Pipes & Workflows

Pipes & Workflows

A Pipe is what happens when a raw ComfyUI workflow meets a usable interface. Sweet Tea imports the workflow graph, analyzes its structure, and generates a clean form from the parameters it finds. The result is a tool you can use without ever opening a node editor — but with all the underlying power intact.

This chapter covers how to bring workflows into Sweet Tea, what happens during import, and how to keep your Pipes reliable over time.

Importing Workflows

You can bring workflows into Sweet Tea from several sources:

From a workflow JSON file

1. Open the Pipes section from the sidebar.
2. Click Import and select a ComfyUI workflow JSON file.
3. Sweet Tea reads the graph, checks dependencies, and creates a new Pipe.
4. The Pipe appears in your library with a form generated from the workflow schema.

From a .tea package

.tea is Sweet Tea's portable workflow format. It bundles the workflow graph with metadata, schema configuration, and packaging information.

1. Click Import and select the .tea file.
2. Sweet Tea unpacks and normalizes the contents.
3. The Pipe is ready to use (assuming dependencies are met).

Tip: .tea packages are the best way to share workflows between machines or users. They carry more context than raw JSON and are normalized for portability.

Comfy migration discovery (batch import)

If you already have a ComfyUI installation with lots of saved workflows and history, you don't need to import them one by one.

1. Open Settings.
2. Look for the Comfy migration section.
3. Studio scans your existing ComfyUI workflow files and history.
4. Candidates are deduplicated by graph fingerprint — if you have three copies of the same workflow with different filenames, Sweet Tea recognizes them as one.
5. Each candidate is classified by readiness: ready to import, incompatible (needs dependency work), or invalid (broken graph).
6. Select the candidates you want and import them in batch.

This is the fastest way for existing ComfyUI users to bring their workflow library over to Sweet Tea.

What Happens During Import

Sweet Tea doesn't just copy the file. The import process includes several steps that improve the Pipe's usability:

• Readiness classification — The workflow is checked against available nodes and models. You'll know before importing whether it's ready to run or needs dependency work.
• Deduplication — Graph fingerprinting catches duplicate workflows even when filenames or minor metadata differ.
• Orphan cleanup — Disconnected graph branches (nodes that aren't connected to any output) are pruned so the resulting Pipe stays clean.
• Schema generation — Sweet Tea reads the graph structure plus ComfyUI's object_info to determine what fields the form should expose, what types they are, and what defaults make sense.

What Makes a Good Pipe

After import, a Pipe is functional but may not be polished. A well-configured Pipe has:

• Clear field labels — Prompt and Negative Prompt instead of KSampler.text and KSampler.text.2
• Sensible core/advanced grouping — The fields you touch every run are in core view; niche parameters are in advanced
• Proven dependencies — All required models and nodes are installed and working
• At least one validation output — You've run it once and confirmed it produces good results

The Advanced Pipe Management chapter covers the schema editor where you rename, reorder, hide, and classify controls.

Import Hardening

After importing a new Pipe, run through this process before relying on it for production work:

1. Check dependencies — Open the Pipe and look for missing node or model warnings. Install what's needed using the Models Manager and Extension Manager.
2. Promote core controls — Open the schema editor and move the most-used fields to core view. Hide anything that doesn't need regular adjustment.
3. Rename confusing labels — Replace raw node parameter names with human-readable labels.
4. Run a validation test — Generate one image with default settings to confirm everything works end-to-end.
5. Save the known-good state — Note the settings that produced your validation output so you can always get back to a working baseline.

Warning: Don't skip the validation test. An import can succeed (no errors) but still produce broken output if a model is wrong or a node behaves differently than expected.

Versioning

Treat your Pipes like production tools. When they work, preserve that state:

• Avoid silently modifying a Pipe that's producing good results. If you want to experiment with the schema or graph, create a new version.
• Keep one known-good fallback version of any Pipe you depend on regularly.
• When you update a Pipe's schema, run a new validation test to confirm the changes don't break anything.

Packaging with .tea

When you want to share a Pipe, back it up, or move it to another machine, export it as a .tea package:

1. Select the Pipe in your library.
2. Choose Export or Package.
3. Sweet Tea normalizes the workflow, includes metadata, and produces a .tea file.

Before publishing a .tea to the web platform, verify that the package passes normalization checks and that dependencies are clearly documented.

Note: .tea packages are also what the web platform uses for distribution. When someone downloads a Pipe from sweettea.co/pipes, they're getting a .tea file.

Next: Advanced Pipe Management