InvoicePilot AI: Smarter Invoice Automation with TEv1 and MCP

Description and use

A local AI assistant that processes invoice images, extracts client, amount, and due dates, stores them in SQLite, answers natural language queries, and drafts polite reminder emails—all without sending data to the cloud.

Problem

Freelancers spend hours chasing late payments and managing invoices manually.

Outcome

The assistant reduces manual invoice processing time, keeps client data local, and provides quick, accurate responses and reminder drafts.

Workflow

• Upload invoice image or screenshot
• Vision model (gemini-flash-latest) reads image and outputs JSON with client, amount, due dates
• User confirms extracted fields
• Data saved to local SQLite database
• User asks natural language query (e.g., "which invoices are overdue?")
• AI agent (TEv1 + local LLM via Ollama) selects appropriate tool via MCP
• Tool executes query on database
• AI drafts reminder email with exact amount and days overdue

Evidence

You upload a photo or screenshot of an invoice, a vision model running on your own laptop reads the client, amount, and due dates, and after you confirm the fields it saves them to a local database.

Source

The project explores whether combining a lightweight decision model with an LLM can improve tool-calling reliability while reducing input-token usage.

Source

By running the decision model locally, the project avoids additional API costs for tool selection while keeping the main LLM responsible for generating tool arguments.

Source