Skip to content

README.md

Latest commit

 

History

History
235 lines (179 loc) · 11.8 KB

README.md

File metadata and controls

235 lines (179 loc) · 11.8 KB
Raw

CodeQL

diffusers-workflow

Your GPU, as something an agent can drive.

diffusers-workflow runs Hugging Face Diffusers image, video and audio generation as jobs on your own GPU. An agent does the work. Claude Code, or any MCP client, picks or writes the workflow, checks what it will cost, runs it, and looks at what came back. You steer from the browser. Every workflow, run and output the agent made is in the web UI, where you can change a variable, run it again, compare takes and keep the one you like.

Python 3.12-3.14 | CUDA (NVIDIA) | MPS (Apple Silicon) | CPU

A workspace's overview in the web UI: recent outputs from an agent-made musical trailer, its jobs, and the workflows the agent wrote

Getting started

1. Install. The script picks the right torch build for your platform, creates a virtual environment and installs everything, MCP server included.

bash ./install.sh && source ./activate      # Linux / macOS
.\install.ps1; .\venv\scripts\activate      # Windows

2. Start the engine. Leave it running; the agent and the browser both talk to it.

dw-serve        # web UI on http://127.0.0.1:8765

3. Connect your agent. Register the MCP server with the absolute path to dw-mcp in the venv you just made (a relative path is the setup detail that most often goes wrong). Then add the dw plugin:

claude mcp add dw -- "$(pwd)/venv/bin/dw-mcp"
/plugin marketplace add dkackman/diffusers-workflow
/plugin install dw@diffusers-workflow

Most of the shipped models (Flux, LTX, MiniMax, ...) are gated. Request access on the model's Hugging Face page and run hf auth login once. Without it a run fails partway through with a 401/403 from the Hub. templates/text-to-image uses a small, ungated model, so it makes a good first run.

GPU on another machine, or in Docker?

Start the engine on the GPU box with --mcp and connect over HTTP. There is nothing to install on the laptop:

# on the GPU box
dw-serve --host 0.0.0.0 --token "$DW_API_TOKEN" --mcp --workspace ~/studio

# on your laptop
claude mcp add --transport http dw http://gpu-box:8765/mcp \
  --header "Authorization: Bearer $DW_API_TOKEN"

The server's Status page writes that line for you. The whole setup is in Remote GPU server.

On Linux with an NVIDIA GPU, the dkackman/diffusers-workflow image runs the server with torch, the engine and the web UI built in. It shares the model cache and workspace with a native install on the same box. See Docker.

Ask the agent

Describe what you want. The agent has 65 tools covering the workflow catalog, the real diffusers pipeline signatures, the job queue, the gallery and the model cache, and it stays with a long render until it finishes:

Claude Code driving the dw MCP server: creating a workspace, authoring a script, and generating from it

A session sounds like this:

  • "Make a 30-second musical trailer: four men in boater hats, a giant eye, a score under it." The agent picks a shape from the catalog (list_workflows), breaks the brief into shots, writes each prompt to the prompt library, and builds the workflow that renders the shots and cuts them to the score.
  • "What will that cost?" validate_workflow checks the JSON against the schema and the real pipeline signatures, with no model loads, and quotes an estimate in GPU minutes from this box's own history. Anything that spends GPU time or disk refuses to run until the agent passes acknowledged_cost, so it has to tell you what it is about to spend first.
  • "How's it going?" wait_for_job waits out a render without polling. Then get_output_frames, get_output_image and get_output_audio bring the result back into the conversation, so the agent judges what it made against what you asked for.
  • "Shot three is the one. Keep it and start the next scene from it." keep_output turns the file into a named asset, and the next workflow references it as asset:hero-frame.png.

The dw plugin adds what a general-purpose agent can't know on its own: one skill per model family (MiniMax H3 and Music 3, LTX-2.5, Kandinsky 6, FLUX 3 Klein) that picks the template for a request and states the rules that bite. Cross-cutting skills cover script-to-video, series with a consistent cast, storyboards and animatics, and a shared camera-and-look vocabulary.

The same session hours later: shots rendering one at a time, with the agent reporting progress between them

One server holds several workspaces, each with its own workflows, assets and outputs. Two agents, or an agent and you in the browser, can share one GPU without saving over each other's work.

