Coming from…

Coming from SQL, pandas or dbt

If you work in SQL, pandas, Polars or dbt, you already know every step Locra offers. The nodes are those steps with a form in front of them. Underneath, each node is code you can read, and you can always drop back to writing SQL or Python yourself.

What feels familiar

It runs on DuckDB. Most nodes run as DuckDB SQL, so DuckDB functions and types behave the way you expect. The Metadata tab shows each column's DuckDB type.
Write SQL when it's faster. A DuckDB Query node takes any query, CTEs and window functions included.
Python is one click away. Every node's code is Python. Click the pencil to edit it by hand, or use a Custom Node for your own function. A Code Gen node writes the Python from a plain-language description.
Your workflow can leave the app. Export as Python library turns the project into an installable Python package. One call runs the whole workflow, and each node can be imported on its own.

What's different

Edges replace ref() and variable names. You wire one node's output to the next instead of naming tables.
Every step keeps its result. Select any node to see its rows under the canvas. Run recomputes only the nodes that are out of date, so you don't re-run a whole notebook to check one step.
Tests are nodes too. dbt-style tests become Assert (stops the run) or Expectations (reports, never stops).
No upserts yet. Write to DB replaces or appends a table. MERGE and ON CONFLICT have no equivalent yet.

SQL/pandas names in Locra

You don't need this table open while you work. Type a SQL/pandas name into the add-node menu (right-click the canvas) or the command palette (⌘K, Ctrl+K on Windows). The matching Locra node shows up with a small chip, such as SQL/pandas: GROUP BY, so you learn our name as you go.

SQL

SQL/pandas name	Locra
SELECT	Select Columns or Add Column
SELECT * EXCLUDE	Select Columns
WHERE	Filter or Replace Values
CASE WHEN	Add Column or Switch
CAST	Cast Types
TRY_CAST	Cast Types or Parse
COALESCE	Fill Missing or Add Column
GROUP BY	Aggregate or Group Aggregate
HAVING	Filter or Aggregate
ORDER BY	Sort
LIMIT	Row Limit or Top / Bottom N
DISTINCT	Distinct
DISTINCT ON	Distinct
JOIN	Join or Lookup
LEFT JOIN	Join or Lookup
CROSS JOIN	Cross Join
ASOF JOIN	As-of Join
ANTI JOIN	Join
UNION ALL	Union
UNION BY NAME	Union
EXCEPT	Join or Compare Tables
PIVOT	Pivot
UNPIVOT	Unpivot
UNNEST	Explode or Split to Rows
ROW_NUMBER	Add Row Number or Rank
RANK	Rank
LAG	Shift / Lag or Multi-row Formula
SUM() OVER	Running Aggregations or Group Aggregate
QUALIFY	Filter or Rank
WITH	DuckDB Query
STRING_AGG	Aggregate
date_trunc	Truncate Date/Time or Resample
generate_series	Generate Series
regexp_extract	Regex Extract
read_csv_auto	CSV Reader
read_parquet	Parquet Reader
COPY TO	Write CSV or Write Parquet
CREATE TABLE AS	Write to DB or Write Locra Table
SUMMARIZE	Describe or Column Profile
ST_Transform	Reproject
ST_Intersects	Spatial Join

pandas and Polars

SQL/pandas name	Locra
read_csv	CSV Reader
read_excel	Excel Reader
read_html	Web Page (Add data source, Web Page)
read_sql	DB Reader or DuckDB Query
read_xml	XML Reader
merge	Join or Lookup
merge_asof	As-of Join
concat	Union or Zip
agg	Aggregate
assign	Add Column
with_columns	Add Column or Multi-field Formula
astype	Cast Types
fillna	Fill Missing
dropna	Fill Missing or Filter
drop_duplicates	Distinct
sort_values	Sort
rename	Rename
drop	Select Columns or Filter
query	Filter or DuckDB Query
isin	Filter
str.contains	Filter
str.split	Split to Columns or Split to Rows
str.replace	Text Operations or Replace Values
explode	Explode or Split to Rows
melt	Unpivot
pivot_table	Pivot or Aggregate
crosstab	Pivot or Value Counts
cut	Bin
shift	Shift / Lag
pct_change	Diff / Pct Change
cumsum	Running Aggregations
rolling	Rolling Window
ewm	Exponential Weighted Mean
resample	Resample
value_counts	Value Counts or Column Profile
describe	Describe, or the Metadata tab under the canvas
nlargest	Top / Bottom N
head	Row Limit
to_datetime	Parse or Convert Serial Date
apply	Code Gen or DuckDB Query
to_parquet	Write Parquet
to_excel	Write Excel
to_sql	Write to DB
plot	Chart
sjoin	Spatial Join
to_crs	Reproject

dbt

SQL/pandas name	Locra
ref	Wire an edge from the upstream node instead
source	DB Reader or CSV Reader
seed	CSV Reader or Constant Table
materialized='table'	Write to DB
dbt run	The Run button (`⌘↵`), or Run pipeline in `⌘K`
dbt test	Assert or Expectations
not_null	Assert or Expectations
unique	Distinct or Assert
accepted_values	Assert or Expectations
relationships	Join or Assert
recency	Assert or Expectations
unique_combination_of_columns	Assert or Expectations
generate_surrogate_key	Surrogate Key or Row Hash
date_spine	Generate Series or Densify
snapshot	Compare Tables or Row Hash
dbt Cloud job	Deploy & schedule to cloud (Cloud runs)
dbt docs generate	Lineage is the canvas itself; see docs per node

No equivalent yet

Some SQL/pandas tools have no Locra node yet. Searching for them shows a hint instead of an empty list. These are the ones you are most likely to miss:

SQL/pandas name	What to do instead
MERGE INTO	Upsert not supported yet; Write to DB can replace or append
ON CONFLICT	Upsert not supported yet; Write to DB can replace or append
GROUPING SETS	Run several Aggregate nodes and stack them with Union
ROLLUP	Aggregate at each level, then Union the results for subtotals
CUBE	Aggregate per combination, then Union; or write it in DuckDB Query
CREATE MACRO	Put reusable SQL in a DuckDB Query node
EXPLAIN	No query plan view; each node's SQL is visible in its code
cumprod	No cumulative product; use a Multi-row Formula ([Row-1:col] * x)
get_dummies	No one-hot node yet; Pivot with count, or a DuckDB Query
read_feather	Convert to Parquet first; Feather isn't read yet
read_fwf	No fixed-width reader; read lines as CSV then Split to Columns
read_sas	SAS files aren't supported; export to CSV first
insert_by_period	Loop over periods and append with Write to DB
get_query_results_as_dict	Use the node's output table directly; no Jinja templating

Tips

Search by function name. Typing melt, QUALIFY or merge_asof finds Unpivot, Filter and As-of Join.
Use a query node as a bridge. Paste a query you already trust into a DuckDB Query node to get going, then split it into nodes where seeing each step's output helps.
Read the code a node runs. Select a node and open the Code tab to see its Python. Click the pencil there to edit it, and the node runs your version; Revert to generated code in the Attributes panel brings back the generated code.
Take it to production as a package. Export as Python library gives you a package you can install with pip. Credentials are never written into it: they come from environment variables, listed in an example file next to the package.