Back to Baml

Comparing AI SDK

fern/01-guide/09-comparisons/ai-sdk.mdx

0.222.09.6 KB
Original Source

AI SDK by Vercel is a powerful toolkit for building AI-powered applications in TypeScript. It's particularly popular for Next.js and React developers.

Let's explore how AI SDK handles structured extraction and where the complexity creeps in.

Why working with LLMs requires more than just AI SDK

AI SDK makes structured data generation look elegant at first:

typescript
import { generateObject } from 'ai';
import { openai } from '@ai-sdk/openai';
import { z } from 'zod';

const Resume = z.object({
  name: z.string(),
  skills: z.array(z.string())
});

const { object } = await generateObject({
  model: openai('gpt-4o'),
  schema: Resume,
  prompt: 'John Doe, Python, Rust'
});

Clean and simple! But let's make it more realistic by adding education:

diff
+const Education = z.object({
+  school: z.string(),
+  degree: z.string(),
+  year: z.number()
+});

const Resume = z.object({
  name: z.string(),
  skills: z.array(z.string()),
+  education: z.array(Education)
});

const { object } = await generateObject({
  model: openai('gpt-4o'),
  schema: Resume,
  prompt: `John Doe
Python, Rust
University of California, Berkeley, B.S. in Computer Science, 2020`
});

Still works! But... what's the actual prompt being sent? How many tokens is this costing?

The visibility problem

Your manager asks: "Why did the extraction fail for this particular resume?"

typescript
// How do you debug what went wrong?
const { object } = await generateObject({
  model: openai('gpt-4o'),
  schema: Resume,
  prompt: complexResumeText
});

// You can't see:
// - The actual prompt sent to the model
// - The schema format used
// - Why certain fields were missed

You start digging through the AI SDK source code to understand the prompt construction...

Classification challenges

Now your PM wants to classify resumes by seniority level:

typescript
const SeniorityLevel = z.enum(['junior', 'mid', 'senior', 'staff']);

const Resume = z.object({
  name: z.string(),
  skills: z.array(z.string()),
  education: z.array(Education),
  seniority: SeniorityLevel
});

But wait... how do you tell the model what "junior" vs "senior" means? Zod enums are just string literals:

typescript
// You can't add descriptions to enum values!
// How does the model know junior = 0-2 years experience?

// You try adding a comment...
const SeniorityLevel = z.enum([
  'junior',  // 0-2 years
  'mid',     // 2-5 years  
  'senior',  // 5-10 years
  'staff'    // 10+ years
]);
// But comments aren't sent to the model!

// So you end up doing this hack:
const { object } = await generateObject({
  model: openai('gpt-4o'),
  schema: Resume,
  prompt: `Extract resume information.
  
Seniority levels:
- junior: 0-2 years experience
- mid: 2-5 years experience
- senior: 5-10 years experience  
- staff: 10+ years experience

Resume:
${resumeText}`
});

Your clean abstraction is leaking...

Multi-provider pain

Your company wants to use different models for different use cases:

typescript
// First, install a bunch of packages
npm install @ai-sdk/openai @ai-sdk/anthropic @ai-sdk/google @ai-sdk/mistral

// Import from different packages
import { openai } from '@ai-sdk/openai';
import { anthropic } from '@ai-sdk/anthropic';
import { google } from '@ai-sdk/google';

// Now you need provider detection logic
function getModel(provider: string) {
  switch(provider) {
    case 'openai': return openai('gpt-4o');
    case 'anthropic': return anthropic('claude-3-opus-20240229');
    case 'google': return google('gemini-pro');
    // Don't forget to handle errors...
  }
}

// And manage different API keys
const providers = {
  openai: process.env.OPENAI_API_KEY,
  anthropic: process.env.ANTHROPIC_API_KEY,
  google: process.env.GOOGLE_API_KEY,
  // More environment variables to manage...
};

Testing without burning money

You want to test your extraction logic:

typescript
// How do you test this without API calls?
const { object } = await generateObject({
  model: openai('gpt-4o'),
  schema: Resume,
  prompt: testResumeText
});

// Mock the entire AI SDK?
jest.mock('ai', () => ({
  generateObject: jest.fn().mockResolvedValue({
    object: { name: 'Test', skills: ['JS'] }
  })
}));

// But you're not testing your schema or prompt...
// Just that your mocks return the right shape

The real-world spiral

As your app grows, you need:

  • Custom extraction strategies for different document types
  • Retry logic for flaky models
  • Token usage tracking for cost control
  • Prompt versioning for A/B testing

