404 Not Found: CC Switch proxy failed /responses - Codex
Quick fix
DeepSeek does not support the Responses API — use the built-in preset to route through Chat Completions. For image errors, drop the image modality.
Symptom
Section titled “Symptom”unexpected status 404 Not Found: CC Switch local proxy failed while handling Codex endpoint /responses. Provider: DeepSeek; model: deepseek-v4-pro; upstream_status: HTTP 404; cause: 上游错误 (404):, url: http://127.0.0.1:15721/v1/responsesunexpected status 404 Not Found: CC Switch local proxy failed while handling Codex endpoint /responses. Provider: DeepSeek; model: deepseek-v4-pro; upstream_status: HTTP 404; cause: upstream error (404):, url: http://127.0.0.1:15721/v1/responses
This cluster holds several HTTP 404 variants. The most common: the official DeepSeek endpoint does not support the Codex `/responses` endpoint, so requesting it directly returns 404. The others: `No endpoints found that support image input` means DeepSeek is a text-only model but the model catalogue config wrongly includes the `image` modality, so the rectifier never strips the image; a 404 on `/v1/models` with an empty model list is a Codex desktop restriction requiring the official login state to be preserved; a 404 from a third-party gateway such as OpenCode Go needs the `/v1` path filled in by hand; and a 404 on a MiMo usage query is simply the vendor not exposing that API.
For a DeepSeek `/responses` 404, do not enter the official API address directly — use the built-in CC Switch DeepSeek preset so it routes through Chat Completions takeover.
For a `No endpoints found that support image input` 404 caused by image input, correct the model catalogue config and remove the unsupported image modality.
// ~/.codex/cc-switch-model-catalog.json{"input_modalities": ["text"]}For a `/v1/models` 404 leaving the model list empty, sign in to an official ChatGPT account in Codex desktop first, turn on 切换第三方时保留官方登录 (keep the official login when switching to a third party) in CC Switch, and restart Codex.
For a 404 from a third-party gateway such as OpenCode Go, add the `/v1` suffix by hand in the CC Switch endpoint config (`https://opencode.ai/zen/go/v1`, for example).
Affected Versions
Section titled “Affected Versions”Source Issues
Section titled “Source Issues”This page is distilled from 5 real issues
- Why does Codex desktop show no third-party models after switching to DeepSeek, with `/v1/models` returning 404?
- That is a Codex desktop restriction: it forcibly falls back without an official account login. Sign in to an official ChatGPT account once, turn on keeping the official login in CC Switch, and restart to restore the display.
- Using OpenRouter for DeepSeek, every request after sending an image returns 404. How do I fix it?
- The model's `input_modalities` in `cc-switch-model-catalog.json` wrongly includes `image`, so the rectifier never strips the image. Change that field to `["text"]` by hand.
- Configuring a custom gateway such as OpenCode Go, requesting `/chat/completions` returns 404. What now?
- CC Switch does not append `/v1` automatically for every custom gateway. Enter the full API root when configuring the endpoint, for example `https://opencode.ai/zen/go/v1`.
- After configuring Xiaomi MiMo, clicking the usage query returns 404. What now?
- MiMo has not opened a balance query API yet, so the query path CC Switch generates by default returns 404. Ignore the error, or disable the usage query in the config.
Related problems
Section titled “Related problems”This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.