DMAI NODES 0.2.0 - Slate redesign

DMAI NODES 0.2.0 — Slate interface update

Larger controls, readable native connections and free Manual sampling replace the cramped 0.1.x layout.

• Roomier Prompter with clear format controls, Custom resolution and Generate.
• Two-row LoRA cards with filenames, order, enable/edit actions and independent Model/CLIP controls.
• Clear captions beside the real Canvas sockets; no duplicate connector toolbar.
• Installed checkpoint/diffusion model selector stays at the top of Engine.
• Manual is the default. JSON upload exists only in DMAI Enhanced; the old preset catalog and export controls are removed from the interface.
• Saved sampling values and older selected presets remain compatible. Legacy DMAI Prompters retain their separate IDs.
• Gallery keeps individual downloads and a visible ZIP footer while images scroll.

Updating a pinned installation

git fetch origin tag v0.2.0
git checkout v0.2.0
python -m pip install -r requirements.txt
git describe --tags --exact-match

Restart ComfyUI, then hard-refresh the browser with Ctrl+F5. Saved nodes may expand to the new minimum sizes; move them apart if needed. Larger user-resized nodes are preserved.

Validation

102 Python tests pass, including real CPU Comfy registration and legacy coexistence checks. Frontend regressions and real Canvas checks cover Manual/Enhanced import and draft restoration, independent LoRA edits and reordering, socket dragging, save/reload, and panel visibility. Exact frontend test count is recorded in docs/VALIDATION.md.

This is a preview release. GPU inference, image quality, Runpod templates and complete Nodes 2.0 rendering remain unverified. See docs/VALIDATION.md for evidence and limits. The package contains no model weights.

DMAI NODES 0.1.3 Preview

Installing DMAI NODES alongside an older DMAI Suite or standalone Prompter could hide or replace the new Prompter because both used the same internal ID. This release gives the new node its own ID, DMAINodesPrompter, while keeping its display name DMAI Prompter and the DMAI-NODES package badge.

• Both Prompters can coexist with separate backend contracts and frontend controls.
• Identifiable DMAI NODES 0.1.0-0.1.2 UI workflows migrate on load, preserving IDs, links and settings. Older Suite/standalone workflows keep their original nodes.
• API starters and the submission helper use the new ID. Direct API callers can convert saved JSON with tools/migrate_workflow.py old.json --output migrated.json.
• New LoRA row styles are isolated from the older DMAI LoRA Loader styles.

Update

Run inside your existing custom_nodes/DMAI-NODES checkout:

git fetch origin tag v0.1.3
git checkout v0.1.3

Restart ComfyUI and hard-refresh the browser. Select DMAI Prompter with the DMAI-NODES badge. Keep older DMAI packages enabled when their workflows are needed.

Validation

99 Python tests passed with optional real-Comfy CPU tests enabled, plus 64 JavaScript tests. Real Comfy loader tests cover legacy V1/V3 Prompters in both registration orders. A live CPU ComfyUI instance loaded the actual older Suite and all seven new nodes together; browser search, distinct controls, migration and save/reload were checked.

GPU generation, full RES4LYF execution and Runpod GPU validation remain pending. This release fixes node coexistence and does not add new model-quality claims. MIT preview; model weights and optional extensions are separate.

DMAI NODES 0.1.2 Preview

Native connections can now be dragged without the node panels intercepting the mouse. The Engine's top Model selector lists actual installed checkpoints and diffusion model files.

• Native sockets have clear hit areas along the node sides.
• Hidden JSON widgets no longer consume socket spacing; resize keeps socket positions aligned.
• Model files, architecture, text encoders and VAE are grouped at the top of the Engine.
• Exact paths, unavailable saved files and architecture-specific settings are preserved.
• Encoder/VAE changes refresh missing-file messages immediately.

Validation: 53 JavaScript tests, native ComfyUI browser connection and model-picker checks, and the existing CPU contract coverage. Model-file browser tests use explicitly named tiny QA fixtures, without image inference. See docs/VALIDATION.md for details and model support limits.

Update the package, restart ComfyUI, then refresh the browser. Existing workflow connections are preserved. MIT license; preview release.

DMAI NODES 0.1.1 Preview

Generate now lives in the Prompter. A softly glowing border follows measured progress across all requested images; 100% appears only after the entire workflow succeeds, including Gallery output.

• Generation Engine keeps its model and sampling controls in the background.
• Queued, loading, sampling, decoding, saving, error and stopped states stay distinct.
• Cached jobs, native Run, subgraphs, multiple Engine branches and removed queue entries are handled.
• The WebSocket progress event is documented for future website clients.

Validation: 81 Python tests (including real Comfy CPU checks), 38 JavaScript tests, native WebSocket error-path verification and browser checks. Full model inference on GPU remains pending. See docs/VALIDATION.md for the complete coverage.

Install the ZIP in ComfyUI/custom_nodes, restart ComfyUI, and refresh the browser. This is a preview release under the MIT license.

DMAI NODES 0.1.0 Preview

DMAI NODES 0.1.0 is the first public preview of the Slate node suite.

Includes seven native V3 nodes: Prompter, LoRA Loader, Generation Engine, Gallery, Krea Enhancer, VAE Encode and VAE Decode. Model profiles cover Krea 2 Turbo, SDXL, Qwen Image and FLUX.1 Dev. Sampling uses the installed Comfy registry and optional external SAMPLER/SIGMAS inputs, with RES4LYF wiring documented.

Download the ZIP, extract it into ComfyUI/custom_nodes, install requirements with ComfyUI's Python, then restart. Local, Windows Portable, Runpod and API instructions are included.

Validation: 70 Python tests and 19 JavaScript tests passed; real ComfyUI v0.37.0 CPU registration, LoRA comparisons and Gallery API/ZIP checks passed. GPU model inference, full RES4LYF execution and Runpod installation remain unverified. This preview contains no model weights.

MIT license. Verify the ZIP using the attached SHA-256 file.