v0.1.8

This release upgrades the ComfyStream docker runtime to target torch 2.8.0+cu128 and addresses recent security fixes

What's Changed

• bump torch==2.8.0 by @eliteprox in #369
• chore(deps): bump next from 15.5.3 to 15.5.7 in /ui by @dependabot[bot] in #547
• fix(docker): update supervisord configuration for Comfystream API by @eliteprox in #549
• refactor(deps): move byoc / pytrickle dependency outside of comfystream package by @Copilot in #551
• fix: update the version of react to patched one by @JJassonn69 in #554
• fix(docker): update base image and ComfyUI version in Dockerfiles by @JJassonn69 in #553
• chore(ci): update GitHub Actions runner to self-hosted by @eliteprox in #556
• chore: update Tailwind CSS to v4 by @JJassonn69 in #557
• chore(ci): move tests back to github runner by @eliteprox in #561
• Bump comfystream package and UI to version 0.1.8 by @Copilot in #566

Full Changelog: v0.1.7...v0.1.8

v0.1.7

This release provides Livepeer network integration via pytrickle, allowing Orchestrators to deploy and advertise custom AI pipeline capabilities using livepeer/comfyui-base:stable.

This release also improves the ComfyStream Pipeline class interface with pause, resume, and stop methods for better control of running prompts.

TensorRT engines are built the same as for development use, refer to comfystream docs

For more information on the Bring-Your-Own-Container (BYOC) initiative, please refer to go-livepeer docs

What's Changed

BYOC Server

• fix(docker): update cuDNN version in Dockerfile to match PyTorch version by @eliteprox in #496
• fix(byoc): Use correct capability url for registration by @eliteprox in #460
• docker(byoc): add default entrypoint, set log level to INFO, fix custom node blacklist by @eliteprox in #464
• feat(pipeline): implement pause and resume functionality for prompt exec control by @eliteprox in #499

Developer Improvements

• add workflow to publish to registry by @eliteprox in #121
• chore: add ruff config to precommit hooks by @eliteprox in #507
• fix(devcontainer): include constraints for pip install -e . by @eliteprox in #538

UI

• ui: fix deprecated husky install command by @eliteprox in #445
• Fix text output panel visibility - hide by default, show on first text output by @Copilot in #428

Dependency updates

• chore(deps): bump actions/checkout from 4 to 5 by @dependabot[bot] in #355
• chore(deps): bump actions/setup-node from 5 to 6 by @dependabot[bot] in #483
• pin setuptools<81 to resolve opentelemetry warning by @eliteprox in #465
• chore(deps): bump actions/upload-artifact from 4 to 5 by @dependabot[bot] in #485
• chore(deps): bump github/codeql-action from 3 to 4 by @dependabot[bot] in #463
• chore(deps): bump actions/setup-python from 5 to 6 by @dependabot[bot] in #375
• chore(deps): bump softprops/action-gh-release from 1 to 2 by @dependabot[bot] in #456
• chore(deps): bump actions/download-artifact from 4 to 6 by @dependabot[bot] in #482
• chore(deps): bump actions/checkout from 5 to 6 by @dependabot[bot] in #515
• chore(deps-dev): bump js-yaml from 4.1.0 to 4.1.1 in /ui by @dependabot[bot] in #511
• Bump version to 0.1.7 by @Copilot in #536

Full Changelog: v0.1.6...v0.1.7

v0.1.6

Release Notes Summery

This release introduces several major features and enhancements, including support for "pytrickle" in ComfyStream BYOC workflows, the addition of FasterLivePortrait and updated opencv-cuda support for Python 3.12, and the debut of ComfyUI-Manager with blacklist loading for improved performance. The UI has been refreshed to remove inline elements, and all documentation links have been updated to point to the new docs.comfystream.org. Models now use TAESD safetensors instead of PTH files for better efficiency. Critical bug fixes include updates to the pyproject.toml for ComfyUI registry compatibility, improved Docker image dependencies, added input timeouts to tensor loading nodes, and enhanced logging for input exceptions. Audio transcription workflows now use the correct sample rate, and unnecessary GitHub workflow triggers have been removed. Numerous dependency updates have also been made, such as bumping pytrickle, replacing deprecated pynvml, and upgrading Next.js, setup-node, and brace-expansion. The version is now bumped to 0.1.6, reflecting these improvements and fixes.

