docs: sync reasoning/thinking section across EN and ZH READMEs

docs: document prompt caching for both Anthropic and OpenAI providers

fix: drop max_tokens and passthrough usage details for OpenAI requests

feat: add prompt caching for OpenAI ChatGPT Backend requests

chore: translate .env.secrets.example comments to English

fix: drop temperature/top_p for OpenAI requests, document limitation

fix: always send instructions field to ChatGPT Backend

docs: update model list in README for Codex CLI models

feat: update OpenAI model list and routing for Codex CLI models

feat: add automatic prompt caching for Anthropic requests

feat: add thinking/reasoning_effort parameter passthrough

feat: add unit tests, OpenAI env fallback, smoke test, and doc improvements

feat: configurable model list + default model

chore: add CD workflow, secret management scripts, and gitignore

feat: add PROXY_API_KEY auth middleware + docs overhaul