Skip to content

Build your personal knowledge base with Trilium Notes

fix(llm): show a streaming thought in full instead of its last four lines

Elian Doran8505d00cbed0
diff --git a/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.css b/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.css
index c4209ff30f..59f6fa7252 100644
--- a/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.css
+++ b/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.css
@@ -239,19 +239,6 @@ body.background-effects.theme-supports-background-effects .llm-chat-message-user
white-space: nowrap;
}
-/* Aligning the content to the end clips the overflow at the top, keeping the newest lines in view. */
-.llm-chat-thinking-live .llm-chat-thinking-content {
- display: flex;
- flex-direction: column;
- justify-content: flex-end;
- max-height: calc(4 * 1.6em);
- overflow: hidden;
-
- > * {
- flex-shrink: 0;
- }
-}
-
/* The text after the last thought or tool call, set apart from the steps that led to it. */
.llm-chat-reply {
margin-top: 1rem;
diff --git a/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.tsx b/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.tsx
index 712be9e412..964d5ab895 100644
--- a/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.tsx
+++ b/apps/client/src/widgets/type_widgets/llm_chat/ChatMessage.tsx
@@ -126,8 +126,7 @@ function CitationsSection({ citations }: { citations: LlmCitation[] }) {
/**
* One stretch of the model's reasoning. A finished one folds to a single muted line: its leading
* `**Title**` (the shape of Codex's reasoning summaries), or else its first line, which the body
- * then leaves out. The one being generated stays open, clamped to its last lines under a spinner
- * and its latest title.
+ * then leaves out. The one being generated stays open in full under a spinner and its latest title.
*/
function ThinkingCard({ content, isLive }: { content: string; isLive?: boolean }) {
if (isLive) {
diff --git a/apps/server/src/assets/doc_notes/en/User Guide/User Guide/AI.html b/apps/server/src/assets/doc_notes/en/User Guide/User Guide/AI.html
index 6e9fa9d75e..8a67175809 100644
--- a/apps/server/src/assets/doc_notes/en/User Guide/User Guide/AI.html
+++ b/apps/server/src/assets/doc_notes/en/User Guide/User Guide/AI.html
@@ -120,8 +120,9 @@ class="admonition note">
search providers such as Exa, Tavily &amp; SearXNG are not yet supported.</p>
</aside>
<h3>Thinking</h3>
-<p>Some models reason before they answer. Their reasoning appears above the
- reply as a collapsible <em>Thought process</em> line.</p>
+<p>Some models reason before they answer. While the model is thinking, its
+ reasoning is shown in full under a spinner; once it is done, it folds into
+ a collapsible <em>Thought process</em> line above the reply.</p>
<p>How much a model thinks is set per chat, in one of two ways depending
on the model:</p>
<ul>
diff --git a/docs/User Guide/User Guide/AI.md b/docs/User Guide/User Guide/AI.md
index 7709b2b567..25f4057841 100644
--- a/docs/User Guide/User Guide/AI.md
+++ b/docs/User Guide/User Guide/AI.md
@@ -92,7 +92,7 @@ This feature is on by default but it can easily be disabled by clicking on the m
### Thinking
-Some models reason before they answer. Their reasoning appears above the reply as a collapsible _Thought process_ line.
+Some models reason before they answer. While the model is thinking, its reasoning is shown in full under a spinner; once it is done, it folds into a collapsible _Thought process_ line above the reply.
How much a model thinks is set per chat, in one of two ways depending on the model: