Add first-class ChatGPT subscription provider support (#881)

* Add first-class ChatGPT subscription provider support

* Address PR feedback: move uuid import to top, reuse parsed config in up()

* Add ChatGPT token watchdog for seamless long-lived sessions

* Address PR feedback: error on stream=false for ChatGPT, fix auth file permissions

* Replace ChatGPT watchdog/restart with passthrough_auth

---------

Co-authored-by: Musa Malik <musam@uw.edu>

demos/llm_routing/chatgpt_subscription/README.md

@@ -0,0 +1,61 @@
+# ChatGPT Subscription Routing
+
+Route requests through your ChatGPT Plus/Pro subscription using Plano. Uses the OpenAI Responses API under the hood, targeting `chatgpt.com/backend-api/codex/responses`.
+
+## Setup
+
+### 1. Authenticate with ChatGPT
+
+```bash
+planoai chatgpt login
+```
+
+This opens a device code flow — visit the URL shown and enter the code. Tokens are saved to `~/.plano/chatgpt/auth.json`.
+
+### 2. Start Plano
+
+```bash
+planoai up config.yaml
+```
+
+### 3. Send a request
+
+```bash
+curl http://localhost:12000/v1/responses \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "gpt-5.2",
+    "input": "Hello, what model are you?"
+  }'
+```
+
+Or use the test script:
+
+```bash
+bash test_chatgpt.sh
+```
+
+## How it works
+
+- `chatgpt/gpt-5.2` in the config tells Plano to use the ChatGPT subscription provider
+- Plano reads OAuth tokens from `~/.plano/chatgpt/auth.json` (auto-refreshes if expired)
+- Requests are proxied to `https://chatgpt.com/backend-api/codex/responses` with the required headers:
+  - `Authorization: Bearer <access_token>`
+  - `ChatGPT-Account-Id: <account_id>`
+  - `originator: codex_cli_rs`
+  - `session_id: <uuid>`
+
+## Available models
+
+```
+chatgpt/gpt-5.4
+chatgpt/gpt-5.3-codex
+chatgpt/gpt-5.2
+```
+
+## Managing credentials
+
+```bash
+planoai chatgpt status   # Check auth status
+planoai chatgpt logout   # Remove stored credentials
+```

demos/llm_routing/chatgpt_subscription/chat.py

@@ -0,0 +1,38 @@
+#!/usr/bin/env python3
+"""Interactive chat with a model through Plano using the OpenAI SDK."""
+
+import sys
+from openai import OpenAI
+
+client = OpenAI(base_url="http://localhost:12000/v1", api_key="unused")
+
+
+def run_chat(model):
+    print(f"Chatting with {model} via Plano (Ctrl+C to quit)\n")
+    history = []
+    while True:
+        try:
+            user_input = input("you> ")
+        except (KeyboardInterrupt, EOFError):
+            print("\nbye")
+            break
+        if not user_input.strip():
+            continue
+
+        history.append({"role": "user", "content": user_input})
+
+        stream = client.responses.create(model=model, input=history, stream=True)
+        print(f"{model}> ", end="", flush=True)
+        full = ""
+        for event in stream:
+            if event.type == "response.output_text.delta":
+                print(event.delta, end="", flush=True)
+                full += event.delta
+        print()
+
+        history.append({"role": "assistant", "content": full})
+
+
+if __name__ == "__main__":
+    model = sys.argv[1] if len(sys.argv) > 1 else "gpt-5.2"
+    run_chat(model)

demos/llm_routing/chatgpt_subscription/config.yaml

@@ -0,0 +1,9 @@
+version: v0.3.0
+
+listeners:
+  - type: model
+    name: model_listener
+    port: 12000
+
+model_providers:
+  - model: chatgpt/*

demos/llm_routing/chatgpt_subscription/test_chatgpt.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+# Test ChatGPT subscription routing through Plano
+# Prerequisites: planoai chatgpt login && planoai up config.yaml
+
+set -e
+
+echo "Testing ChatGPT subscription via Plano Responses API..."
+echo ""
+
+curl -s http://localhost:12000/v1/responses \
+  -H "Content-Type: application/json" \
+  -d '{
+    "model": "gpt-5.2",
+    "input": "What is 2 + 2? Reply in one word."
+  }' | python3 -m json.tool
+
+echo ""
+echo "Done."