The Wire · Showcase
NEXT.JS REQUEST INSIGHTS LANDS IN THREE PIECES, SLACK AGENTS SHIP, CLI HARDENS CODING SETUP
By RepoJournal · Filed · About Vercel
Next.js is building out its developer observability stack with a multi-part Request Insights feature that lets dev servers expose real-time span data and request history to tools and the browser, while Vercel's CLI is locking down coding agent setup across Codex, OpenCode, and now macOS Keychain.
The Request Insights work dropped as a coordinated three-part stack across the Next.js monorepo [1] [2] [3]. Part one records local OTEL framework spans without requiring a running provider, part two maintains a 100-request bounded history with deduped fetch records, and part three exposes snapshots over HMR and a private dev endpoint so tools can consume them in real time. This is the plumbing for what's coming next: agent diagnosis access and DevTools integration. Over on the CLI side, `ai-gateway coding-agents setup` locked down three high-friction pieces. Codex and OpenCode support [4] [5] now ship with proper config file handling (including byte-for-byte comment preservation in TOML), and the macOS Keychain integration [6] keeps API keys out of plaintext config by default, resolving them at runtime from shell rc exports instead. The setup command also got harder to break [7]: non-interactive runs now detect what's actually installed and error cleanly when nothing is found, and [8] warns users before configuring agents that would break the Codex desktop app. On the chat side, Slack Agent messaging experience support is in [9], complete with normalized app context events that let apps know what the user is viewing (channel, canvas, list, message), and code ownership is now scoped to package source [10] so docs and examples stop blocking reviews. [email protected] shipped [11] with proper framework attribution in the Build Output API, fixing a long-standing gap where `framework_slug` never populated in AI Gateway logs, while [email protected] [12] renames the attributes feature from experimental status and fixes namespaced queue recovery for persisted runs [13].
Action items
- → Review Next.js Request Insights PR stack if you're building developer tools or want real-time span visibility in dev mode vercel/next.js [plan]
- → Test `ai-gateway coding-agents setup` with Codex or OpenCode if you use those agents; Keychain integration changes the key storage model vercel/vercel [monitor]
- → Upgrade workflow to beta.31 if you use namespaced queues with persisted runs vercel/workflow [plan]
References
- [1] request insights: record local framework spans (1/5) ↗ vercel/next.js
- [2] request insights: derive request history and fetch data (2/5) ↗ vercel/next.js
- [3] request insights: expose dev snapshots to tools and HMR (3/5) ↗ vercel/next.js
- [4] [cli] coding-agents setup: add Codex support ↗ vercel/vercel
- [5] [cli] coding-agents setup: add OpenCode support ↗ vercel/vercel
- [6] [cli] coding-agents setup: store the key in the macOS Keychain (#16863) vercel/vercel
- [7] [cli] coding-agents setup: safer non-interactive defaults and failures ↗ vercel/vercel
- [8] [cli] coding-agents setup: warn before breaking the Codex desktop app ↗ vercel/vercel
- [9] feat(slack): add Slack Agent messaging experience (agent_view) support ↗ vercel/chat
- [10] chore: scope CODEOWNERS to package source (#694) vercel/chat
- [11] [email protected] ↗ vercel/eve
- [12] [email protected] ↗ vercel/workflow
- [13] Fix namespaced active run recovery ↗ vercel/workflow