From 36278ffc7d882d5f902f3959ce3ec0294b563d66 Mon Sep 17 00:00:00 2001
From: Fsocietyhhh <1211904451@qq.com>
Date: Mon, 27 Apr 2026 23:39:17 -0700
Subject: [PATCH] fix(imagegen): bump image-to-image timeout to 180s (was 60s
 shared with t2i)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The single 60s timeout on the main /v1/images/{generations,image2image}
fetch covered both x402 retry attempts AND the actual generation. That
budget is generous for text-to-image (gpt-image-1 finishes in 10–30s)
but consistently insufficient for image-to-image:

- gpt-image-2 is reasoning-driven; an edit easily takes 60–150s on its
  own
- the request body carries a few MB of base64 reference image, adding
  measurable upload time on slower links
- the timeout has to share its budget across two POSTs (the 402 round
  trip plus the paid one)

End result: every gpt-image-2 reference-image call I tried hit AbortError
before the upstream model returned, with a misleading message ("Try a
simpler prompt") that made it look like a server problem.

This patch:
- splits the budget — image-to-image gets 180s, text-to-image keeps 60s
- updates the timeout error message to mention the 180s limit and the
  most likely causes (image too large, model under load) when the
  reference-image path is taken

The existing 30s timeouts on resolveReferenceImage (URL fetch) and on
the result download stay unchanged — those are network-bound and
shouldn't routinely be slow.
---
 src/tools/imagegen.ts | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)

diff --git a/src/tools/imagegen.ts b/src/tools/imagegen.ts
index cf04bac..c1353ab 100644
--- a/src/tools/imagegen.ts
+++ b/src/tools/imagegen.ts
@@ -283,7 +283,14 @@ function buildExecute(deps: ImageGenDeps) {
   };
 
   const controller = new AbortController();
-  const timeout = setTimeout(() => controller.abort(), 60_000); // 60s timeout
+  // Reference-image mode (gpt-image-2 edits) is meaningfully slower than
+  // pure text-to-image: the model is reasoning-driven and the request
+  // body carries a few MB of base64. The shared 60s budget has to cover
+  // both x402 retry attempts plus the actual generation, which made
+  // image-to-image effectively always time out. Image-to-image gets 3
+  // minutes; text-to-image keeps the original 60s.
+  const timeoutMs = referenceImage ? 180_000 : 60_000;
+  const timeout = setTimeout(() => controller.abort(), timeoutMs);
 
   try {
     // First request — will get 402
@@ -404,7 +411,12 @@ function buildExecute(deps: ImageGenDeps) {
   } catch (err) {
     const msg = (err as Error).message || '';
     if (msg.includes('abort')) {
-      return { output: 'Image generation timed out (60s limit). Try a simpler prompt.', isError: true };
+      return {
+        output: referenceImage
+          ? 'Image-to-image timed out (180s limit). The reference image may be too large or the model under load — try a smaller image or simpler prompt.'
+          : 'Image generation timed out (60s limit). Try a simpler prompt.',
+        isError: true,
+      };
     }
     return { output: `Error: ${msg}`, isError: true };
   } finally {