LU v2.6.7

2.6.7 is a repair release. Every fix went back to a fresh tester who did not know what had been changed, driving the real installed build on Windows, and whatever they broke went into the next round. That loop ran seventeen times. Most of what follows is the app no longer claiming things that were not true.

What is new

• A ComfyUI that dies while the app is running restarts itself. Three attempts with a growing pause between them, and the render that triggered it carries on afterwards. A ComfyUI on a foreign host, a missing installation and an environment broken at import each get their own sentence instead, because none of those is ours to restart.
• An idle app notices that ComfyUI is gone. It used to sit silent for as long as you left it and heal only when you asked for the next picture. A quiet line now appears within about half a minute and says what will happen: a ComfyUI the app started restarts with your next render, a remote one does not, and the line says so without promising a rescue.
• A cold ComfyUI start explains itself. Once the loading phase passes twelve seconds a line tells you the model is going into memory. Warm runs and small checkpoints never see it.
• A test button under Settings, AI Backends checks the built-in engine end to end, repairs what it can and then reports what it found.

Fixed

Create and rendering

• The very first render after starting the app shows its loading texts again. The websocket was connected after the job had already been submitted, and that first connect costs up to five seconds, so every phase message in that window was lost. On the test machine that meant 74 seconds of model loading with nothing on screen but Queued.
• Sampling is only claimed once ComfyUI is really sampling. The sampler's executing event arrives before the work starts, and the app read it as step one, so the progress line ran forty seconds ahead of reality. It waits for a reported step now, and that early event says the model is loading.
• A still image gets an honest decode line. It used to announce that frames were being decoded and call that the last long stretch. A picture has no frames, and on measurement it is the shortest phase of the render, around three seconds.
• The loading line stopped promising a warm cache. It said later renders reuse the loaded model. Measured on a large model, a warm run took 22 seconds against 23 cold, because ComfyUI unloads after every render. The promise is gone rather than dressed up.
• A dismissed cross origin warning stays dismissed. The yellow bar came back after every single image, even right after you closed it, because nothing remembered the click. It is now tied to the host and ComfyUI version that caused it and survives a restart of the app.
• The render watchdog stopped throwing away work that was still running. It asks ComfyUI whether the job is still queued before calling five quiet minutes a hang, allows more time while a first checkpoint is loading, and gives a render in its final steps another minute.
• A render on the processor leaves the chat engine where it is. ComfyUI started with the CPU flag touches no video memory, so unloading and reloading the chat model around every picture cost a wait for nothing.
• The CPU notice names the reason it is actually there. It has one text for Force CPU, which also says how to switch back, one for no usable card, and one for an AMD card without ROCm. AMD help stopped showing up on NVIDIA machines.

Models and backends

• A downloaded model shows up as installed (#113). A bundle counted as installed as soon as a neighbouring bundle had brought one shared file along, so a card said Installed while its own main model was missing. Every file is checked now, and after a download the app keeps looking in the background for a full minute instead of giving up after four seconds.
• The Installed inventory names every ComfyUI folder, not only checkpoints and diffusion models. LoRAs, VAEs, text encoders, CLIP vision, ControlNet, upscalers, embeddings and style models were invisible.
• Your own file stops disappearing behind a catalogue price. A model whose catalogue entry lists 16 GB was thrown away as a partial download when the file on your disk was smaller, even when it was a perfectly good different build. Pickers still filter, the inventory never does.
• The counters stopped reporting a zero they had not counted. A cold start could leave Installed at zero until you refreshed by hand.
• A deleted model disappears from the list immediately instead of standing there for another ten seconds.
• The built-in engine starts on a fresh installation (#118). Text downloads picked their destination from the model you were chatting with, and a fresh install has none, so the file landed where the engine never looks. The download goes to the right place, models that already went missing are found again, and an engine that cannot start says why in under a second instead of timing out.
• Adding your own provider no longer erases the built-in engine. It kept the same slot, so the card vanished without a trace. The built-in engine now waits in standby with a labelled way back, and Disable on whichever provider holds the slot hands it over instead of leaving both sides wrong.
• A disabled provider keeps its card and an Enable button. Disable used to delete the card and the Enable along with it, which was a one way street on Ollama, Anthropic and LU Cloud as well.
• A provider you added can be removed from the interface. Until now the only ways out were Disable or a full reset.
• The LM Studio button in the picker sets the provider up. It started a server that nothing was configured to ask, so the model list stayed empty. The click now takes the slot over the same way the backend selector does, and says beforehand what it replaces and how to get back.
• Hints point at controls that exist. Four of them named a power button on Create, a gear icon and other things that were never there. Open Settings now lands in the right tab with the right section already open.
• The LU Cloud row shows the same five models every time and names how many more there are. The order was whatever arrived first from the server.

