[mgarmash]

Problem

fetchImageEndpoint() in open-sse/handlers/imageGeneration.ts uses raw fetch() without any timeout control. When calling upstream image generation APIs (OpenAI, Nebius, etc.) that take 20-30+ seconds, Next.js self-hosted default timeout kills the connection, producing:

{"error":{"message":"terminated","type":"upstream_error","code":"upstream_error"}}

The chat/SSE endpoint already uses fetchWithTimeout() with FETCH_TIMEOUT_MS (default 600s) via open-sse/config/constants.ts, but image generation does not.

Fix

• Import fetchWithTimeout and getConfiguredTimeout from @/shared/utils/fetchTimeout (the same utility already used in the SSE code path)
• Replace raw fetch() with fetchWithTimeout() in fetchImageEndpoint(), using timeoutMs: getConfiguredTimeout()

Default timeout: 120s (OMNIROUTE_DEFAULT_FETCH_TIMEOUT_MS env var) or FETCH_TIMEOUT_MS env var.

Files changed

• open-sse/handlers/imageGeneration.ts — add import, replace fetch with fetchWithTimeout in fetchImageEndpoint

Testing

Reproduced issue with /v1/images/generations requests taking >30s. After fix, requests complete successfully within the configured timeout window.

[gemini-code-assist]

Code Review

This pull request integrates fetchWithTimeout and getConfiguredTimeout into the fetchImageEndpoint function within open-sse/handlers/imageGeneration.ts to enforce request timeouts. The review feedback recommends adding unit tests for this new behavior to comply with the repository style guide, and catching FetchTimeoutError or AbortError to return a 504 Gateway Timeout status code instead of a misleading 502 error.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

According to Rule 9 of the Repository Style Guide, we must always include tests when changing production code in open-sse/. Please add unit tests to verify the new timeout behavior of fetchImageEndpoint and ensure that fetchWithTimeout is correctly invoked with the configured timeout.

Suggested change

	import { fetchWithTimeout, getConfiguredTimeout } from "@/shared/utils/fetchTimeout";
	import { fetchWithTimeout, getConfiguredTimeout, FetchTimeoutError } from "@/shared/utils/fetchTimeout";

References

1. Always include tests when changing production code (src/, open-sse/, electron/, bin/). ^(link)

[gemini-code-assist]

When an upstream image generation request times out, returning a 502 (Bad Gateway) status code is misleading. To align with other timeout handlers in this file (which return 504), we should catch FetchTimeoutError or AbortError and return a 504 (Gateway Timeout) status code.

    let response;
    try {
      response = await fetchWithTimeout(url, {
        method: "POST",
        headers,
        body,
        timeoutMs: getConfiguredTimeout(),
      });
    } catch (err: any) {
      if (err instanceof FetchTimeoutError || err?.name === "AbortError") {
        if (log) {
          log.error("IMAGE", `${provider} fetch error: ${err.message}`);
        }
        return {
          success: false,
          status: 504,
          error: `Image provider error: ${sanitizeErrorMessage(err.message || err)}`,
        };
      }
      throw err;
    }

[kilo-code-bot]

Code Review Summary

Status: No Issues Found | Recommendation: Merge

Files Reviewed (1 file)

• open-sse/handlers/imageGeneration.ts

---

_{Reviewed by nemotron-3-super-120b-a12b-20230311:free · 372,991 tokens}

[diegosouzapw]

Thanks, @mgarmash! 🙏 Wrapping fetchImageEndpoint in fetchWithTimeout is the right call — image-gen was the one upstream path with no AbortController bound, so a stuck provider could hang until Next's server timeout. Verified the timer is cleared on both success and error, the error is sanitized (no stack leak), and a timeout correctly surfaces as a 504. The FETCH_TIMEOUT_MS env var governs the bound for operators running slower backends (ComfyUI/video). Merging into release/v3.8.9 for the next release.