Uranium: AI-powered Q&A for Spreadsheets

Welcome to Uranium! This project helps you automate Q&A for your spreadsheets by leveraging AI. It fetches questions from a specified Google Sheet, processes answers using the Gemini AI engine, and populates the answers in another column.

Benefits:

• Save time on repetitive tasks like RFP responses and research analysis.
• Leverage AI for insightful answers to your questions.
• Simplify workflow by automating data extraction and response generation.

Getting Started:

1. Prerequisites:

• A Google Cloud Platform (GCP) project with a service account.
• A Gemini AI application and API key.
• Java installed on your system.

2. Configuration:

• Edit application.yml located in the /release/config directory.

• Configure the following sections:

  a. Google Sheet Integration:

  sources:
    google:
      enabled: "true"
      name: "<ApplicationName>"  # Replace with your application name
      credential: "<CredentialFileName>"  # Replace with service account credential file name

  • Create a service account in your GCP project and download the credential file.
  • Place the downloaded credential file in the /release/config directory.

  b. Gemini AI Integration:

  processors:
    gemini:
      enabled: "true"
      url: "https://generativelanguage.googleapis.com/v1beta/models/gemini-pro:generateContent"
      key: "<KEY>"  # Replace with your Gemini API key
      delay: 5000  # Delay in milliseconds between requests (optional)

  • Create a Gemini application and get your API key.

  c. Start Application:

  • Navigate to 'release' directory and execute the following

   sh ./start.sh
  

3. Sending a Request:

Use a tool like Postman or curl to send a JSON request to http://<host>:8080/qna/sheet url. Replace <host> with your server's address. The request body should look like this:

{
  "sheetName": "<STRING, NAME_OF_SHEET>",
  "questionColumn": "<CHAR, COLUMN LABEL OF QUESTION>",
  "answerColumn": "<CHAR, COLUMN LABEL OF ANSWER>",
  "startColumnIndex": <INT, START INDEX OF COLUMN TO READ>,
  "endColumnIndex": <INT, END INDEX OF COLUMN TO READ>,
  "suffix": "<STRING, LITERAL TO SUFFIX QUESTION (optional)>",
  "url": "<STRING, URL OF SHEET>"
}

Example Request:

{
  "sheetName": "Monitor & Manage",
  "questionColumn": "C",
  "answerColumn": "D",
  "startColumnIndex": 24,
  "endColumnIndex": 67,
  "suffix": ",explain from WSO2 point of view",
  "url": "https://docs.google.com/spreadsheets/d/1eRJqk2P6b1hqu1hoQFKfQ6KCi5uyR5KHAbHdTMX8EU0/edit#gid=1420065883"
}

4. Response:

The Spring Boot service console will display messages indicating progress:

Gathering questions from spreadsheet..
Questions collected, now processing answers
Processing completed, please check the spreadsheet

5. Check the Spreadsheet:

Open your Google Sheet and check the designated answer column for the AI-generated responses.

Also check this video https://drive.google.com/file/d/1HnCUxGtkQMBaTehveooTpJ9Pd2HrjDn5/view?usp=drive_link

Important Notes:

• If the processing engine encounters issues, "ERROR" will be written in the answer cell.
• Currently, only Google Sheets are supported.
• Ensure the Google Sheet has appropriate access permissions for the service account.

Further Development:

• Support for additional spreadsheet formats like XLS.
• Error handling and logging improvements.
• User interface for configuration and request submission.

We hope Uranium empowers you to streamline your data analysis and research tasks!

Feel free to contribute to this project and suggest improvements.