jev-demos (DataGobes)
Data-engineering demos putting Jev inside DuckDB SQL, dbt tests, Databricks and charting, with live versus simulated runs labelled.
Overview
Jev demos by DataGobes are small, self-contained demos that place Jev inside tools data and analytics engineers already use, each scored honestly with live and simulated runs labelled and numbers taken from logged runs. Cringe-o-Meter scores a LinkedIn draft with eight parallel judgements and lets sliders reweight the composite without new calls. semsql adds semantic SQL to DuckDB, so a plain-English Noul such as 'asks for a refund' becomes a filterable column.
VISUALIZE ends a SQL query with an intent, has code propose only valid charts, and lets Jev score which best answers it. dbt semantic tests express data tests as English sentences to catch rows that pass structural tests but are wrong or leak PII, scored against a hidden answer key and a regex baseline; a Databricks version runs Jev in a Unity Catalog function with Delta caching and a 50,000-review production run audited against planted errors. Most demos run without a key in a SIMULATED mode.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this row's text match the English predicate (for example, asks for a refund)? | — |
| Score | How well does each candidate chart answer the query intent? | — |
| Yes / no | Does this row violate the English-sentence dbt test or leak PII? | — |
How it works
- Pick a demo folder and install its dependencies
- Run in SIMULATED mode or add TYPESAFE_API_KEY to its .env
- Jev answers typed questions inside SQL, dbt or a chart picker
- Compare against answer keys or baselines in logged runs
Strengths
- Integrates with DuckDB, dbt and Databricks
- Honest labelling of live versus simulated results
- Results scored against hidden keys and baselines
Caveats
- Each demo is small and self-contained, not a library
- Live runs need a TypeSafe key
Getting started
Open a demo folder, follow its README, and put TYPESAFE_API_KEY in that folder's .env for live results.At a glance
- Type
- Service
- Status
- prototype
- Language
- Python
- Runtime
- DuckDB, dbt, Databricks
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Every demo lives in its own folder with its own README, dependencies and instructions. Most run without an API key in a clearly labelled \`SIMULATED\` mode; put \`TYPESAFE\_API\_KEY=...\` in the demo folder's \`.env\` for live results.
logicrw/awesome-jev-projectsSmall, honest demos of TypeSafe's Jev inside tools data engineers already use.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Row-level predicates in SQL and dbt tests, chart fitness, and post-quality judgements.
Can I run them without a key?
Most run in a clearly labelled SIMULATED mode.
Has it been run at scale?
The Databricks demo reports a 50,000-review production run audited against planted errors.
Related projects
dbt package whose macros run Jev Choice, Noul and Score over SQL values in DuckDB and ClickHouse models.
ProjectbetaSQL (Jinja)DuckDB extension that asks Jev about every row in SQL and returns answers as ENUM, DOUBLE or STRUCT columns.
PluginbetaC++SQLite and Node library adding Jev judgments to SQL for matching, evidence picking, ranking, review queues and semantic CI checks.
ProjectbetaJavaScriptCLI that scans tabular, structured and text files or databases and asks Jev which fields look anomalous, building a review queue.
CLIbetaPythondbt package and Terraform module adding jev_noul, jev_choice, jev_score and jev_ask SQL functions to Snowflake.
LibrarybetaSQLNative MySQL plugin adding an AILIKE operator that asks Jev whether a value satisfies a natural-language condition.
PluginbetaC++