BAML Has a Free API: Define LLM Functions as Typed Interfaces

Prompt engineering is string manipulation. BAML makes it type-safe function definitions.

What Is BAML?

BAML (Basically A Made-up Language) is a DSL for defining LLM functions with typed inputs and outputs. You define the interface, BAML generates the client code and handles parsing, retries, and validation.

// extract_resume.baml
class Resume {
  name string
  email string
  skills string[]
  experience Experience[]
}

class Experience {
  company string
  role string
  years int
}

function ExtractResume(resume_text: string) -> Resume {
  client GPT4o
  prompt #"
    Extract structured data from this resume:
    {{ resume_text }}

    {{ ctx.output_format }}
  "#
}

# Generate client code
npx @boundaryml/baml-cli generate

// Auto-generated, fully typed client
import { b } from './baml_client'

const resume = await b.ExtractResume("John Doe, [e-mail], 5 years at Google as Staff Engineer, skills: Python, Go, K8s")
console.log(resume.name)       // "John Doe" — typed as string
console.log(resume.skills)     // ["Python", "Go", "K8s"] — typed as string[]
console.log(resume.experience) // [{company: "Google", role: "Staff Engineer", years: 5}]

Why BAML

• Type-safe — LLM outputs are validated against your schema
• Auto-retry — if parsing fails, BAML retries with corrective prompt
• Any LLM — OpenAI, Anthropic, Gemini, Ollama
• IDE support — VS Code extension with live preview
• Testing — unit test your LLM functions with fixtures
• Streaming — partial typed objects as they generate

pip install baml-py  # or npm install @boundaryml/baml

---

Building AI applications? Check out my AI tools or email [e-mail].