Type something to search...
to navigateto selectESC to close

Warung Assistant

Summary

Warung Assistant is an AI-powered assistant built with Large Language Models (LLMs), LangChain, and Retrieval-Augmented Generation (RAG) to help small retail store owners (warung owners) handle customer inquiries without losing their heads.

It answers the questions customers ask all day — product prices, stock availability, shelf locations, estimated total costs. That leaves store owners free to focus on finalized purchases and payment processing, with fewer errors and faster service.

Problem

Small retail stores hit the same operational friction again and again:

  1. Repeated customer questions
    For example:

    • “How much is Indomie Ayam Bawang?”
    • “Is there any egg left?”
    • “Which shelf is the soy sauce on?”

  2. Distractions during checkout

    • Store owners are often interrupted while calculating purchases
    • Every interruption raises the risk of miscalculation

  3. Unstructured customer input
    Customers speak naturally, like:

    • “Half a kilo of eggs, 3 Indomie, and 2 bundles of spinach”

    Product names, units, and quantities vary widely.

  4. Rigid traditional POS systems

    • Cannot understand natural language
    • Cannot assist customers before the order is finalized

Solution

Warung Assistant acts as an AI front-desk for customers, while the store owner focuses only on payment and fulfillment.

Solution Workflow

  1. Customer interacts with Warung Assistant

    • Ask for prices
    • Check stock availability
    • Ask product locations
    • Request estimated totals

  2. Natural Language Understanding

    • Extracts product names, quantities, and units
    • Supports:
      • Kilograms, halves, quarters
      • Grams / ounces
      • Unit-based quantities
      • Budget-based requests (e.g., “10,000 IDR worth of sugar”)
      • Product name synonyms

  3. Reasoning & Validation

    • Fuzzy product matching using RapidFuzz
    • Stock validation (requested quantity ≤ available stock)
    • Subtotal and grand total calculation

  4. Structured Output

    • Product list
    • Unit price
    • Quantity
    • Subtotal per item
    • Grand total
    • Shelf location (if requested)

  5. Owner-focused Checkout

    • Fewer things to hold in your head
    • Faster checkout
    • Fewer calculation errors

Impact

What this actually changes for a small retail business:

  • Faster customer service
  • Fewer calculation errors
  • Lower cognitive load for store owners
  • Better customer experience
  • A foundation for POS and inventory systems

As a portfolio piece, it also covers:

  • Practical LLM system design
  • Tool-based reasoning
  • Handling real-world constraints
  • End-to-end AI application development

Tech Stack

LLM & AI

  • Ollama (Local LLM: Gemma / LLaMA / Mistral)
  • LangChain (Agents, Tools, Memory)
  • Prompt Engineering (ReAct, Few-shot prompting)

Retrieval & Data

  • CSV-based Inventory (RAG-style retrieval)
  • RapidFuzz for fuzzy product matching
  • Custom quantity and unit parsing logic

Backend

  • Python
  • Pandas
  • Modular tool-based architecture:
    • check_stock
    • get_price_from_csv
    • calculate_total_price
    • find_product_location

Interface

  • Gradio ChatInterface
  • Interactive real-time chat UI

Extras (Optional / Future Work)

Things I’d like to build on top of it:

  • Role-based access (Customer mode / Owner mode)
  • Transaction export (receipt, CSV, PDF)
  • Automatic stock updates after payment
  • Vector database integration (FAISS / Chroma)
  • Voice input (Speech-to-Text)
  • Multi-turn memory (e.g., “What did I buy earlier?”)

Why This Project Matters

This one is where the pieces come together:

  • Applied LLM reasoning
  • LangChain agent orchestration
  • Tool calling and state validation
  • Practical AI for real-world small businesses

It’s a chatbot only on the surface. Underneath it is a production-oriented AI assistant system, and building it meant dealing with everything a real store throws at it.