From 2b7f57a7884fde1b13bb57c8d81e6d7039dc6d41 Mon Sep 17 00:00:00 2001 From: Waleed Date: Tue, 16 Jun 2026 10:27:07 -0700 Subject: [PATCH] improvement(providers): tighten Gemini and vLLM agent-attachment ceilings (#5095) A live-doc audit of the merged large-file feature found two ceilings that were higher than the provider actually accepts: - Gemini: 100MB -> 50MB. Gemini hard-caps PDFs at 50MB, so a 50-100MB PDF passed our gate, got uploaded + polled, then failed at generateContent. 50MB respects the documented limit and is more memory-safe. - vLLM: 50MB -> 25MB. vLLM's default image-fetch timeout is 5s; a 50MB remote fetch routinely exceeds it. 25MB aligns with that reality and matches Baseten (the other vLLM-backed provider). --- apps/sim/providers/models.ts | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/apps/sim/providers/models.ts b/apps/sim/providers/models.ts index 9381b7317e..a827c32fe8 100644 --- a/apps/sim/providers/models.ts +++ b/apps/sim/providers/models.ts @@ -195,7 +195,7 @@ export const PROVIDER_DEFINITIONS: Record = { }, vllm: { id: 'vllm', - fileAttachment: { maxBytes: 50 * 1024 * 1024, strategy: 'remote-url' }, + fileAttachment: { maxBytes: 25 * 1024 * 1024, strategy: 'remote-url' }, name: 'vLLM', icon: VllmIcon, description: 'Self-hosted vLLM with an OpenAI-compatible API', @@ -1319,7 +1319,7 @@ export const PROVIDER_DEFINITIONS: Record = { }, google: { id: 'google', - fileAttachment: { maxBytes: 100 * 1024 * 1024, strategy: 'files-api' }, + fileAttachment: { maxBytes: 50 * 1024 * 1024, strategy: 'files-api' }, name: 'Google', description: "Google's Gemini models", defaultModel: 'gemini-2.5-pro',