Skip to content

AI Image Calls Documentation

Menu Location: AI > AI Image Calls

Access Level: Kiva Admin only

Last Updated: 2026-03-02

Overview

The AI Image Calls page (ai-image-summary.php) is a personal productivity utility that processes phone screenshots using AI vision. It scans a designated Dropbox folder for screenshots, sends them to the OpenAI vision API for content analysis, generates searchable text summaries, and organizes processed images into separate folders.

Note: This is a Kiva Admin personal tool, not a standard business feature.

Primary Functions:

  • Automatically identify phone screenshots by dimensions (1080x2316 pixels)
  • Send images to OpenAI vision API for content analysis
  • Generate structured JSON output (suggested filename + markdown description)
  • Save searchable text summaries as .md files
  • Organize processed images into dedicated folders
  • Process up to 5 images per run

How the System Works

Processing Pipeline

  1. Scan: Reads JPG files from the Dropbox Camera Uploads folder
  2. Filter: Only processes images matching phone screenshot dimensions (1080x2316)
  3. Skip: Bypasses already-processed images (checks for existing summary files)
  4. Analyze: Sends each image to OpenAI vision API with a detailed prompt
  5. Parse: Extracts JSON response with filename and body fields
  6. Save: Creates .md summary file in the Summaries folder
  7. Move: Relocates processed image to the Images folder

AI Analysis Prompt

The AI is prompted to:

  • Identify what the screenshot is about
  • Explain why it might have been captured
  • Extract key details (prices, dates, names, etc.)
  • Suggest a human-friendly filename
  • Provide a searchable markdown body with keywords

Output Format

Each processed image generates a markdown file containing:

  • Original image filename reference
  • AI-generated description and analysis
  • Embedded image link for reference
  • Suggested descriptive filename

Folder Structure

Folder Purpose
~/Dropbox/Camera Uploads/ Source folder - phone screenshots
~/Dropbox/Screenshots/Summaries/ Output - generated .md summary files
~/Dropbox/Screenshots/Images/ Output - processed images moved here

How to Use

Running the Processor

  1. Navigate to AI > AI Image Calls
  2. The page automatically processes up to 5 unprocessed screenshots
  3. Progress is displayed showing each image being processed
  4. Successfully processed images are moved and summaries are created

Finding Summaries

  • Summaries are saved as .md files in the Summaries folder
  • Filenames follow the pattern: [date prefix] [AI-suggested name].md
  • Search within summary files to find specific screenshots

Common Use Cases

Screenshot Organization

Automatically categorize and describe phone screenshots that would otherwise accumulate without context.

Information Retrieval

Search through .md summary files to find screenshots of events, bookings, product ideas, or other saved information.

Digital Memory

Create searchable records of screenshots taken as reminders - the AI extracts details like prices, dates, and names that make future searching possible.

Tips & Best Practices

  1. Only phone screenshots (1080x2316) are processed - other images are skipped
  2. The system processes 5 images per run to manage API costs
  3. Already-processed images are automatically skipped
  4. Check the Summaries folder periodically for your organized notes
  5. The system creates folders automatically if they don't exist

Troubleshooting

Images Not Being Processed

  • Verify images are in JPG format in the Camera Uploads folder
  • Check image dimensions are exactly 1080x2316 pixels
  • Ensure the image hasn't already been processed (check Summaries folder)

API Errors

  • OpenAI API errors are displayed on-screen and logged
  • The system stops processing on errors to prevent data issues
  • Wait and retry if the API is temporarily unavailable

Permission Errors

  • Verify write permissions on the Summaries and Images folders
  • The system checks for writable destination before moving files

FAQ

Why only 1080x2316 images?

This specific resolution matches phone screenshots, filtering out regular photos and other images in the Camera Uploads folder.

Is this available to all admin users?

No, this is a Kiva Admin personal utility tool. It only runs on localhost.

How much does it cost per image?

Each image requires one OpenAI vision API call. Costs depend on the current API pricing for vision models.

  • API Calls - Log of all OpenAI API calls including image analysis calls
  • AI Overview - Central AI dashboard