Skip to content

Add Vercel AI SDK adapter #239

@hidai25

Description

@hidai25

Why

The Vercel AI SDK is one of the most common ways teams ship agents today, especially TypeScript/Next.js shops. It's the most-requested missing framework in our adapter list.

EvalView currently has adapters for HTTP, Anthropic, OpenAI, LangGraph, CrewAI, Pydantic AI, Aider, Goose, MCP, and more — see evalview/adapters/. Vercel AI is the gap.

What

Add evalview/adapters/vercel_ai_adapter.py that wraps a Vercel AI SDK agent endpoint (or local script invocation) and produces the same AgentResult shape every other adapter returns.

Acceptance criteria

  • New adapter file following the AgentAdapter ABC in evalview/adapters/base.py
  • Registered in evalview/adapters/registry.py
  • Tool calls + final output captured and round-tripped through the diff engine
  • At least one integration test under tests/adapters/ mirroring the existing adapter tests
  • README adapter table updated

Hints

  • Look at anthropic_adapter.py and openai_assistants_adapter.py for the closest patterns.
  • Tool calls in the Vercel AI SDK live on the streamed response — capture them into the trace as they arrive.

Size

~1-2 hours for someone who's used the Vercel AI SDK before.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or requesthelp wantedExtra attention is needed

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions