The Wire · Showcase
VERCEL SHIPS AGENT SETUP, PYTHON WORKFLOWS, AND EVE STREAMING FIX
By RepoJournal · Filed · About Vercel
Claude Code agents now auto-configure through the CLI, Python services declare workflows in pyproject.toml, and eve 0.22.5 cuts streaming latency by decoupling provider reads from durable writes.
The CLI gained three agent-focused features overnight: `coding-agents setup` mints AI Gateway keys and configures Claude Code in one command [1], `ai-gateway coding-agents setup` idempotently writes gateway config to `~/.claude/settings.json` [1], and a new `setup` command provisions keys without requiring `--key` upfront [2]. Python developers got two layers of relief: pyproject.toml now supports subscriber services (Celery workers alongside FastAPI apps) [3] and workflow declarations [4], eliminating manual vercel.json wiring. Eve 0.22.5 shipped the fix everyone's been waiting for: provider streams no longer block on durable event writes, cutting perceived latency to near-zero while preserving strict FIFO ordering [5]. The release also added `experimental_chatgpt` model support (bill to your ChatGPT subscription instead of API keys) [6] and `eve extension` commands to scaffold and build reusable extensions [7]. Workflow core got replay lineage tracking [8] and fixed a critical dashboard hang: queue namespace is now threaded through o11y actions so Eve runs replay without 504s [9]. Next.js docs landed a comprehensive Building guide explaining route table symbols and prerendering workflows [10], plus an incremental adoption path for Cache Components migration [11].
Action items
- → Test `vercel ai-gateway coding-agents setup` with Claude Code if you support agent use cases vercel/vercel [plan]
- → Migrate Python services to pyproject.toml declarations before next deploy vercel/vercel [plan]
- → Upgrade eve to 0.22.5 to fix streaming latency on production agents vercel/eve [immediate]
- → Review Workflow core changes if replaying eve runs from dashboard vercel/workflow [monitor]
References
- [1] [cli] ai-gateway: add coding-agents setup (core, Claude Code) ↗ vercel/vercel
- [2] [cli] coding-agents setup: provision an API key ↗ vercel/vercel
- [3] [services] compose `pyproject.toml` subscribers into `services` ↗ vercel/vercel
- [4] [python] support declaring workflows in `pyproject.toml` ↗ vercel/vercel
- [5] fix(eve): streaming - preserve throughput during durable writes ↗ vercel/eve
- [6] feat(eve): add `experimental_chatgpt` model support ↗ vercel/eve
- [7] feat(eve): add eve extension init and eve extension build ↗ vercel/eve
- [8] feat(core): record replay lineage in run executionContext ↗ vercel/workflow
- [9] [core] Thread queue namespace through o11y run actions and bound healthCheck stream reads ↗ vercel/workflow
- [10] docs: add Building guide ↗ vercel/next.js
- [11] docs: add incremental adoption path to Cache Components migration guide ↗ vercel/next.js