the models that needed a diffusers this install did not have (v0.5.11)

Krea-2-Turbo and LTX-2.5 have sat disabled with the same note: the classes
they name -- Krea2Pipeline, LTX2DurationHead, LTX2VideoDiffusionDecoderModel
-- did not exist in diffusers 0.38. They are in 0.42.1, a release rather than
git main, so the floor moves.

The pin that matters is the one next to it. diffusers 0.42 wants
huggingface_hub >=1.32, transformers 5.x refuses >=2.0, and pip resolves 2.3.0
given the chance -- which leaves transformers unimportable and every model
with it. The usable window is 1.33 and it is written down here rather than
rediscovered.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014T3rZnLhydX9QqUcfBbqLu

codai/__init__.py

@@ -16,7 +16,7 @@
 
 # Canonical product version for CoderAI — single source of truth. Both the API
 # metadata and the admin web UI read from here.
-__version__ = "0.5.9"
+__version__ = "0.5.11"
 __tagline__ = "Complete orchestration, distribution and escalation of remotizable advanced inference."
 
 # Configure the CUDA caching allocator BEFORE torch is imported anywhere.

packaging/runpod/profiles/audio.txt

@@ -8,7 +8,7 @@ transformers>=4.35.0
 # real A40 for the embeddings image: 303s of boot paid before anyone could find
 # out. Every profile that carries transformers carries this with it.
 accelerate>=0.26.0
-diffusers>=0.25.0
+diffusers>=0.42.1  # Krea2Pipeline and the LTX-2.5 classes; see requirements.txt
 demucs>=4.0.1
 deepfilternet>=0.5.6
 soundfile>=0.12.0

packaging/runpod/profiles/core.txt

@@ -16,7 +16,8 @@ setproctitle>=1.1
 Pillow>=10.0.0
 numpy>=1.24.0
 safetensors>=0.4.0
-huggingface_hub>=0.25.0
+huggingface_hub>=1.33.0,<2.0  # the window diffusers 0.42 (>=1.32) and
+# transformers 5.x (<2.0) share; resolving outside it breaks one of them
 
 # Imported at MODULE level somewhere in the graph `import codai.api.app` walks,
 # so every pod needs them regardless of profile — app.py mounts every router, and

packaging/runpod/profiles/images.txt

@@ -2,7 +2,7 @@
 torchvision
 transformers>=4.35.0
 accelerate>=0.24.0
-diffusers>=0.25.0
+diffusers>=0.42.1  # Krea2Pipeline and the LTX-2.5 classes; see requirements.txt
 ftfy>=6.0
 timm>=0.9.0
 realesrgan>=0.3.0

packaging/runpod/profiles/video.txt

@@ -3,7 +3,7 @@ torchvision
 torchaudio
 transformers>=4.35.0
 accelerate>=0.24.0
-diffusers>=0.25.0
+diffusers>=0.42.1  # Krea2Pipeline and the LTX-2.5 classes; see requirements.txt
 ftfy>=6.0
 imageio[ffmpeg]>=2.33.0
 timm>=0.9.0

requirements.txt

@@ -42,7 +42,12 @@ accelerate>=0.24.0
 geoclip>=1.2.0
 geopy>=2.4
 geographiclib>=2.0
-diffusers>=0.25.0  # For Stable Diffusion image generation
+diffusers>=0.42.1  # Stable Diffusion and every other diffusion pipeline.
+# The floor is 0.42 because that is where Krea2Pipeline and LTX-2.5's
+# LTX2DurationHead / LTX2VideoDiffusionDecoderModel landed; 0.38 loads neither.
+# It also moves the huggingface_hub floor: diffusers 0.42 wants >=1.32, while
+# transformers 5.x refuses >=2.0, so the usable window is narrow and pinned.
+huggingface_hub>=1.33.0,<2.0
 safetensors>=0.4.0  # Required by diffusers
 # Reads GGUF-quantized pipeline COMPONENTS (a Q4_K_M transformer served on the
 # base pipeline's VAE and text encoder — see build_gguf_pipeline_components).