Image-to-Memory
Upload images, get AI analysis, save descriptions as searchable memories.
Upload images, get AI analysis, save descriptions as searchable memories.
What It Does
- Accepts image uploads (JPEG, PNG, GIF, WebP, BMP, TIFF up to 20MB)
- Sends images to configured vision AI (OpenAI GPT-4o, Gemini, Ollama LLava, OpenRouter)
- 8 context modes: general, screenshot, whiteboard, document, diagram, photo, chart, meme
- Extracts structured descriptions with tags (TAGS: ["tag1", "tag2"] format)
- Stores analyses in
image_analysestable with thumbnail - Save any analysis as a full embedded memory
- Supports re-analysis with different context or custom prompt
- Gallery view with stats
How To Use It
- Configure a vision-capable AI provider in Settings
- Upload an image in the Image-to-Memory page
- Select context type (screenshot, whiteboard, etc.)
- View the AI analysis and tags
- Save as a memory for future retrieval
Port Architecture
- Prompts, tag extraction, request builders in
src/server/plugins/ports/image-to-memory.ts - Vision API calls + route handling at
src/app/api/v1/plugins/image-to-memory/route.ts - Test at
tests/unit/image-to-memory.test.ts