From 88170a6b12bdbb0a3c4bf83730dea21ffe952d45 Mon Sep 17 00:00:00 2001
From: Anthony Stirling <77850077+Frooodle@users.noreply.github.com>
Date: Wed, 2 Sep 2026 10:35:55 +0100
Subject: [PATCH] feat(editor): warn about small context windows on self-hosted
AI providers
---
.../public/locales/en-GB/translation.toml | 1 +
.../public/locales/en-US/translation.toml | 1 +
.../configSections/AdminAiModelsSection.tsx | 17 +++++++++++++++++
3 files changed, 19 insertions(+)
diff --git a/frontend/editor/public/locales/en-GB/translation.toml b/frontend/editor/public/locales/en-GB/translation.toml
index d2b2aa38e3..71510f202a 100644
--- a/frontend/editor/public/locales/en-GB/translation.toml
+++ b/frontend/editor/public/locales/en-GB/translation.toml
@@ -818,6 +818,7 @@ openai = "OpenAI API key"
setPlaceholder = "Saved - leave blank to keep the current key"
[admin.settings.ai.models.baseUrl]
+contextWindow = "Check the model's context window. The engine sends prompts of several thousand tokens and a window below 16,384 will silently truncate them - the oldest part of the prompt is dropped, so the model answers without having seen the request. Ollama defaults to 4,096 and divides that between parallel requests; raise it with OLLAMA_CONTEXT_LENGTH on the server."
description = "Base URL of the OpenAI-compatible / Ollama endpoint, e.g. http://ollama:11434/v1."
label = "Provider base URL"
warning = "The base URL must point at a trusted internal endpoint. The engine will make server-side requests to it, so an untrusted value is SSRF-sensitive."
diff --git a/frontend/editor/public/locales/en-US/translation.toml b/frontend/editor/public/locales/en-US/translation.toml
index 681bca4dd8..8a724a2a48 100644
--- a/frontend/editor/public/locales/en-US/translation.toml
+++ b/frontend/editor/public/locales/en-US/translation.toml
@@ -818,6 +818,7 @@ openai = "OpenAI API key"
setPlaceholder = "Saved - leave blank to keep the current key"
[admin.settings.ai.models.baseUrl]
+contextWindow = "Check the model's context window. The engine sends prompts of several thousand tokens and a window below 16,384 will silently truncate them - the oldest part of the prompt is dropped, so the model answers without having seen the request. Ollama defaults to 4,096 and divides that between parallel requests; raise it with OLLAMA_CONTEXT_LENGTH on the server."
description = "Base URL of the OpenAI-compatible / Ollama endpoint, e.g. http://ollama:11434/v1."
label = "Provider base URL"
warning = "The base URL must point at a trusted internal endpoint. The engine will make server-side requests to it, so an untrusted value is SSRF-sensitive."
diff --git a/frontend/editor/src/proprietary/components/shared/config/configSections/AdminAiModelsSection.tsx b/frontend/editor/src/proprietary/components/shared/config/configSections/AdminAiModelsSection.tsx
index acba1c9571..ce30b1df82 100644
--- a/frontend/editor/src/proprietary/components/shared/config/configSections/AdminAiModelsSection.tsx
+++ b/frontend/editor/src/proprietary/components/shared/config/configSections/AdminAiModelsSection.tsx
@@ -370,6 +370,23 @@ export default function AdminAiModelsSection() {
)}
+
+ {showBaseUrl && (
+
+ }
+ >
+
+ {t(
+ "admin.settings.ai.models.baseUrl.contextWindow",
+ "Check the model's context window. The engine sends prompts of several thousand tokens and a window below 16,384 will silently truncate them - the oldest part of the prompt is dropped, so the model answers without having seen the request. Ollama defaults to 4,096 and divides that between parallel requests; raise it with OLLAMA_CONTEXT_LENGTH on the server.",
+ )}
+
+
+ )}