JSON Schema to BAML

A Python utility that converts JSON Schema definitions to BAML types for use with the BAML functions.

BAML BAML Dyanmic Types

Overview

This tool allows you to automatically convert Pydantic models and their JSON schemas into BAML types, making it easy to use your existing data models with BAML's LLM functions.

⚠️ This is just sample code! You can copy and paste the library in your own codebase and modify as needed.

Installation

# Install using pip
pip install json-schema-to-baml

# Install dependencies
pip install -r requirements.txt

# Install using uv
uv sync

Generate the baml_client

Using uv:

uv run baml-cli generate

If you installed via pip:

baml-cli generate

Setup VSCode

• In command palette (command + shift + p), select "Python: Select interpreter", and select the .venv directory that was generated (if using uv)

Run

Next make sure you add a .env file, and you can run this script:

uv run dotenv run python test_one.py

dotenv run python test_one.py

Understanding the Code

# my_types.py
# File containing your Pydantic model
from pydantic import BaseModel
from enum import Enum

class Gender(str, Enum):
  male = 'male'
  female = 'female'
  other = 'other'
  not_given = 'not_given'

class MainModel(BaseModel):
  name: str
  age: int
  gender: Gender | None = None

from baml_client.type_builder import TypeBuilder
from json_schema_to_baml import parse_json_schema
# Create a type builder
tb = TypeBuilder()

# Convert your model's JSON schema to BAML types
from my_types import MainModel
schema = MainModel.model_json_schema()
baml_type = parse_json_schema(schema, tb)

// baml_src/example.baml
class DynamicContainer {
  @@dynamic
}

from baml_client import b
# Add the type to your BAML Type annotated with `@@dyanamic`
tb.DynamicContainer.add_property("data", baml_type)
# Use in BAML function calls
response = b.ExtractDynamicTypes(text, {"tb": tb})
# Parse the response
result = MainModel.model_validate(response.data)

# result.data is now a key-value pair that should work with your Pydantic model
# This will always work!
parsed = MainModel.model_validate(result.data)

# Required environment variables
OPENAI_API_KEY=your_openai_key_here