Your code evolves into:

typescript
class ResumeExtractor {
  private tokenCounter: TokenCounter;
  private promptTemplates: Map<string, string>;
  private retryConfig: RetryConfig;
  
  async extract(text: string, options?: ExtractOptions) {
    const model = this.selectModel(options);
    const prompt = this.buildPrompt(text, options);
    
    return this.withRetry(async () => {
      const start = Date.now();
      const tokens = this.tokenCounter.estimate(prompt);
      
      try {
        const result = await generateObject({
          model,
          schema: Resume,
          prompt
        });
        
        this.logUsage({ tokens, duration: Date.now() - start });
        return result;
      } catch (error) {
        this.handleError(error);
      }
    });
  }
  
  // ... dozens more methods
}

The simple AI SDK call is now buried in layers of infrastructure code.

Enter BAML

BAML was designed for the reality of production LLM applications. Here's the same resume extraction:

baml
class Education {
  school string
  degree string
  year int
}

enum SeniorityLevel {
  JUNIOR @description("0-2 years of experience")
  MID @description("2-5 years of experience")
  SENIOR @description("5-10 years of experience")
  STAFF @description("10+ years of experience, technical leadership")
}

class Resume {
  name string
  skills string[]
  education Education[]
  seniority SeniorityLevel
}

function ExtractResume(resume_text: string) -> Resume {
  client GPT4
  prompt #"
    Extract the following information from the resume.
    
    Resume:
    ---
    {{ resume_text }}
    ---
    
    {{ ctx.output_format }}
  "#
}

Notice what you get immediately:

  1. The prompt is right there - No digging through source code
  2. Enums with descriptions - The model knows what each value means
  3. Type definitions that become prompts - Less tokens, clearer instructions

Multi-model made simple

baml
// All providers in one place
client<llm> GPT4 {
  provider openai
  options {
    model "gpt-4o"
    temperature 0.1
  }
}

client<llm> Claude {
  provider anthropic  
  options {
    model "claude-3-opus-20240229"
    temperature 0.1
  }
}

client<llm> Gemini {
  provider google
  options {
    model "gemini-pro"
  }
}

client<llm> Llama {
  provider ollama
  options {
    model "llama3"
  }
}

// Same function, any model
function ExtractResume(resume_text: string) -> Resume {
  client GPT4  // Just change this
  prompt #"..."#
}

Use it in TypeScript:

typescript
import { b } from '@/baml_client';

// Use default model
const resume = await b.ExtractResume(resumeText);

// Switch models based on your needs
const complexResume = await b.ExtractResume(complexText, { client: "Claude" });
const simpleResume = await b.ExtractResume(simpleText, { client: "Llama" });

// Everything is fully typed!
console.log(resume.seniority); // TypeScript knows this is SeniorityLevel

Testing that actually tests

With BAML's VSCode extension, you can:

  1. Test prompts without API calls - Instant feedback
  2. See exactly what will be sent - Full transparency
  3. Iterate on prompts instantly - No deploy cycles
  4. Save test cases for regression testing

No mocking required - you're testing the actual prompt and parsing logic.

The bottom line

AI SDK is fantastic for building streaming AI applications in Next.js. But for structured extraction, you end up fighting the abstractions.

BAML's advantages over AI SDK:

  • Prompt transparency - See and control exactly what's sent to the LLM
  • Purpose-built types - Enums with descriptions, aliases, better schema format
  • Unified model interface - All providers work the same way, switch with one line
  • Real testing - Test in VSCode without API calls or burning tokens
  • Schema-Aligned Parsing - Get structured outputs from any model
  • Better token efficiency - Optimized schema format uses fewer tokens
  • Production features - Built-in retries, fallbacks, and error handling

What this means for your TypeScript apps:

  • Faster development - Test prompts instantly without running Next.js
  • Better debugging - Know exactly why extraction failed
  • Cost optimization - See token usage and optimize prompts
  • Model flexibility - Never get locked into one provider
  • Cleaner code - No wrapper classes or infrastructure code needed

AI SDK is great for: Rapid prototyping, simple use cases BAML is great for: Production structured extraction, multi-model apps, cost optimization, streaming UIs with semantic streaming

We built BAML because we were tired of elegant APIs that fall apart when you need production reliability and control.

Limitations of BAML

BAML does have some limitations:

  1. It's a new language (but learning takes < 10 minutes)
  2. Best experience requires VSCode

Ready for bulletproof structured extraction with full control? Try BAML.