Underdoc Tutorial Expense Analytics MCP Sqlite

🚀 Expense Analysis Using LLM and Receipts/Invoices Images (Part 2)

This project focuses on leveraging UnderDoc and Claude LLM to extract text from receipts and invoices images, store the data in a database, and conduct comprehensive expense analysis.

🚀 Quick Start

In Part 1, we explored how to use UnderDoc to extract text information from receipts and invoices and store this data in an SQLite database for further analysis. The next step is to build on this foundation and use Claude LLM to connect to appropriate tools through the Multi-Model Contextual Search (LMF) protocol for a comprehensive analysis of expense data.

📦 Installation

First, make sure you have installed the following software:

1. UnderDoc: Used to process and extract text information from receipts/invoices images.
2. Claude LLM: Connects to the required tools and services via the LMF protocol.

💻 Usage Examples

Using UnderDoc for Expense Data Storage

Step 1: Create an SQLite Database

Create a new SQLite database, such as expense.db, and set up a table to store expense data. Here is the table structure:

CREATE TABLE myexpense (
    id INTEGER PRIMARY KEY AUTOINCREMENT,
    name TEXT NOT NULL,
    category TEXT,
    amount REAL NOT NULL,
    date TEXT NOT NULL,
    currency TEXT NOT NULL,
    description TEXT
);

Step 2: Insert Expense Data

Insert the information extracted from receipts and invoices into the database. For example:

INSERT INTO myexpense (name, category, amount, date, currency, description) VALUES ('Grocery Store', 'Food', 150.0, '2023-10-01', 'CNY', 'Purchased food');

Using Claude for Expense Analysis

Step 1: Display Tables in the Database

View the tables in the database through the following query:

Show me the tables in the sqlite db

Claude will use the list_tables tool of the SQLite MCP server to list all tables.

Step 2: Describe the Table Structure

Select a table to describe, such as myexpense:

Describe the schema of myexpense table

Claude will use the appropriate tool to display the table structure.

Step 3: Summarize Expenses by Category and Currency

Analyze the total expenses for different categories and currencies:

Show me the total amount of expense by category, by currency

Claude will use the read_query tool to query the data and display the results in a table format.

Step 4: Generate Charts

Request to convert the results into a multi-line chart:

Show me the result in a multi-line chart

Claude will generate and run the code to display an interactive chart. You can adjust the chart's axes, colors, etc.

Step 5: Currency Conversion

Convert all expense amounts to the target currency (e.g., HKD):

Convert all expense amount to HKD currency

Claude will recalculate and display the results using the latest exchange rates.

📚 Documentation

By using UnderDoc and Claude LLM, we can easily extract data from receipts/invoices images and conduct complex expense analysis. This approach not only saves time but also provides profound insights to help make more informed financial decisions.

If you have any questions or feedback, please feel free to contact me at clarence@underdoc.io. I'm happy to assist you!