Pipeline Steps

How pipelines work in Alphacast, and a map of every step available — from importing data to publishing the result.

A pipeline is an ordered chain of steps that turns source data into something you can publish, chart or download. Each step takes the output of the previous one as its input, so the data is reshaped progressively from the first step to the last.

Pipelines are reproducible: when a source dataset updates, the pipeline reruns and everything downstream follows. That is what makes a published dataset stay current without anyone touching it.

The shape of a pipeline

1. Import

   Every pipeline starts by bringing data in — from an existing Alphacast dataset, or from an external source like FRED or Yahoo Finance.

2. Filter and reshape

   Narrow the data to what you need and get it into the right shape: select columns, filter rows and entities, resample, join or pivot.

3. Transform and analyze

   Compute new variables, apply economic transformations, or run a domain model.

4. Output

   Publish a dataset, build a chart, or produce a downloadable snapshot.

Every step

Import data

Step	What it does
Fetch Alphacast Dataset	Pull an existing Alphacast dataset into the pipeline.
Fetch FRED Data	Import economic series from the St. Louis Fed by series code.
Fetch Financial Data	Import stocks, ETFs and indices by ticker.
Fetch Alphacast Series	Import individual series by ID, wide or long.
Fetch API Explorer Data	Import directly from a connected external provider.

Filter and select

Step	What it does
Select Columns	Keep only the columns you need.
Filter Entity	Keep only certain countries, categories or tickers.
Filter Dates	Restrict to an absolute or relative date range.
Filter Rows	Drop rows that don’t meet a condition.
Sort Rows	Order rows by one or more columns, each with its own direction.
Select Series	Reduce the data to specific variable and entity combinations.

Reshape

Step	What it does
Reorder Columns	Set the order columns appear in.
Rename Columns	Give columns new names.
Regroup Entities	Drop an entity and aggregate — the equivalent of GROUP BY.
Change Frequency	Resample to a different time frequency.
Merge with Dataset	Join two datasets on their common entities.
Append Dataset	Stack datasets with the same structure.
Long to Wide (Unstack)	Pivot from long format to wide.
Wide to Long (Melt)	Collapse from wide format to long.
Rename Values	Replace values inside an entity or string column.
Reorder Rows	Put entity values in a deliberate, non-alphabetical order.

Transform and analyze

Step	What it does
Apply Transform	Prebuilt transformations — YoY change, moving averages, per capita, constant prices.
Apply Country Transform	Country-specific conversions, with explicit control over the country.
Apply Formula	Apply one formula to many columns at once.
Calculate variable	Create columns with an Excel-like formula parser.
X13 ARIMA Seasonal Adjustment	Seasonal adjustment with full model control.
Validate Countries	Normalize country names against a canonical list.
Technical Analysis	132 technical indicators for financial assets.
Portfolio Analysis	Tear sheets and rolling performance stats.

The formula reference documents the grammar shared by Calculate variable, Filter rows and Apply Formula.

AI steps

Step	What it does
Transform with AI	Describe a transformation and let AI write the Python.
Summarize with AI	Generate a written summary of the data.
Create AI View	Build a view from the output of several earlier steps.

Outputs

Step	What it does
Publish to Dataset	Write the result to an auto-updating dataset.
Create a Chart	Build and customize a chart.
Create a Table	Render the data as a table for a dashboard.
Create a Deck of views	Several views over one dataset, with shared filters.
Data Snapshot	Produce a downloadable, embeddable snapshot.
Add markdown text	Document the pipeline inline.

Pipelines can also be built and run by AI agents. The Alphacast MCP server exposes the full pipeline lifecycle — creating pipelines, adding and editing steps, previewing output and triggering runs. See Pipeline tools.

Next steps

Fetch dataset

Start a pipeline from an existing dataset.

Formula reference

Every operator and function available in a formula field.