[Bug] Z.ai API persistent "Connecting" hang in ZCode 3.1.4 — entitlement endpoint fails, model returns HTTP 529

提交前确认 · Pre-submission checklist

• 我已搜索过现有 issue，确认这不是重复 / I searched existing issues and confirmed this isn't a duplicate.
• 我已阅读 CONTRIBUTING.md / I've read CONTRIBUTING.md.

问题类别 · Category

UI / 界面体验 · UI / UX

涉及的 Agent 框架 · Agent framework

ZCode Agent（自研）

严重程度 · Severity

阻塞使用 · Blocking (无法使用核心功能 / core function unusable)

复现频率 · Reproducibility

必现 · Always

问题描述 · Description

Problem: ZCode desktop 3.1.4 stays stuck on "Connecting" when using Z.ai models directly (api.z.ai).

• GET /api/monitor/usage/quota/limit fails repeatedly (11–22824 ms timeouts)
• POST /api/anthropic/v1/messages returns HTTP 529 (overloaded_error) intermittently
• Same models work fine when routed through local NVIDIA NIM proxy (127.0.0.1:8082)

Impact: Daily occurrence (189 failures on 2026-06-22, 76 on 2026-06-23). ZCode unusable with Z.ai without third-party proxy.

Evidence: Logs attached — entitlement check is the primary blocker. Direct curl from same machine succeeds (200/0.83s), so server is reachable.

Requested fixes:

• Z.ai: stabilize /api/monitor/usage/quota/limit, add Retry-After header for 529
• ZCode: don't block session startup on entitlement failure, add retry with exponential backoff, show actionable error instead of infinite spinner

Full details below ↓

zai-connecting-issue.md

复现步骤 · Steps to reproduce

up

期望表现 · Expected behavior

up

实际表现 · Actual behavior

up

ZCode 版本 · ZCode version

3.1.4

设备 / 系统 / 浏览器 · Device / OS / Browser

截图 / 录屏 / 日志 · Screenshots / Recordings / Logs

No response

Start with the attached zai-connecting-issue.md and its logs, then verify the reported quota and messages endpoints against the ZCode 3.1.4 and local NVIDIA NIM proxy details. The issue spans Z.ai and ZCode, so confirm which project owns each requested change; done requires an agreed scope and a reproducible resolution for the persistent Connecting state.