Skip to content

v0.6.0 — Story-driven generation, 7 domains, 1M-row scale

Choose a tag to compare

@rasinmuhammed rasinmuhammed released this 10 Apr 16:25
· 129 commits to main since this release

What's new in 0.6.0

One-liner API

import misata
tables = misata.generate("A SaaS company with 5k users and 20% churn")

### 7 supported domains
SaaS · Ecommerce · Fintech · Healthcare · Marketplace · Logistics · Pharmaeach with domain-realistic distributions (lognormal for monetary columns, Zipf for categories, real blood-type probabilities, etc.)

### Pre-generation schema validation
validate_schema() catches misconfigured schemas before generation and reports every issue at once with a human-readable error.

### LLM parser hardened
Retry logic with exponential backoff, markdown fence stripping, and graceful skip of malformed tables/relationships.

### Performance
1M+ rows across 5 related tables in ~1.5s. Full referential integrity preserved.

### New exports
misata.parse() · misata.generate() · misata.generate_from_schema() · SchemaConfig.summary() · LLMSchemaGenerator · StoryParser · validate_schema · validate_data