image-viewerlisted

# Image Analysis & Organization in EverFern ## 🚨 CRITICAL: WHEN TO USE VISION VS WHEN TO SKIP IT ### Use file extension (NO vision needed) — format/organize-by-type queries If the user asks to organize or filter **by file format/type**, file extension is sufficient. Do NOT waste time/cost on vision. | User says | Do this | |-----------|---------| | "organize all SVG files" | Filter by `.svg` extension → move to SVG folder | | "find all JPEGs" | Filter by `.jpg`/`.jpeg` extension | | "separate PNGs from JPGs" | Filter by extension, zero vision calls | | "find huge images" | Use Pillow for dimensions, no vision | ### Use vision (MUST use analyze_image) — content-based queries If the user asks to classify **what's IN the image**, you MUST use vision. **NEVER guess content from file names, file size, or metadata.** **Wrong** (filename guessing — banned): > File is called "anime.jpg" → classify as anime ✗ **Correct** (must use this): > Call `analyze_image` with question "Is this image anime/manga, photograph, > illustration, or screenshot?" → vision model actually sees the pixels ✓ | User says | Must use vision? | |-----------|-----------------| | "organize by file type" | NO — use extension | | "organize by content" (anime vs photos, etc.) | YES — analyze_image | | "find screenshots" | YES — analyze_image | | "find photos of people" | YES — analyze_image | | "separate memes from real photos" | YES — analyze_image | | "what images are in this folder?" | NO — just list file