Introduction to Laminar SQL Editor
Laminar SQL Editor is a tool that allows you to write and execute SQL queries on your data stored at Laminar.
The main use cases for this are:
- Advanced analytics on your data
- Exporting any trace or evaluation result to a dataset
The SQL Editor is currently in beta. Some functionality may change. We value your feedback and suggestions.
Laminar provides some useful syntactic sugar over standard PostgreSQL so that you don’t
have to write explicit join
s and you don’t depend on our internal data model too much.
For detailed reference of the syntax, see the reference page.
Features and benefits
- Write and execute SQL queries on your data stored at Laminar
- Filtering:
- By date range
- By span name and path
- By span tags
- By trace costs, metadata, duration, etc.
- By evaluator scores
- By evaluation results
- Fast analytics on your data, e.g.
- Detailed breakdown of token count and cost by operation
- Detailed latency analysis
- Exporting results to Laminar datasets or labelling queues
Allowed queries and limitations
- Only
SELECT
queries are allowed. - Allowed tables (to select from):
spans
traces
evaluation_results
evaluation_scores
evaluations
datasets
dataset_datapoints
- Currently, adding
WITH
CTEs is not supported.
For detailed information about the tables and columns, see the reference page.
Getting started
Prerequisites
- You must have some data stored at Laminar, such as traces, evaluation results, or datasets.
Using the SQL Editor
In any page in your dashboard, click the Terminal icon in the top right corner.
You will see a sheet with an SQL editor.
Example query
This query will return the input and output of the spans in the last 3 days.
For more examples, see the examples page.
Viewing results
Results are displayed in a table or raw JSON view.
Exporting results
Once you have selected the results you want to export, click the “Export to Dataset” button.
Choose the dataset you want to export to and map the columns to the dataset data
, metadata
, and target
fields.