Describe and design · Autessa Synthetic Data
Sample rows as few-shot examples
Autessa Synthetic Data accepts sample rows as few-shot examples. Paste a JSON array of a few example rows and the generator follows their style, tone and format in every batch, and the schema proposal uses them too.
The problem
Some things are easier to show than describe
The exact voice of a customer review, the way your SKUs are formatted, or how terse your agents’ notes are can take a paragraph to describe and still come out wrong. Two or three examples make it obvious.
What you get
- Paste a JSON array of example rows in "Sample rows (optional few-shot examples)".
- The examples steer style and format for every batch of the job.
- Proposing a schema with sample rows also shapes the suggested columns.
- The field is checked before the job starts: it must be valid JSON and an array.
- Sample rows are stored encrypted with your company key.
- Also available on the CLI and API as
--sampleRowsJson.
How it works
How does sample rows as few-shot examples work?
Open sample rows
In "Describe the data", expand "Sample rows (optional few-shot examples)".
Paste examples
Paste a JSON array, for example [{"name": "Ada Lovelace", "role": "Engineer"}]. A few varied rows work better than many similar ones.
Propose or generate
Propose a schema with the examples, or start the job. Every batch is generated with the examples as guidance.
Every option
Sample row rules
| Option | What it does |
|---|---|
| Format | A JSON array of objects, one object per example row. |
| Validation | "Sample rows must be valid JSON" and "must be a JSON array" errors are shown before the job starts. |
| Used by | Dataset jobs, schema proposal, and the generate command on the CLI and API. |
| Storage | Encrypted with your company key alongside the description. |
Permissions
Who can do what?
| Who | Can |
|---|---|
| People in your company | Create datasets, see the company list of jobs, open any job and review, approve, append, cancel or discard it. |
| API keys | Can run quick generate, propose schemas and run dataset file jobs. Table targets need a signed-in person. |
Where AI does the work
How the generator handles this
Sample rows are added to the generation prompt as few-shot examples for style. They guide the model; they are not copied, and generated rows are still checked and de-duplicated.
Generation runs on your company default model, and every row is checked against your columns before it is staged or written.
Your model and PII stripping →Connected
How sample rows as few-shot examples connects to the rest of Autessa
SYNTHETIC DATA FEATURE
Generate from a description
Examples and the description work together.
SYNTHETIC DATA FEATURE
LLM training data
Paste two real-quality pairs to set the voice of a training set.
SYNTHETIC DATA FEATURE
Your model and PII stripping
Strip personal data from prompts if you paste real examples.
FAQ
Questions about sample rows as few-shot examples
Should I paste real customer records as samples?
Only if you are comfortable sending them to your model. Hand-written examples work just as well. If you do paste real data, turn on company-wide PII stripping and treat the output with the same care.
How many sample rows should I add?
A few varied rows is usually enough. They show style and format; the description and column settings still decide the content.
Are sample rows required?
No. They are optional and hidden behind a toggle on the New dataset screen.
100% customizable, code and all
Make Autessa Synthetic Data fully your own.
Use it out of the box, then change anything. Ask Autobot, Autessa’s AI builder, for a new field, screen, report, workflow or agent and it builds it into your copy of the app. Or open the code and edit it yourself. Nothing is locked: the app, its data and its code are yours.
Start free and make it yours