language_models: Add thinking support to LM Studio provider (#32337)

Umesh Yadav created 6 months ago

It works similar to how deepseek works where the thinking is returned as
reasoning_content and we don't have to send the reasoning_content back
in the request.

This is a experiment feature which can be enabled from settings like
this:
<img width="1381" alt="Screenshot 2025-06-08 at 4 26 06 PM"
src="https://github.com/user-attachments/assets/d2f60f3c-0f93-45fc-bae2-4ded42981820"
/>

Here is how it looks to use(tested with
`deepseek/deepseek-r1-0528-qwen3-8b`

<img width="528" alt="Screenshot 2025-06-08 at 5 12 33 PM"
src="https://github.com/user-attachments/assets/f7716f52-5417-4f14-82b8-e853de054f63"
/>


Release Notes:

- Add thinking support to LM Studio provider

Change summary

crates/language_models/src/provider/lmstudio.rs | 25 ++++++++++++------
crates/lmstudio/src/lmstudio.rs                 |  2 +
2 files changed, 18 insertions(+), 9 deletions(-)

Detailed changes

crates/language_models/src/provider/lmstudio.rs

@@ -250,15 +250,15 @@ impl LmStudioLanguageModel {
         for message in request.messages {
             for content in message.content {
                 match content {
-                    MessageContent::Text(text) | MessageContent::Thinking { text, .. } => messages
-                        .push(match message.role {
-                            Role::User => ChatMessage::User { content: text },
-                            Role::Assistant => ChatMessage::Assistant {
-                                content: Some(text),
-                                tool_calls: Vec::new(),
-                            },
-                            Role::System => ChatMessage::System { content: text },
-                        }),
+                    MessageContent::Text(text) => messages.push(match message.role {
+                        Role::User => ChatMessage::User { content: text },
+                        Role::Assistant => ChatMessage::Assistant {
+                            content: Some(text),
+                            tool_calls: Vec::new(),
+                        },
+                        Role::System => ChatMessage::System { content: text },
+                    }),
+                    MessageContent::Thinking { .. } => {}
                     MessageContent::RedactedThinking(_) => {}
                     MessageContent::Image(_) => {}
                     MessageContent::ToolUse(tool_use) => {
@@ -471,6 +471,13 @@ impl LmStudioEventMapper {
             events.push(Ok(LanguageModelCompletionEvent::Text(content)));
         }
 
+        if let Some(reasoning_content) = choice.delta.reasoning_content {
+            events.push(Ok(LanguageModelCompletionEvent::Thinking {
+                text: reasoning_content,
+                signature: None,
+            }));
+        }
+
         if let Some(tool_calls) = choice.delta.tool_calls {
             for tool_call in tool_calls {
                 let entry = self.tool_calls_by_index.entry(tool_call.index).or_default();

crates/lmstudio/src/lmstudio.rs

@@ -277,6 +277,8 @@ pub struct ResponseMessageDelta {
     pub role: Option<Role>,
     pub content: Option<String>,
     #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub reasoning_content: Option<String>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
     pub tool_calls: Option<Vec<ToolCallChunk>>,
 }