Chat and models in the conversation

• Thinking reaches the engine. The switch was set, the engine never saw it, and no bubble appeared. The signal now goes out on every path, including group chat, workflows, A/B compare, the benchmark and the phone relay, and it was proven on the wire in both positions.
• Local models with strict chat templates stopped failing. Tool results and repeated roles broke templates that refuse anything but a clean alternation, which is where a long chat or a group round from turn two would die. The system prompt reaches the engine first and in one piece, whatever the conversation looked like before.
• Every answer records the model that wrote it, and the chat reads that instead of your current selection, so an old conversation stops claiming a model it never ran on. Your pick also survives a restart.
• One stray click can no longer move the app into the cloud and bill you for it. The switch sits next to the model picker. Going into the cloud now takes a second click within six seconds, going back out stays a single click, and the composer shows which side you are on.
• The engine refuses a request that names a model it is not holding, and group chat loads each speaker's model before that speaker's turn instead of answering everything from whatever was loaded.
• An agent workspace folder follows the chat, not its title, so an automatic rename no longer loses the folder.

Hardware

• AMD cards on Windows are detected again. The check ran through wmic, which Microsoft removed in the August 2026 update, so on a current Windows the app believed there was no AMD card at all. It reads the registry now, with wmic left as a fallback.
• Windows with an RDNA3, RDNA3.5 or RDNA4 card gets AMD's own ROCm wheel channel instead of processor wheels and a recommendation for the frozen DirectML build. The channel is AMD's published one and is marked as researched rather than proven, because we have no such card here.
• On Linux, AMD cards the wheels do not cover stay on the processor build honestly. They used to pass detection, report a working device and then crash on the first kernel, which would have made 2.6.7 worse than 2.6.6 for those cards. The trainer refuses on them with a reason instead of starting a run that cannot finish.
• Cards from Turing up use the cu130 channel with cu126 as the fallback, checked live before it is chosen.
• The Debian and Ubuntu package stopped fighting llama.cpp over a file name (#120). The engine ships as lu-llama-server, so both packages can be installed in either order. We reproduced the collision against the real packages in a virtual machine and checked the fix there.

Updates, chats and stability

• Updating no longer risks your chats. The app waits for open chat writes before handing over to the installer, saves a fresh copy at that moment, and keeps three rotating backups instead of one. Proven against a hard kill in the middle of a write and against reinstalling over a running app.
• Closing the window gives the video memory back. The cross hides the app by design, but the engine sat there holding its model. After a short grace period it unloads the same way it does on a cloud switch, and comes back on its own when you use it again.
• The engine dies with the app on Windows. A crash used to leave it behind holding several gigabytes of video memory, and every restart leaked a handle. A crash also leaves a witness file now.
• Voice input says what actually went wrong. A dead transcription server reloads its model by itself and reports honestly if it cannot, refusals reach you in their own words, and the microphone hint no longer blames a setting the app had switched off itself.
• Error messages are English on a non English system. Windows writes its error text in the system language and we were passing it through, so a German or French machine showed a half translated failure in an English app. The number stays, the text is ours.

Security

• The model size check stopped following file names it was given. A name like an absolute path made the check look at that file and answer whether it exists and how large it is. A hostile or intercepted ComfyUI had an existence and size oracle for arbitrary paths on your machine. Names now run through the same filter the delete path uses. A rejected name gets the ordinary not found answer and the app never touches the disk for it. Nested names like sdxl/pony.safetensors keep working.

Cloud

• A very long hosted conversation is trimmed to fit instead of refused. The system prompt and the newest turns are kept, tool call pairs are never split, and only if the provider still refuses does a clear message with the code context_exceeded appear.

Download

Windows: Download the .exe installer. During setup, choose "Install for current user only" for a portable/no-install experience. Or use the .msi for system-wide installation.

Linux: Download the .AppImage, make it executable (chmod +x), and run. No installation needed.

All versions are portable-friendly, no admin rights required. LU by LU Labs.

Updating: the app updates itself, or take the installer for your platform below. Your chats come along either way, which is the one thing in this release we tested hardest.