Pull requests / #759
#759 Add native Responses API support for Codex clients
closed · @Averyyy · 0 comments · View on GitHub
Setup & installServer & APISecurityWindows
Description
Strata's server supports Chat Completions and Anthropic Messages, but clients such as Codex require the Responses API. Add `POST /v1/responses` so those clients can use the existing model and engine directly. The endpoint supports streaming and non-streaming text, reasoning effort, function and custom tools, explicit tool namespaces, and full-history replay of tool calls and results. It reuses the existing authentication, cancellation, request monitoring, and output parser. The client owns conversation history; stored responses and structured output are outside this endpoint's supported scope. Validation: 125 Responses and server tests, plus 28 monitoring and security tests, passed. Protocol tests cover streaming lifecycle, text and reasoning, output limits, authentication, function/custom tool replay, and namespaced calls. On Windows with Codex CLI 0.160.0 and Qwen3.8 Flash Next, a real streaming turn completed an `exec_command` call, replayed its output, and returned the requested result. The initial prompt contained 92,672 tokens from the installed tools and skills. A separate local coding profile completed shell execution and a native `spawn_agent` / `wait_agent` / `close_agent` workflow; the Qwen child returned the requested marker.
Related on strata.com
Editorial links to help you install, pick models, or read release notes — not part of the upstream thread.