AutoScrum — Agentic AI Sprint Planner (Streamlit + LangChain)

A multi-agent Scrum assistant that reads a backlog CSV and produces:

• a 2-week sprint plan (working-days aware)
• a risk report
• a final sprint summary

using memory-enabled agents and prompt templates.

Agents collaborate. The user decides. Output stays explainable and reviewable

---

Architecture

---

Screenshots

---

What It Does

✅ 1) Backlog Understanding (CSV → Text Context)

• Upload backlog CSV
• Convert rows into compact text for AI reasoning
• Ask optional general questions about backlog

✅ 2) Agentic Actions

• Sprint Planning Agent → assigns tasks with workload balance
• Risk Detection Agent → finds blockers & overload
• Report Generator Agent → executive sprint summary

✅ 3) Memory-Aware Conversations

• Uses ConversationBufferMemory
• Enables follow-up reasoning

Example:

“Which tasks are biggest risk and why?”

---

Tech Stack

• Python
• Streamlit UI
• LangChain Core + Memory
• OpenAI GPT-4 / GPT-4o
• Pandas (CSV ingestion)
• Prompt templates

---

Quick Start

git clone https://github.com/lavanya1402/AutoScrum-AgenticAI.git
cd AutoScrum-AgenticAI

python -m venv venv
venv\Scripts\activate

pip install -r requirements.txt
streamlit run streamlit_app.py

---

Configuration

Create .env file:

OPENAI_API_KEY=your_key_here

---

Project Structure

AutoScrum/
├── agents/
├── prompts/
├── utils/
├── assets/
│   ├── architecture.svg
│   ├── screenshot1.png
│   └── screenshot6.png
├── streamlit_app.py
├── requirements.txt
├── README.md

---

This project demonstrates:

✅ Multi-agent orchestration
✅ Memory-enabled reasoning
✅ Prompt templating
✅ Gated execution workflow
✅ Real-world PM automation

This is not a chatbot.
This is an agentic workflow system.

---

License

MIT