Download the docker images for comfystream from official livepeer dockerhub.

What's Changed

• fix: update pyproject.toml metadata for comfyui registry by @eliteprox in #406
• chore(deps): bump next from 15.2.4 to 15.5.3 in /ui by @dependabot[bot] in #392
• workflows: fix sample rate in audio transcription by @eliteprox in #408
• feat(byoc): add trickle to comfystream by @eliteprox in #357
• remove ai-runner gh workflow trigger by @eliteprox in #414
• fix(nodes): add input timeout to LoadTensor and LoadAudioTensor nodes by @eliteprox in #423
• cleanup(nodes): Improve input exception logging by @eliteprox in #427
• Update documentation links to docs.comfystream.org by @eliteprox in #431
• launch.json: update capability name for byoc by @eliteprox in #437
• fix(docker): add missing development libraries for Cairo and Pango by @eliteprox in #438
• feat: GitHub workflow opencv-cuda compilation by @JJassonn69 in #432
• chore: update opencv-cuda for Python 3.12, add FasterLivePortrait by @JJassonn69 in #413
• chore(deps): bump actions/setup-node from 4 to 5 by @dependabot[bot] in #377
• chore(deps-dev): bump brace-expansion from 1.1.11 to 1.1.12 in /ui by @dependabot[bot] in #443
• Feat/updated UI remove inline by @JJassonn69 in #444
• models: use taesd safetensors instead of pth by @eliteprox in #440
• feat: add ComfyUI-Manager, blacklist loading for performance by @eliteprox in #439
• chore(deps): bump pytrickle 0.1.4, replace deprecated pynvml with nvidia-ml-py by @eliteprox in #453
• Chore/update workflow manual by @JJassonn69 in #457
• Bump version to 0.1.6 by @JJassonn69 in #458

Full Changelog: v0.1.5...v0.1.6

v0.1.5

This release adds support for audio driven workflows and text outputs. An example real-time text transcription demo workflow using Whisper has been added.

The docker ComfyUI workspace has been upgraded to v0.3.56. ComfyStream is updated to support newer versions of ComfyUI

What's Changed

• Fixes issues with importing the custom nodes and having comfyui in site-packages by @doctorpangloss in #374
• fix: bump python to 3.12 for comfyui upgrade by @eliteprox in #379
• fix: upgrade comfyui workspace in dockerfile to be compatible by @eliteprox in #380
• fix launch config, patch comfyui workspace with install.py by @eliteprox in #381
• fix: resolve convert_prompt failure by @eliteprox in #391
• feat: add rtc data channel and text output for audio transcriptions by @eliteprox in #388
• bump pkg ver to 0.1.5 by @eliteprox in #393

New Contributors

• @doctorpangloss made their first contribution in #374

Full Changelog: v0.1.4...v0.1.5

v0.1.4

This release adds support for RTX 5090 GPUs using PyTorch 2.7.1+cu128. The minimum supported NVIDIA CUDA driver version is 570.124.06 with CUDA 12.8

This is a fast-follow release to v0.1.3

What's Changed

• deps: update pytorch for blackwell architecture by @eliteprox in #327
• chore: Revert the remote and branch names by @JJassonn69 in #344
• bump package to ver 0.1.4 by @eliteprox in #358

Full Changelog: v0.1.3...v0.1.4

v0.1.3

What's Changed

• docs: Adding workflow for FasterLivePortrait by @JJassonn69 in #286
• fix: pin numpy<2.0 by @eliteprox in #288
• fix devcontainer stuck loading in post-script by @eliteprox in #284
• fix: bind to correct host in server manager by @eliteprox in #73
• fix conda env install by @eliteprox in #310
• improve volume mount for persistent storage by @eliteprox in #184
• feat: add --api and --ui flags to entrypoint.sh by @eliteprox in #317
• fix: hostname arg for ui entrypoint flag by @eliteprox in #322
• refactor: update base image to CUDA 12.8 and TensorRT 10.12 by @JJassonn69 in #321
• release: 0.1.3 bump comfystream ver by @eliteprox in #325

