karimz1/imgcompress

Technical Requirements for Enhancement: Ability to define a few presets Technical Ticket for Feature #810, for questions tag: @karimz1 as the architect. Summary Add persistent compression presets. Users should be able to save the currently selected compression settings under a custom name and apply them again later. Presets are global to the imgcompress instance. There is currently no user separation, so all users can view, create, update and delete presets. Presets must be persisted to /data/presets.json so /data can be mounted as a Docker volume. Feature flag Add: DISABLE_CUSTOM_PRESETS=false Default: false. Follow the existing DISABLE_* configuration pattern and expose the flag through /config/runtime.json. When set to true: Do not render any preset UI. Preset API endpoints return 403. Existing preset data must remain untouched. Preset data A preset stores the compression parameters currently supported by /api/compress: format quality width target_size_kb use_rembg pdf_preset p

Hi @karimz1 . I have been looking for this kind of tool for a while but never came across imgcompress. I found it on Awsome Selfhosted. Quick, straightforward and excellent results. I have a feature suggestion. I often need to convert individual images and it is a bit tedious to set the required parameters every time I open imgcompress. Would it be possible to have an option to pre-define a few parameter sets and with one click select the preset. I don't mind if presets have to be defined in a config file on the server. Do you think that's reasonable?

I have been using imgcompress to convert large number of images trough drag-and-dropping directories into the Web UI. It's worked great for my use-cases so far but the biggest issue that I've run into with large conversion are that the directory structures are flattened post-conversion: The pre-converted directory structure exists as: . ├── test-nested1 │   ├── test-nested2 │   │   └── test3.jpg │   └── test2.jpg └── test1.jpg which then becomes: . ├── test1.avif ├── test2.avif └── test3.avif It would be extremely helpful if the original folder structure was preserved in the output zip file

Add GPU (CUDA) support for local AI background removal Context ImgCompress now ships selectable local AI background-removal models (u2net, isnet-anime, isnet-general-use, u2net_human_seg, birefnet-general-lite) and a slim vs latest image split. All of that runs on CPU only. Some of these models (especially birefnet-general-lite) are noticeably slow on CPU. On machines with an NVIDIA GPU we could run them far faster through onnxruntime's CUDA execution provider. GPU support was intentionally left out of the model-selection PR to keep that change focused and CPU-only. This ticket tracks adding it properly as its own PR. Goal Offer an optional CUDA-accelerated image that uses the GPU when one is available and falls back to CPU otherwise, without changing the default (CPU) experience for anyone. It must be reproducible, covered by CI/CD, and published as a dedicated image tag. Requirements Provider auto-detection. The app should query onnxruntime for the available execution providers and

Summary I'd like to propose adding a lightweight MCP (Model Context Protocol) wrapper server so LLMs like Claude can use imgcompress as a tool directly, without a human operating the UI. Motivation imgcompress is a capable image processing backend, but it currently only has a browser-based frontend as its client. With the rise of AI agents and LLM-driven workflows, there's a growing need for tools that can be invoked programmatically by models. An MCP server bridges that gap — it exposes the existing REST API as structured tools that any MCP-compatible LLM can call. Proposed approach Transport: stdio (the LLM client spawns the server as a local process — no new ports, no new attack surface) Architecture: pure proxy — no business logic, no changes to the backend. All calls go through the existing REST API, exactly as the browser does New dependencies: mcp[cli], httpx (Python, isolated in mcp/) Structure: mcp/ pyproject.toml ← dependencies (mcp[cli], httpx) server.