The Python fallback path was reading amplitude directly from effect dict
instead of checking dynamic_params first like the CUDA kernel path does.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Downstream code expects arrays with .flags attribute, not GPUFrame.
Extract the underlying gpu/cpu array before returning.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The fallback path was passing raw numpy/cupy arrays to GPU functions
that expect GPUFrame objects with .cpu property.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Audio playback path was being resolved during parsing when database
may not be ready, causing fallback to non-existent path. Now resolves
lazily when stream starts, matching how audio analyzer works.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace slow scipy.ndimage operations with custom CUDA kernels:
- gpu_rotate: AFFINE_WARP_KERNEL (< 1ms vs 20ms for scipy)
- gpu_blend: BLEND_KERNEL for fast alpha blending
- gpu_brightness/contrast: BRIGHTNESS_CONTRAST_KERNEL
- Add gpu_zoom, gpu_hue_shift, gpu_invert, gpu_ripple
Preserve GPU arrays through pipeline:
- Updated _maybe_to_numpy() to keep CuPy arrays for GPU primitives
- Primitives detect CuPy arrays via __cuda_array_interface__
- No unnecessary CPU round-trips between operations
New jit_compiler.py contains all CUDA kernels with FastGPUOps
class using ping-pong buffer strategy for efficient in-place ops.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Build decord with -DUSE_CUDA=ON for true NVDEC hardware decode
- Use DLPack for zero-copy transfer from decord to CuPy
- Frames stay on GPU throughout: decode -> process -> encode
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace decord (CPU-only pip package) with PyNvCodec which provides
direct NVDEC access. Frames decode straight to GPU memory without
any CPU transfer, eliminating the memory bandwidth bottleneck.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Install decord in GPU Dockerfile for hardware video decode
- Update GPUVideoSource to use decord with GPU context
- Decord decodes on GPU via NVDEC, avoiding CPU memory copies
- Falls back to FFmpeg pipe if decord unavailable
- Enable STREAMING_GPU_PERSIST=1 for full GPU pipeline
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Central config with logging on startup
- Hot reload support for GPU worker (docker-compose.gpu-dev.yml)
- Quick deploy script (scripts/gpu-dev-deploy.sh)
- GPU/CPU frame compatibility tests
- CI/CD pipeline for GPU worker (.gitea/workflows/gpu-worker.yml)
- Standardize GPU_PERSIST default to 0 across all modules
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
All color_ops primitives now auto-convert GPU frames to numpy,
fixing compatibility with geometry_gpu primitives.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
GPU persistence returns CuPy arrays but most primitives expect numpy.
Disable until all primitives support GPU frames.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
streaming_gpu.py was being loaded on GPU nodes but had no PRIMITIVES dict,
causing audio-beat, audio-energy etc. to be missing. Now imports and
includes all primitives from the CPU streaming.py module.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The background IPFS upload task was running on workers that don't have
the file locally, causing uploads to fail silently. Now uploads go to
IPFS synchronously so the IPFS CID is available immediately.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add IPFSHLSOutput class that uploads segments to IPFS as they're created
- Update streaming task to use IPFS HLS output for distributed streaming
- Add /ipfs-stream endpoint to get IPFS playlist URL
- Update /stream endpoint to redirect to IPFS when available
- Add GPU persistence mode (STREAMING_GPU_PERSIST=1) to keep frames on GPU
- Add hardware video decoding (NVDEC) support for faster video processing
- Add GPU-accelerated primitive libraries: blending_gpu, color_ops_gpu, geometry_gpu
- Add streaming_gpu module with GPUFrame class for tracking CPU/GPU data location
- Add Dockerfile.gpu for building GPU-enabled worker image
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Update save_run_cache to also update actor_id, recipe, inputs on conflict
- Add logging for actor_id when saving runs to run_cache
- Add admin endpoint DELETE /runs/admin/purge-failed to delete all failed runs
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove legacy_tasks.py, hybrid_state.py, render.py
- Remove old task modules (analyze, execute, execute_sexp, orchestrate)
- Add streaming interpreter from test repo
- Add sexp_effects with primitives and video effects
- Add streaming Celery task with CID-based asset resolution
- Support both CID and friendly name references for assets
- Add .dockerignore to prevent local clones from conflicting
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
