@vercel/agent-eval@0.0.12

Patch Changes

• #18 85bfb21 Thanks @paoloricciuti! - feat: add editPrompt config to experiment

@vercel/agent-eval-playground@0.0.5

Patch Changes

• 6159d01 Thanks @allenzhou101! - Run playground in production mode (next start) instead of dev mode (next dev) to fix React version conflicts and "Cannot read properties of null (reading 'useInsertionEffect')" errors when running via npx.

@vercel/agent-eval-playground@0.0.4

Patch Changes

• 23e2d43 Thanks @allenzhou101! - Add repository field to package.json to fix npm provenance verification error during publishing.

v0.0.11

Patch Changes

• 558abe5 Thanks @paoloricciuti! - feat: accept array of models in experiment #10

v0.0.8

Version 0.0.8

v0.0.7

What's Changed

Fixes

• CLI loads .env.local: CLI now loads .env.local before .env (matching integration test behavior)
• CLI version from package.json: Version is now read dynamically instead of being hardcoded

Changes

• Default timeout increased: Changed from 5 minutes to 10 minutes (600s)

v0.0.6

What's Changed

Fixes

• [OpenCode] Fix model format: OpenCode now requires vercel/ prefix in model strings (e.g., vercel/minimax/minimax-m2.1)
• [Docker] Fix file permissions: Added chown after file upload so agents can edit files in the sandbox
• Timeout enforcement: Added Promise.race at runner level and signal abort to agent on timeout for proper resource cleanup

New

• Minimax model support: Added vercel/minimax/minimax-m2.1 to supported models
• Parallel integration tests: All agents now tested on both Docker and Vercel sandboxes concurrently
• [Config] Sandbox backend selection: Moved sandbox backend selection from env var to experiment config (sandbox: 'docker' | 'vercel' | 'auto')

Documentation

• Updated README with correct model format and examples

v0.0.5

What's Changed

• Added Docker sandbox as alternative to Vercel Sandbox
• Added test:integration:docker and test:integration:vercel scripts