Add a host-managed wait tool for Codex sessions

[Jules-Astier]

What Changed

• Register a namespaced t3.wait dynamic tool on new Codex threads.
• Run waits in the T3 server for 1 second through 1 hour, without periodic model inference.
• Cancel outstanding waits before turn interruption and when the provider session closes, including calls that race with cancellation.
• Decode the experimental raw thread/start response at the same boundary used by typed requests.
• Document the live-session contract and a separate follow-up architecture for durable timer/webhook resumes in docs/project/deferred-thread-resume.md.

The existing Codex adapter and web timeline already normalize and render dynamic-tool lifecycle items, so this does not add a parallel UI state.

Why

Agents currently monitor long-running commands, CI, deployments, and review work by repeatedly polling. Most checks contain no new information but still wake the model and consume inference. This gives T3 a focused host-owned primitive that keeps the current tool call open until a bounded timer expires.

This PR intentionally does not claim durable wake-up semantics. Restart-safe timers, external triggers, and PR webhooks need persisted waitpoints, idempotent delivery, and an explicit waiting lifecycle; those are designed in the included project doc as follow-up work.

Testing

• vp test run apps/server/src/provider/Layers/CodexDynamicTools.test.ts apps/server/src/provider/Layers/CodexSessionRuntime.test.ts — 28 passed
• vp run --filter t3 typecheck — passed (three existing Effect suggestions in orchestration/decider.ts)
• Targeted vp lint, vp fmt --check, and git diff --check — passed
• Live ephemeral smoke turn with codex-cli 0.144.6 — one t3.wait call, 1,006 ms elapsed, completed turn, expected assistant sentinel
• Macroscope correctness and Effect service convention checks — passed after addressing the cancellation-registration race

Checklist

• This PR is small and focused
• I explained what changed and why

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: 3a84ea43-e5bf-4263-9727-3fe0cdd5f85a

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}