Research AI Sdk

Research Vercel AI SDK Patterns

Use this skill when you need to:

• Learn AI SDK streaming patterns
• Understand multi-provider integration (Anthropic, OpenAI, Google, etc.)
• Research tool calling and function execution
• Find structured output patterns
• Learn RAG (Retrieval Augmented Generation) patterns

Process

1. Identify AI Use Case

   • Text generation, streaming, or structured output?
   • Tool calling or function execution?
   • Multi-modal (text, images, audio)?
   • Which provider (Anthropic, OpenAI, Google, etc.)?

2. Search Documentation (Ref)

   Query patterns:
   - "Vercel AI SDK [feature]"
   - "AI SDK streaming response"
   - "AI SDK tools function calling"
   - "AI SDK structured output Zod"
   

3. Find Implementation Examples (Exa)

   Query patterns:
   - "Vercel AI SDK streaming React component example"
   - "AI SDK tool calling Next.js server action"
   - "AI SDK Anthropic Claude function execution"
   - "AI SDK structured output Zod schema validation"
   

4. Consider Project Stack

   • Integration with Next.js Server Components/Actions
   • Type safety with TypeScript and Zod
   • Provider selection based on use case

Common Research Topics

Streaming

// Documentation
Query: "Vercel AI SDK streaming response useChat"

// Code examples
Query: "AI SDK streamText Next.js route handler streaming response"

Tool Calling

// Documentation
Query: "AI SDK tools function calling schema"

// Code examples
Query: "AI SDK tool execution Zod schema Next.js example"

Multi-Provider

// Documentation
Query: "AI SDK provider abstraction Anthropic OpenAI"

// Code examples
Query: "AI SDK switch providers Claude GPT-4 configuration example"

Structured Output

// Documentation
Query: "AI SDK structured output generateObject Zod"

// Code examples
Query: "AI SDK generateObject Zod schema typed response example"

RAG Patterns

// Documentation
Query: "AI SDK embeddings vector search RAG"

// Code examples
Query: "AI SDK RAG pattern document retrieval context injection example"

Provider-Specific Research

Anthropic (Claude)

Query: "AI SDK Anthropic Claude tool use streaming example"
Query: "AI SDK Claude 3.5 Sonnet function calling"

OpenAI (GPT)

Query: "AI SDK OpenAI GPT-4 function calling example"
Query: "AI SDK OpenAI streaming response Next.js"

Google (Gemini)

Query: "AI SDK Google Gemini multi-modal example"
Query: "AI SDK Gemini function calling tools"

React Integration

useChat Hook

// Documentation
Query: "AI SDK useChat hook React"

// Code examples
Query: "AI SDK useChat streaming messages Next.js example"

useCompletion Hook

// Documentation
Query: "AI SDK useCompletion hook"

// Code examples
Query: "AI SDK useCompletion text generation form example"

Server Actions

// Documentation
Query: "AI SDK Next.js server actions generateText"

// Code examples
Query: "AI SDK server action form submission AI response example"

Advanced Patterns

Context Management

Query: "AI SDK conversation history context management example"

Error Handling

Query: "AI SDK error handling retry fallback provider example"

Streaming with Tools

Query: "AI SDK streaming tool calls partial results example"

Multi-step Agents

Query: "AI SDK agent pattern multiple tool calls reasoning example"

Output Format

Provide:

1. Feature explanation - How the AI SDK feature works
2. Code examples - Real implementations from Exa
3. Provider comparison - When to use which provider
4. Type safety - Zod schemas and TypeScript integration
5. Best practices - Error handling, rate limiting, costs
6. Integration guide - How to implement in this project

Project Context

Your project uses:

• Vercel AI SDK 5.0.51
• Multiple providers: Anthropic, OpenAI, Google, Groq, Mistral, XAI
• Next.js 15 Server Components and Actions
• TypeScript with Zod validation
• React 19 with streaming

Common Use Cases

Chat Interface

Research: "AI SDK useChat streaming Next.js chat interface"

Document Analysis

Research: "AI SDK document analysis tool calling structured output"

Content Generation

Research: "AI SDK generateText streaming content generation"

Code Generation

Research: "AI SDK code generation Anthropic Claude structured output"

When to Use

• Implementing AI features
• Choosing AI providers
• Setting up streaming responses
• Creating tool/function calling
• Optimizing AI performance
• Debugging AI integrations
• Learning new AI SDK features

Related Skills

• research-nextjs (for Next.js integration)
• research-react (for React hooks)
• research-typescript (for type safety)