Show HN: A local-first financial auditor using IBM Granite, MCP, and SQLite (github.com)

A new local-first financial auditing platform has been introduced, utilizing IBM's Granite, Model Context Protocol (MCP), and SQLite to transform raw bank data into accurate financial insights. This innovative project employs a Local LLM architecture that simulates a Senior Auditor, intelligently navigating between database tools to deliver mathematically precise summaries while filtering out non-essential internal transactions like transfers and credit card settlements. Notably, the platform runs entirely on a user's machine, ensuring data privacy by keeping financial information within the local environment. The significance of this development lies in its combination of privacy, accuracy, and user-friendliness in financial analysis. Built with a modular architecture, the platform features a React-powered dashboard alongside a Python RESTful API that facilitates tasks such as PDF bank statement uploads, transaction categorization, and manual inputs. Unique tools include the use of the granite model for normalizing vendor names and performing agentic reasoning, as well as leveraging SQLite for arithmetic processes to guarantee the accuracy of financial calculations. Future enhancements are planned to introduce automated categorization based on user behavior, further optimizing personal financial management.