Lab_Receipt_Bot

A quick tool to help Jon Simon group people to process invoices and receipts, powered by GPT4o-mini.

Installation

pip install -r requirements.txt

Running

streamlit run app_streamlit.py

Usage

secret.yaml:

openai_api_key: "<openai api key>"
sender_username: "xxx@gmail.com"
sender_password: "<app password>"

config.yaml:

invoice_analysis_prompt: |
  You are given an invoice. You should parse it and do the following:
  1) Identify the final total amount charged in the invoice (the amount paid). Output a float or integer number with no currency symbols or extraneous text.
  2) Describe the main thing of the purchase, with its manufacturer or brand, do not include amounts or quantities.
  Return your result as valid JSON in the form:
  {{
    "amount": "float or int number here",
    "rationale": "short summary or fallback"
  }}

summary_prompt: |
  Please describe the main thing of the purchase in a very precise sentence, use "and" to connect multiple items.
  Here are the rationales:

contacts:
  - name: "PhD"
    email: "<phd email>"
    role: "purchaser"
  - name: "Admin"
    email: "<admin email>"
    role: "admin"

grants:
  - name: "<grant display name>"
    code: "<grant code>"