v0.4.1

v0.4.1 — Codex app-server stability

Fixes the Codex MCP-stack memory leak by disposing per-call app-server threads (reap-on-drop: keepThread + thread/delete) instead of letting them accumulate, and fast-fails wedged Codex requests with a 120s request timeout to end the intermittent ~30-minute hangs. See PR #3.

---

   🚀 Features

• Add gpt-5.6-sol/gpt-5.5 to model enum; schema tells callers to omit model so config.toml default applies  -  by @rbutera (2adca)
• Establish standalone harness-bridge v0.4.0  -  by @rbutera (4db15)
• codex: Phase 2 app-server backend, behind CODEX_BRIDGE_BACKEND  -  by @rbutera (fe242)

   🐞 Bug Fixes

• Report the cli package version  -  by @rbutera (2b1ea)
• codex:
  • Drop ChatGPT-rejected *-codex models, raise+rename exec timeout, clearer timeout error  -  by @rbutera (5c0fc)
  • Per-key thread isolation, sandbox-matched resume, bounded concurrency  -  by @rbutera (c90a6)
  • Serialize calls sharing a threadKey  -  by @rbutera (838aa)
  • Dispose app-server threads to stop the MCP-stack leak; fast-fail wedged requests  -  by @rbutera in #3 (07978)

    View changes on GitHub