The tool reference, setup for other MCP hosts and the troubleshooting table are in MCP Server.

Then tweak it in the browser

The web UI shows everything the agent made and lets you take over at any point. It runs on the same persistent GPU worker, so models stay loaded between runs.

A workflow the agent wrote is a form you can edit. The workflow's description, recent outputs and variables are on one page. Change a prompt, a frame count or a LoRA scale and press Run.

A workflow page: the agent's description of the trailer, its recent shots, and its variables ready to edit and run

Every job is a recipe. A job keeps the exact workflow it ran and draws it as a graph of how data flows between steps. Click a step to see what it produced. The run's warnings sit above the graph, so a problem the agent missed, such as an audible level jump at a cut, is in front of you with the argument that fixes it. Run again repeats the job, and New seed gives you another take. Progress streams live for each denoising step, and a cancel stops the job cleanly at the next step.

A finished cut-and-score job: a warning about a level jump at a cut, then the workflow as a data-flow graph from the edit and soundtrack through resampling, mixing and normalizing to the final film

The gallery groups outputs by the workflow that made them, with final and intermediate files kept apart. Any image reopens as the workflow that produced it. Keep as asset adds a file to the asset library for later workflows to use.

The gallery: a poster workflow's four takes and the frames grabbed from the trailer's shots

The editor works on the agent's JSON. You can view the workflow as a form, as JSON, side by side, or as a flow graph. The forms and autocomplete come from the real diffusers pipeline signatures, so every argument a pipeline takes is there with its documentation. Validate catches schema errors and argument typos before any model loads.

The editor in split view: variables as a form beside the workflow JSON

There is also a prompt library, with an Enhance with AI panel that expands an idea into a full prompt using a local language model. The model manager shows the Hugging Face cache (size, last use, free disk) and downloads or deletes models with live progress.

The model manager: every cached model with its size and when it was last used

The pages are described in Server & Web UI. The UI and the agent both use the server's HTTP API, which you can browse at /docs on a running server (see The HTTP API).

Underneath

Every front end reads and writes the same thing: a JSON workflow of named steps, each a diffusers pipeline or a utility task. Its arguments are references to variables, earlier steps' results, stored prompts and assets, not hard-coded values. That is how text-to-image chains into image-to-video, and how a generated image reopens as the recipe that made it. You rarely need to write one by hand. When you do, the Workflow Guide is the reference, the JSON Schema is the exact structure, and workflows/ has runnable examples for each model family. There is also a command-line client, python -m dw.run workflow.json prompt="a cat", which queues a job on a running server (see Server & Web UI).

The engine also covers:

  • Quantization: BitsAndBytes, TorchAO, GGUF, SDNQ and optimum-quanto, set per component.
  • Acceleration: FirstBlockCache, FasterCache, MagCache, TaylorSeerCache, torch.compile and attention backends.
  • Adapters and prompts: LoRA, IP-Adapter and A1111-style prompt weighting.
  • Long video and caching: long-video chaining with audio-driven length, and step-output caching, so re-running a fixed-seed workflow finishes instantly.
  • Utility tasks: upscaling, face restoration, segmentation, captioning, frame interpolation, audio slicing, cut-and-score assembly and more.

A workflow can reach any diffusers pipeline or quantization backend by dynamic import, so loading one can execute arbitrary Python. Treat a workflow file from someone else the way you'd treat a .py script. See Trust model.

Field reports. At the end of a session, ask the agent what got in its way, and have it file each bug or gap as an issue on dkackman/diffusers-workflow with the field-report label. The label marks a report as coming from real use. A report from anyone but the maintainer is reviewed before the agent loop acts on it.

Documentation

Using it

  • MCP Server - the agent's tool surface (Claude Code, Claude Desktop, other MCP hosts)
  • dw plugin - the model-family and composition skills
  • Server & Web UI - the web UI, the command line, and the HTTP API with its OpenAPI spec
  • Workspaces - where your content lives, run directories, several workspaces on one server
  • Remote GPU server - the server, UI and MCP from another machine
  • Docker - the server in a container on a Linux NVIDIA box

Workflows in depth

Reference