Jun 22, 2026 β€’ ai-productivity

VapePurchaseSystem Review: AI-Powered Procurement for Small Manufacturers

How a small vape manufacturer built an AI-integrated procurement system that automates BOM management, supplier risk scoring, and purchase order generation.

VapePurchaseSystem Review: AI-Powered Procurement for Small Manufacturers

Most small manufacturers still manage procurement with spreadsheets, WeChat messages, and gut instinct. VapePurchaseSystem is an open-source procurement platform built specifically for small-to-medium vape manufacturers that want to bring structure, intelligence, and auditability to their supply chain without the complexity and cost of traditional ERP systems.

What It Does

VapePurchaseSystem is a Flask-based web application that manages the full procurement lifecycle:

  • BOM Management β€” Bill of Materials with multi-level product structures, material specifications, and cost tracking
  • Purchase Orders β€” Automated PO generation based on production plans and inventory gaps
  • Supplier Management β€” Supplier profiles, risk scoring, and performance tracking
  • Production Planning β€” Production order creation linked to sales forecasts
  • Shortage Detection β€” Real-time inventory gap analysis with automatic reorder suggestions

The system runs on SQLite, making it deployable on a single machine with zero database administration. The entire stack β€” Flask, SQLite, vanilla HTML templates β€” keeps operational complexity minimal.

The AI Layer: LVE Decision Framework

What makes VapePurchaseSystem different from a typical CRUD procurement app is its integration with LVE (Lydia’s Vape Emporium Decision Framework), an AI-assisted decision engine that adds three intelligence layers:

1. Risk Scoring

Every supplier and purchase decision gets an automated risk score based on historical failure cases, red-line rules, and decision genes. The system doesn’t just store data β€” it evaluates decisions before they’re made.

2. Red-Line Enforcement

The system enforces hard rules (red lines) that cannot be overridden without explicit justification. For example: never order from a supplier with unresolved quality complaints, never approve a PO above a threshold without dual approval. These rules are codified from real failure cases, not theoretical best practices.

3. Decision Gene Pool

LVE maintains a β€œgene pool” of successful procurement patterns β€” decision templates that have proven effective across multiple order cycles. When a new purchase decision matches a known pattern, the system recommends the proven approach.

Architecture

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚           Flask Web UI                  β”‚
β”‚  (15 templates: PO, SO, BOM, Suppliers) β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
               β”‚
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β–Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚         SQLite Database                 β”‚
β”‚  (materials, suppliers, orders, BOM)    β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
               β”‚
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β–Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚         LVE Bridge Module               β”‚
β”‚  (risk scoring, red lines, gene pool)   β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
               β”‚
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β–Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚         Local LLM (optional)            β”‚
β”‚  (natural language queries, summaries)  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Real-World Performance

In a production environment managing a vape manufacturing operation with:

  • 200+ raw materials across 15 product lines
  • 20+ suppliers with varying reliability
  • Weekly production cycles with tight timelines

The system reduced procurement decision time from 2-3 hours of manual spreadsheet work to 15 minutes of reviewing AI-scored recommendations. More importantly, it eliminated two categories of costly errors: ordering from blacklisted suppliers and missing reorder windows for critical materials.

Who It’s For

VapePurchaseSystem is designed for:

  • Small manufacturers (10-50 employees) running procurement on spreadsheets
  • Teams that need structure but can’t justify a full ERP implementation
  • Operations that want AI-assisted decision-making without vendor lock-in

Limitations

  • Single-user design β€” No multi-user authentication or role-based access (yet)
  • SQLite constraints β€” Not suitable for high-concurrency multi-location deployments
  • Niche focus β€” Built for vape manufacturing; adapting to other industries requires code changes
  • No mobile app β€” Web-only interface

Getting Started

git clone <repo>
cd VapePurchaseSystem
pip install -r requirements.txt
python main.py
# Open http://localhost:5000

The system includes seed data for demo purposes and a setup_bom.py script to initialize product structures.

Verdict

VapePurchaseSystem proves that AI-assisted procurement doesn’t require a six-figure ERP budget. For small manufacturers drowning in spreadsheet chaos, it offers a practical path to structured, intelligent purchasing decisions. The LVE integration is the real differentiator β€” it transforms a simple order management tool into a decision support system that learns from past failures.

Rating: 4/5 β€” Excellent for its target audience. The lack of multi-user support and the niche focus limit broader adoption, but for small vape manufacturers, this is a genuinely useful tool.