sample Data

The repository includes a synthetic Excel file to help you get started instantly:

• Multiple sheets
• Numeric & categorical columns
• Missing values and outliers
• Time-based data

⚠️ No real client or sensitive data is included.

---

🛠️ Tech Stack

• Python
• Streamlit – Interactive UI
• LangGraph – Agent orchestration
• Pandas / NumPy – Data processing
• Matplotlib – Visualizations
• LLMs (via Ollama) – Reasoning & Q&A
• ReportLab – PDF report generation

---

▶️ Installation & Setup

1️⃣ Clone the repository

git clone https://github.com/AmritaPanjwani/contextlens-agentic-analyst.git
cd contextlens-agentic-analyst

2️⃣ Create virtual environment

python -m venv venv
source venv/bin/activate   # macOS/Linux
venv\Scripts\activate      # Windows

3️⃣ Install dependencies

pip install -r requirements.txt

4️⃣ Install and run Ollama

ollama pull llama3

▶️ Run the App

streamlit run app.py

Open the URL shown in the terminal to access the UI.

💬 Example Questions You Can Ask

• “Tell me about the distribution of key numeric variables.”
• “Are there any outliers in this dataset?”
• “How do the main metrics behave over time?”
• “Which variables are most correlated?”
• “Summarize the most important patterns you see.”

---

📄 Report Generation

The app can generate a PDF report containing:

• Dataset summary
• Key charts
• Short explanations below each graph
• Contextual notes

This report is intended as a baseline analytical artifact, not a final decision document.

---

⚠️ Limitations & Disclaimer

• Performs exploratory analysis, not predictive modeling
• Insights should be validated by domain experts
• Not intended for automated decision-making

---

🧭 Roadmap

• Multi-agent orchestration (planner + specialist agents)
• Domain-specific extensions
• Database / SQL connectors
• Auto-generated executive summaries
• Report template customization

---

📜 License

This project is licensed under the MIT License.
Feel free to use, modify, and extend it.

---

🙌 Acknowledgements

Inspired by modern agentic AI design patterns and the open-source ecosystems around LangGraph, Streamlit, and Python data tools.