qlm/transformer

EDUCATION DATA, MADE INSPECTABLE

Different products.
One evidence format.

Watch a product export become Learnlog. Keep the original measures, trace every mapping, and bring records together for research.

Try a transformation

No sign-up inside the demo. No log uploads to a server. Get the Airflow pipeline ↗

MANY FORMATS IN
{ }product_export.json
▦progress_report.csv
↳practice_rows.jsonl
L↗
learnlog.jsonsource · mapping · record
SHARED STRUCTURE OUT
01

Transform an export

Processed in this tab

Loading mapping templates…

INPUT

Source export

sample.json
Provided synthetic sample
OUTPUT

Learnlog record

Awaiting input
Your data, with its context.

Run the transformation to inspect the
actual Learnlog output and provenance.

SOURCE → MAPPING → RECORDOriginal values preserved
  1. 1 Read export
  2. 2 Apply mapping
  3. 3 Seal records
02

Bring the records together

CROSS-PRODUCT EXPLORER

Many products.
A shared starting point.

Combine imports, inspect data coverage, and keep each product’s original measures. Start with the five provided samples.

0products selected
0imported rows
0Learnlog records
0product-specific metrics
Add a transformed export or combine the provided samples.

Imported rows by product

COUNT

No records in the dataset yet.

Product identity, source hashes and mapping profiles travel with the records. Learner identifiers are scoped by product; matching identifiers across products does not establish the same learner.

TAKE IT INTO YOUR WORKFLOW

Run this on your own exports.

Download the Airflow pipeline with five synthetic inputs, product mappings, Learnlog outputs and a shared coverage report. Start with the samples, then connect your authorized exports.

Apache AirflowCSV · JSON · JSONLSource provenance included
Download the Airflow pipeline 60 KiB ZIP · DAG, CLI runner, source and setup guideNo email required.
A FORMAT IS A STARTING POINT

Keep the conditions
with the number.

Learnlog gives records a common structure. A study still needs a defined population, comparable measures, assistance conditions, and an appropriate design.

What are the provided samples?

They are synthetic examples for the five mapping templates, not student data or validated vendor integrations. MATHia’s progress-report columns are based on published documentation. The other product templates are illustrative.

Can I use my own export?

Yes, if its fields match the selected mapping profile. Open “Field mapping” or download the sample to inspect the expected structure. CSV, JSON arrays and JSONL are supported. Data is processed in this tab and cleared when the page reloads.

What can I analyze across products?

You can combine imported observations, examine missingness and metric coverage, and trace values to their source. The listed metric values retain their source representation; unmapped source columns are omitted. Common JSON does not establish that products measure the same construct or that learners can be linked.

QLM TRANSFORMER

Mapping and research notes

This browser implementation demonstrates the five summary-import profiles from QLM Transformer v0.2.1. Provided examples are synthetic. Compatibility requires the expected source fields; no live connection to the named products is made.

What travels with a record

The source-byte hash, mapping hash, source row, original values, and product-scoped pseudonyms. The Learnlog envelope contains the transformed records and their hash chains. Download the mapping profile to retain its source-field names.

What summary exports do not establish

An imported progress row is an import observation. A cumulative hint count does not establish individual help events, assistance absence, or learning effects. Unknown assistance remains unknown.

Designing a product-agnostic study

  1. Define the outcome, population and comparison before pooling data.
  2. Check how each export measures that outcome and records help.
  3. Preserve missingness and exclude incompatible measures explicitly.
  4. Collect additional event or assessment data where the question requires it.

The combined view is descriptive. It supplies neither a product ranking nor a causal effect estimate. Original scores are not pooled across products.

Privacy and pseudonyms

Processing uses browser-local code. There is no analytics, network upload or persistent browser storage in this demo. Use de-identified inputs. Pseudonyms are scoped to the product and demo session. Reloading creates a new session key; this is not a production anonymization service.

Learnlog is a proposed format. This demo does not claim an adopted Internet Standard, certified interoperability or independent validation.