🧠 Excell-AI — Professional Local Stack

PhD-Grade Inventory Intelligence (Ollama + Qwen + LangGraph + Streamlit)

A high-performance, fully local AI pipeline designed for professional inventory analysis and data-driven decision support.

🏗️ Architecture

User (UI)  ──►  Streamlit App  ──►  LangGraph (Logic)  ──►  Ollama (Qwen LLM)
     ▲              │                    │                       │
     └──────────────┴────────────────────┴───────────────────────┘

📁 Project Structure

Excell-AI/
├── app/
│   ├── main.py              # Premium Streamlit UI
│   ├── chains.py            # LangGraph / LangChain logic
│   ├── prompts.py           # PhD-grade system prompts
├── core/
│   ├── llm.py               # Unified Ollama interface
│   ├── config.py            # Centralized settings
│   ├── tools.py             # Data analysis tools
├── data/
│   └── docs/                # Local knowledge base (RAG)
├── scripts/
│   └── ingest.py            # RAG ingestion logic
├── docker-compose.yml       # Production orchestration
└── requirements.txt         # Streamlined dependencies

🚀 Quick Start

1. Requirements

• Docker (Recommended)
• Ollama (If running locally without Docker)

2. Run with Docker (Recommended)

docker-compose up --build

• Access the UI at http://localhost:8504.
• Ollama with Qwen 2.5 will be automatically configured inside the container.

3. Run Locally (Manual)

# Pull Qwen
ollama pull qwen2.5:3b

# Install dependencies
pip install -r requirements.txt

# Start Streamlit
streamlit run app/main.py

💡 Expert Persona: Dr. Elena Voss

The system uses a PhD-grade reasoning protocol (ASCM/CIPS standards) to ensure every inventory recommendation is:

1. Rule-based: Adheres to global supply chain standards.
2. Validated: Calculations are cross-checked.
3. SMART: Specific, Measurable, Achievable, Relevant, and Time-bound.

🛠️ Customization

• System Prompt: Edit app/prompts.py to change the AI's persona.
• Graph Logic: Edit app/chains.py to modify the LangGraph workflow.
• Tools: Add new data analysis functions in core/tools.py.

Built With

• dockerfile
• python
• shell