localhost.
+ Open a folder of images to begin.
+Every image is shown with a badge telling you whether it has a prompt.
+The full prompt as plain text.
+ +Open a folder of prompt .txt files to begin.
Select a prompt on the left, or create a new one.
+Each file is one reusable prompt; the filename is its title.
++ Pick a label type, a relationship marker and a reference image from disk. + The model writes the definition line, and a matching retention_analysis + entry is added for you to complete. Indices auto-increment per type. +
+No references yet — add one above.
+ +Added automatically per reference (no AI). Fill in the shots and replace each TODO.
+ Connect a local OpenAI-compatible LLM (LM Studio, Ollama, …) to generate, + complete, and improve prompts across the app. In LM Studio: start the local + server and enable CORS. Stored locally as settings only — never your content. +
+ + + +
+ Sends cache_prompt=false so the server doesn't reuse the previous
+ request's context — prevents a prior image "leaking" into later text
+ generations (LM Studio). Turn off for slightly faster repeated prompts.
+