Create your first workflow in the Bethemesh Workspace

A standalone tool is ideal for a one-off operation. When the same task requires three or four consecutive transformations, repeatedly downloading and re-importing results becomes inefficient. The Bethemesh Workspace is designed to compose compatible operations into an explicit sequence and save the recipe as a pipeline once the workflow is stable.

The point is not to make simple tasks complicated. The Workspace becomes useful when repetition or chaining starts to cost time.

Tool, Workspace or pipeline?

Need	Best fit
One operation	standalone tool
Several operations to test together	Workspace
A stable sequence to reuse	pipeline

A pipeline stores the recipe-modules, order and parameters-not a general cloud copy of the files being processed.

Step 1: define the goal

Write the intended result in one sentence.

For example:

Import an image, resize it, convert it to a web format, then export it.

Or:

Import audio, keep one section, normalize the level, add a fade and export it.

That sentence already suggests the basic pipeline order.

Step 2: identify the input

Every workflow starts with a source: a file, text, structured data or another compatible input.

The source determines what later modules can accept. An image module cannot arbitrarily consume a CSV table. Compatibility rules exist to prevent workflows that could never execute successfully.

Step 3: add one transformation

Start small.

For an image workflow, add resizing first. Configure it, run the pipeline and inspect the result before adding the next module.

Build incrementally:

source → transformation 1 → test

then:

source → transformation 1 → transformation 2 → test

If the result becomes wrong after the second step, you immediately know where to investigate.

Step 4: think about order

Compatible operations are not always interchangeable.

For images:

resize → compress

is usually more sensible than:

compress → resize → compress again

For audio, trimming a long unused section before certain later operations may avoid processing data that will be discarded.

Order should reflect the result you want, not just the list of operations.

Step 5: inspect intermediate results

When the final output is surprising, ask:

1. is the input correct?
2. at which step does the result start to diverge?
3. is that step configured as intended?

Changing several parameters at once makes debugging harder.

Step 6: reorder instead of rebuilding

A workflow is an executable draft.

If you have:

import → A → B → C → export

and B should come after C, try:

import → A → C → B → export

Then verify that data types remain compatible.

Step 7: add the right output

A workflow is useful only if the result can be consumed.

Depending on the data, the output may be a download, preview or transformed value. Do not add outputs by habit; choose what matches the original goal.

Example: image preparation

A web-preparation pipeline can look like:

Image → Resize → Convert → Compress → Download

Each step has a distinct purpose:

• resize: remove unnecessary pixels;
• convert: select target format;
• compress: control file size;
• download: retrieve the deliverable.

This becomes valuable when the same rule is applied repeatedly to new images.

Example: audio

A simple audio chain might be:

Import → Trim → Normalize → Fade → Export

Again, the value is repeatability. Tune the parameters once, verify the result, then preserve the structure.

Example: tabular data

A recurring export workflow could be:

Import → Clean → Deduplicate → Sort → Export

What used to require several manual passes becomes an explicit recipe.

When should you save?

Do not necessarily save the first draft.

Save when:

• the pipeline runs correctly;
• step order is stable;
• parameters make sense for future use;
• the name describes the result.

Prefer Prepare catalogue images 1200px over Test pipeline 4.

Local First and the Workspace

The Workspace organizes processing. It does not require turning Bethemesh into general cloud storage for your files.

Compatible local transformations run in the browser. Optional synchronization can concern reusable recipes and account-level features without changing the principle that file processing itself can stay local.

Mistakes to avoid

Building ten steps before the first run. Test progressively.

Using the Workspace for a one-step job. A standalone tool is often faster.

Naming everything “test”. Future you will not know what it does.

Repeating lossy conversions. Media quality can degrade.

Ignoring order. A pipeline is a sequence, not a bag of modules.

The best first workflow has only two or three transformations. Test it step by step and save it only when it reliably produces the intended result.