Full Changelog: v0.1.2...v0.1.3

v0.1.3-pre

What's Changed

• docs: Adding workflow for FasterLivePortrait by @JJassonn69 in #286
• fix: pin numpy<2.0 by @eliteprox in #288
• fix devcontainer stuck loading in post-script by @eliteprox in #284
• fix: bind to correct host in server manager by @eliteprox in #73
• fix conda env install by @eliteprox in #310
• feat: improve volume mount for persistent storage by @eliteprox in #184
• feat: add --api and --ui flags to entrypoint.sh by @eliteprox in #317
• fix: hostname arg for ui entrypoint flag by @eliteprox in #322
• refactor: update base image to CUDA 12.8 and TensorRT 10.12 by @JJassonn69 in #321

Full Changelog: v0.1.2...v0.1.3-pre

v0.1.2

This releases adds a token-gated external MPEG stream egress endpoint to ComfySteam. A demo page is available to view the MPEG stream, which can also be used as an OBS Video Source to capture output for live streaming. Additionally a bug fix was made to comfystream.client.update_prompts and python packages have been pinned to improve stability.

What's Changed

• fix: error log by @varshith15 in #247
• pin xformers==0.0.30 and pytorch==2.7.0 by @eliteprox in #252
• Obs separate tab by @JJassonn69 in #85
• fix: update_prompts lock by @varshith15 in #251
• chore: bump version to 0.1.2 in pyproject.toml and package.json by @JJassonn69 in #253
• fix: version bug resolve for ComfyUI blocked at 0.3.27 by @JJassonn69 in #244
• Revert "fix: version bug resolve for ComfyUI blocked at 0.3.27 " by @eliteprox in #259
• fix: pin onnxmltools==1.13.0 by @eliteprox in #262
• docs: add contributing guide by @eliteprox in #42

Full Changelog: v0.1.1...v0.1.2

v0.1.1

What's Changed

• qualify volume mount copy with --server flag by @eliteprox in #152
• [COMFY-77] workflows: Remove separate build workflows for base and final images by @hjpotter92 in #178
• refactor(dev): remove deprecated Tensordock spinup script by @rickstaa in #111
• [COMFY-77] Update docker.yaml by @hjpotter92 in #181
• [COMFY-80] docker.yaml: Trigger workflow in ai-runner after comfyui-base is built by @hjpotter92 in #185

Full Changelog: v0.1.0...v0.1.1

v0.1.0

ComfyStream v0.1.0 includes multi-resolution support, an improved depth map and an exportable pipeline class for easier integration. This release is available as a docker image at livepeer/comfyui-base:v0.1.0 and livepeer/comfystream:v0.1.0

New Features / Nodes

• Super Resolution node with OpenCV
• Multi-resolution support with tensorrt engine compilation examples (default engine supports ranges of height/width from 448px to 744px)
• Exports the comfystream.pipeline class from Pipeline.py to make it easier to integrate ComfyStream into your own application through the python package namespace

New Models

• Depth Anything V2 large model with tensorrt compilation

NOTE: Due to a version dependency, ComfyStream must be run with ComfyUI version 0.3.27 or older https://github.com/comfyanonymous/ComfyUI/releases/tag/v0.3.27

What's Changed

• feat(models): add Depth Anything V2 Large model by @rickstaa in #79
• Support for super-resolution node with opencv-cuda by @JJassonn69 in #65
• update build_trt.sh args with min/max width/height by @pschroedl in #107
• Feat/dynamic resolution by @ryanontheinside in #38
• chore: update comfyui registry logo by @eliteprox in #115
• refactor(package): comfystream to export server and pipeline by @eliteprox in #114
• fix: exclude app.py from exported server package, add torch init by @eliteprox in #131
• fix: pin comfyui version by @eliteprox in #147
• release: update ver to v0.1.0 by @eliteprox in #151

New Contributors

• @pschroedl made their first contribution in #107
• @ryanontheinside made their first contribution in #38
• @pwilczynskiclearcode made their first contribution in #110

Full Changelog: v0.0.5...v0.1.0