Mark project as paused

README.md

@@ -4,12 +4,25 @@
 
 Codex App auth-split proxy for third-party OpenAI-compatible APIs.
 
+## Project Status
+
+This project is paused and is not recommended for new installs right now.
+
+It remains available as a reference implementation for Codex App auth-split behavior,
+provider-auth isolation, service-tier handling, redacted diagnostics, and safe restore/uninstall
+flows for providers that already support OpenAI's Responses API.
+
+Most simple "keep ChatGPT login while using a third-party provider" workflows are now covered by
+general provider switchers. The main reason to resume this project would be a robust
+Responses API to Chat Completions adapter for providers such as DeepSeek, Kimi, GLM, MiniMax, and
+similar APIs. See [Roadmap](docs/roadmap.md).
+
 Use Codex App while you sign in with ChatGPT for the full App UI, but keep model requests on your
 third-party OpenAI-compatible API provider. `codex-fast-proxy` routes provider traffic through a
 local proxy, applies an optional provider-auth override, preserves streaming, and keeps the App's
 own Fast controls intact when they are available.
 
-[Chinese Guide](docs/README.zh-CN.md) · [Quick Start](#quick-start) · [Common Workflows](#common-workflows) · [Dashboard](#dashboard) · [Safety](#safety) · [Advanced Usage](docs/advanced-usage.md) · [Sponsor](#sponsor)
+[Chinese Guide](docs/README.zh-CN.md) · [Roadmap](docs/roadmap.md) · [Advanced Usage](docs/advanced-usage.md) · [Sponsor](#sponsor)
 
 ![Codex Fast Proxy overview](docs/assets/codex-fast-proxy-promo.gif)
 
@@ -54,7 +67,10 @@ measured workload got faster, and whether provider response metadata explicitly
 The benchmark also records whether the control split was valid, so default samples must omit
 `service_tier` while priority samples send the expected value.
 
-## Quick Start
+## Historical Install
+
+New installs are not recommended while the project is paused. Existing users or maintainers who
+explicitly need this reference implementation can still use the historical install flow.
 
 Paste this into Codex:
 

docs/README.zh-CN.md

@@ -2,11 +2,21 @@
 
 `codex-fast-proxy` 面向使用兼容 OpenAI API 的第三方供应商的 Codex App 用户。核心用途是让 Codex App 可以保持 ChatGPT 账户登录，继续使用插件市场、GitHub/Apps/connectors、Fast 手动选择、状态提示和语音输入等 UI 能力，同时把模型请求转到第三方上游。
 
+## 项目状态
+
+项目已暂停，目前不建议新用户安装。
+
+它仍然可以作为 Codex App auth split、provider auth 隔离、service tier 处理、脱敏诊断和安全恢复/卸载流程的参考实现。简单的“保留 ChatGPT 登录，同时切换第三方供应商”需求，现在已经有通用切换器覆盖了一大部分。
+
+如果后续继续做，最有价值的方向是把 Codex Responses API 稳定转换到 DeepSeek、Kimi、GLM、MiniMax 等 Chat Completions API。详见 [Roadmap](roadmap.md)。
+
 Fast/Priority 是重要能力，但实际是否加速取决于上游 API 提供商是否支持；请以 A/B benchmark 的结果为准。
 
 [返回英文 README](../README.md)
 
-## 快速开始
+## 历史安装方式
+
+项目暂停期间不建议新用户安装。已有用户或维护者如果明确需要这个参考实现，可以继续使用历史安装流程。
 
 把这句话贴给 Codex：
 

docs/roadmap.md

@@ -0,0 +1,49 @@
+# Roadmap
+
+## Current Status
+
+This project is paused. No stable public release or main-branch merge is planned while the product
+scope is unclear.
+
+The current implementation is useful as a reference for:
+
+- Keeping Codex App signed in with ChatGPT while model requests use a third-party provider.
+- Splitting provider auth into a proxy-managed local file.
+- Handling `service_tier` conservatively for Responses API traffic.
+- Redacted diagnostics, request metadata, benchmark records, update, restore, and uninstall flows.
+
+## Not In Scope Today
+
+This project does not currently translate Codex Responses API traffic into Chat Completions traffic.
+That means providers that only expose Chat Completions-style APIs, such as common DeepSeek, Kimi,
+GLM, MiniMax, and similar configurations, are not the primary target unless an upstream proxy already
+offers a Responses-compatible route.
+
+## Resume Criteria
+
+The project becomes worth resuming if it adds a reliable, provider-agnostic adapter:
+
+```text
+Codex Responses API
+        |
+        v
+local adapter
+        |
+        v
+provider Chat Completions API
+```
+
+The adapter should be data-driven rather than a pile of provider-specific branches. It would need to
+preserve streaming behavior, tool calls, errors, usage metadata, model selection, reasoning-related
+fields where possible, and Codex's expectations for incremental Responses events.
+
+## Maintenance Policy
+
+Until then, keep changes small:
+
+- Security fixes.
+- Broken uninstall or restore fixes.
+- Documentation updates that prevent user confusion.
+- Compatibility notes for existing users.
+
+Avoid broad feature work unless it directly supports the Responses-to-Chat-Completions adapter.