Repository navigation
Image Organizer AI Tagging and Search
New in 1.5.0. Optional, and off until you turn it on.
A vision AI model can describe your photos in more detail than the face models, and check them for anything you name, such as red sweater. RulZ talks to an AI server you choose. While AI is off, RulZ never sends a photo anywhere.
Open Library ▸ AI tagging and search and pick one:
| Option | What it uses |
|---|---|
| Off | Nothing (the default). |
| My own server | Any OpenAI-compatible server: Ollama (http://127.0.0.1:11434/v1, the default), LM Studio (http://127.0.0.1:1234/v1), vLLM or a running llama-server. Type the model name or press Load models. It must be a vision model; for Ollama, for example, qwen2.5vl:7b or gemma3 (install with ollama pull qwen2.5vl:7b). Several addresses separated by commas share the work. |
| Training Captioner's server | For power users who already set up the [[Rulz Training Captioner |
Test connection sends a small red test picture and checks that the model names its colour, so you know it can see images.
If the server is on another computer, the card says so in red, and RulZ asks before sending the first photo to each such server, every time RulZ starts.
Use the AI button on the Photos page, the selection bar, a photo's right-click menu, or Tag with AI in the viewer. Pick this photo, the selected photos, the photos shown, or all photos.
The model lists what it sees: people and their clothing, objects, places, weather, colours and the kind of picture. The tags show in the Photo Viewer (click one to search for it), and the search box finds photos by any word of a tag: sweater finds red sweater, and dogs finds dog.
From the same places, choose Look for... and type what to look for, for example red sweater. To look for several things, separate them with commas. The model answers YES or NO for each photo. When you start from the Photos page, the matches show up there as they are found.
- Tags and YES/NO answers are stored in your library, so nothing is asked twice. Searching for the same text later shows the results straight away, and the dialog's button reads Show results (nothing is sent).
- Photos already tagged or checked are skipped unless you tick Tag again / Check again.
- An edited photo counts as not tagged or checked yet. A deleted photo takes its results with it.
- Forget the selected search (in the Look for dialog) and Clear AI results... (Library) remove results.
- Unclear answers (neither YES nor NO, "not sure", a reply cut off) and failed requests are not kept, so the next run asks again.
The normal search box finds AI results too (see Search and Filters):
- Someone's full name, and groups made only of expressions, directions or poses (
smiling), still mean what they did. - Other words are looked up in your Look for searches and AI tags, as written (
woman in red dress), before matching part of a name:carfinds the tag "car", not Carl. -
Mom, red sweaterfinds photos with Mom where the AI found a red sweater. - Text in quotes, like
“smiling”, means only the AI's results for exactly that text. - For a Look for search, a NO answer wins over a tag with the same words.
Advanced... holds the instructions (system prompts) and questions RulZ sends to the model, the token limits, temperature, image size (1024 px), requests at a time, the reply timeout, thinking on/off and the port for RulZ's own llama-server.
Warning: RulZ reads the model's replies by their format. Tagging must reply with a comma-separated list, and Look for with a reply that starts with YES or NO. Changing the prompts can make every reply unreadable.
The page is locked until you tick I understand, and saving lists what changed and asks again. {query} in the Look for question is replaced by what you look for. Restore defaults puts the built-in prompts back. Results saved before a change stay as they are.
One AI job runs at a time. Its progress shows in the sidebar with Stop; requests already sent finish first, and everything found so far is kept. It can run while a library scan runs.
If a server doesn't accept the connection within 10 seconds, RulZ counts it as down: with several addresses the others take over its photos, and when none answers the job stops. Model replies and llama-server messages go to ai.log in the library data folder.
RulZ AI Tools Suite
The apps
RulZ AI Trainer
RulZ AI Trainer Optimizer
Using the app
Help
Rulz Training Captioner
Getting started
Using the app
Models and GPUs
Reference
RulZ LLM Suite
Using the app
- Getting Started
- Building the Engine
- Model Library
- Loading Models
- Chat
- MCP Tools
- Reverse Engineering
- Local API Server
Reference
Contributing
RulZ Coding Agent
Using the app
- Getting Started
- Projects and Tasks
- Approvals and Safety
- Model Settings
- Run and Share
- Notes and Instructions
- Extra Tools (MCP)
- Phone Access
- Android App
Reference
Contributing
RulZ Image Organizer
Using RulZ
- Getting Started
- Library and Scanning
- People and Corrections
- Search and Filters
- AI Tagging and Search
- Photo Viewer
- Appearance
More
RulZ Image Tournament
Using the app
- Getting Started
- How the Tournament Works
- Comparing Images
- Controls and Shortcuts
- Options and Settings
- Supported Formats
- Troubleshooting and FAQ
Developers
RulZ Image Crawler
Getting started
Using the app
Help
RulZ Image Preparer
Getting started
Using the app
Reference
RulZ Image Grid
Using the app
Help
RulZ Huggingface Manager
Using the app
Help
RulZ Simple Video Clip Extractor
Getting started
User guide
- Playback and Marks
- Fixed-Length Clips
- Exporting
- Clip List
- Scene Cuts
- Sound Waveform
- Crop, Rotate and Flip
- Colour Matching
- Distortion Repair
- Preferences and Themes
Help