Getting Started

Installation

Using pip


pip install glassgen

Local Development Installation

Clone the repository:


git clone https://github.com/glassflow/glassgen.git
cd glassgen

Create and activate a virtual environment:


python -m venv venv
source venv/bin/activate  # On Windows use: venv\Scripts\activate

Install the package in development mode:


pip install -e .

Install development dependencies:


pip install -r requirements-dev.txt

Run tests to verify installation:


pytest

Basic Usage

Python SDK

Here’s a simple example of using GlassGen to generate user data and save it to a CSV file:


import glassgen
 
config = {
    "schema": {
        "name": "$name",
        "email": "$email",
        "country": "$country",
        "id": "$uuid",
        "address": "$address",
        "phone": "$phone_number",
        "job": "$job",
        "company": "$company"
    },
    "sink": {
        "type": "csv",
        "params": {
            "path": "output.csv"
        }
    },
    "generator": {
        "rps": 1500,
        "num_records": 5000
    }
}
 
result = glassgen.generate(config=config)
# result is a dict: {"time_taken_ms": ..., "num_records": ..., "sink": "csv"}

Configuration File

You can also load the configuration from a JSON file:


import glassgen
import json
 
with open("config.json") as f:
    config = json.load(f)
 
glassgen.generate(config=config)

CLI


glassgen generate-data --config config.json

generate() Reference


glassgen.generate(config, schema=None, sink=None)

config (required): A configuration dict or GlassGenConfig object.
schema (optional): A custom BaseSchema instance. When provided, the schema block in config is ignored.
sink (optional): A BaseSink instance or a sink config dict. When provided, the sink block in config is ignored.

Return value:

For all sinks except yield: returns a dict with time_taken_ms, num_records, and sink.
For the yield sink: returns a generator that yields one record at a time. See the Yield Sink page for details.

generate_one() Reference

To generate a single record without any sink or generator config, use generate_one():


import glassgen
 
record = glassgen.generate_one({
    "id": "$uuid",
    "name": "$name",
    "email": "$email"
})
# {"id": "...", "name": "...", "email": "..."}

Generator Configuration

The generator block controls how records are produced:


{
  "generator": {
    "rps": 1000,
    "num_records": 5000,
    "bulk_size": 5000
  }
}

Field	Default	Description
`rps`	`0`	Target records per second. `0` means generate as fast as possible with no rate limiting. Values above 2500 also skip rate limiting.
`num_records`	`100`	Total records to generate. Set to `-1` for infinite generation.
`bulk_size`	`5000`	Internal batch size. Tune this to adjust memory usage vs. throughput.

Infinite generation

Set num_records to -1 to generate records indefinitely (useful with the yield sink or streaming sinks):


config = {
    "schema": {"id": "$uuid", "value": "$int"},
    "sink": {"type": "yield"},
    "generator": {"rps": 100, "num_records": -1}
}
 
for record in glassgen.generate(config=config):
    process(record)

Event Duplication

GlassGen can simulate real-world data streams that contain duplicate events. Configure it under generator.event_options:


{
  "generator": {
    "rps": 1000,
    "num_records": 10000,
    "event_options": {
      "duplication": {
        "enabled": true,
        "ratio": 0.1,
        "key_field": "id",
        "time_window": "1h"
      }
    }
  }
}

Field	Required	Description
`enabled`	yes	Turn duplication on or off.
`ratio`	yes	Fraction of records that will be duplicates (0–1). `0.1` means ~10% duplicates.
`key_field`	yes	The schema field used to identify a record for duplication. Must exist in the schema.
`time_window`	no (default `1h`)	How far back to look when picking a record to duplicate. Supports `s`, `m`, `h`, `d` suffixes.

Next Steps

Learn about available generators
Explore sink options