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).
This commit is contained in:
Waleed
2026-06-16 10:27:07 -07:00
committed by GitHub
parent 2c1392e3a0
commit 2b7f57a788
+2 -2
View File
@@ -195,7 +195,7 @@ export const PROVIDER_DEFINITIONS: Record<string, ProviderDefinition> = {
},
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<string, ProviderDefinition> = {
},
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',