FlowMind AI --- Autonomous Workflow Agent

FlowMind AI is a multi-agent autonomous system that converts unstructured business inputs into executable workflows, validates them, and self-recovers from failures --- all while maintaining a complete audit trail.

It is designed as a scalable architecture for enterprise workflow automation using Agentic AI.

---

Problem Statement

This project addresses:

"Agentic AI for Autonomous Enterprise Workflows"
from the ET Gen AI Hackathon 2026.

The goal is to build a system that: - Processes complex multi-step workflows\

• Works with unstructured input (text / transcripts)\
• Detects failures automatically\
• Recovers without human intervention\
• Maintains complete traceability

---

Solution Overview

FlowMind AI uses a multi-agent architecture orchestrated via LangGraph.

Pipeline:

User Input / File Upload
→ Preprocessing (cleaning + chunking)
→ Extractor Agent
→ Planner Agent
→ Executor Agent
→ Validator Agent
→ Recovery Agent (if failure)
→ SLA Agent (auto-fix)
→ Final Validation
→ Output + Logs

---

Core Architecture

Input Layer

• Text input
• File upload (.txt, .md)

Preprocessing

• Cleaning
• Chunking for long transcripts

Agents

• Extractor: Converts text → tasks\
• Planner: Tasks → steps\
• Executor: Simulates execution\
• Validator: Rule-based validation\
• Recovery: Fix suggestions\
• SLA Agent: Auto-deadline assignment\
• Final Validator: Ensures correctness

---

LLM Usage

Used only for: - Task extraction - Planning

API: Groq API
Models: LLaMA / Mixtral variants

---

Logging

• Stored in data/logs.json
• Contains full agent trace

---

Output

• Tasks\
• Steps\
• Validation status\
• Issues\
• Fixes

---

Tech Stack

• Python\
• LangGraph\
• Groq API\
• Streamlit

---

Setup

git clone https://github.com/aryasm13/flowmind-ai.git
cd flowmind-ai
pip install -r requirements.txt

Create .env:

GROQ_API_KEY=your_api_key

Run:

python -m streamlit run app.py

---

Author

Arya Mulay