(feat) add reasoning model support in openai provider

src/providers/openai/convert.rs

@@ -34,16 +34,27 @@ pub fn to_openai_request(req: &GenerateRequest, stream: bool) -> ChatCompletionR
         }),
     });
 
+    let mut temperature = req.options.temperature;
+    let mut top_p = req.options.top_p;
+    let mut reasoning_effort = None;
+
+    if OPENAI_REASONING_MODELS.contains(&req.model.as_str()) {
+        temperature = None;
+        top_p = None;
+        reasoning_effort = Some(OpenAIReasoningEffort::Medium);
+    }
+
     ChatCompletionRequest {
         model: req.model.clone(),
         messages: req.messages.iter().map(to_openai_message).collect(),
-        temperature: req.options.temperature,
-        max_tokens: req.options.max_tokens,
-        top_p: req.options.top_p,
+        temperature,
+        max_completion_tokens: req.options.max_tokens,
+        top_p,
         stop: req.options.stop_sequences.clone(),
         stream: Some(stream),
         tools,
         tool_choice,
+        reasoning_effort,
     }
 }
 

src/providers/openai/provider.rs

@@ -79,7 +79,12 @@ impl Provider for OpenAIProvider {
             )));
         }
 
-        let openai_resp: ChatCompletionResponse = response.json().await?;
+        let response_text = response.text().await?;
+        println!("Response body: {}", response_text);
+
+        let openai_resp: ChatCompletionResponse = serde_json::from_str(&response_text)?;
+        println!("Parsed response: {:#?}", openai_resp);
+
         from_openai_response(openai_resp)
     }
 

src/providers/openai/types.rs

@@ -2,6 +2,16 @@
 
 use serde::{Deserialize, Serialize};
 
+pub static OPENAI_REASONING_MODELS: &[&str] = &[
+    "o1-2024-12-17",
+    "o3-2025-04-16",
+    "o3-mini-2025-01-31",
+    "o4-mini-2025-04-16",
+    "gpt-5-2025-08-07",
+    "gpt-5-mini-2025-08-07",
+    "gpt-5-nano-2025-08-07",
+];
+
 /// Configuration for OpenAI provider
 #[derive(Debug, Clone)]
 pub struct OpenAIConfig {
@@ -48,9 +58,7 @@ pub struct ChatCompletionRequest {
     pub model: String,
     pub messages: Vec<ChatMessage>,
     #[serde(skip_serializing_if = "Option::is_none")]
-    pub temperature: Option<f32>,
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub max_tokens: Option<u32>,
+    pub max_completion_tokens: Option<u32>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub top_p: Option<f32>,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -61,6 +69,23 @@ pub struct ChatCompletionRequest {
     pub tools: Option<Vec<serde_json::Value>>,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub tool_choice: Option<serde_json::Value>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub temperature: Option<f32>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub reasoning_effort: Option<OpenAIReasoningEffort>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default)]
+pub enum OpenAIReasoningEffort {
+    #[serde(rename = "minimal")]
+    Minimal,
+    #[serde(rename = "low")]
+    Low,
+    #[default]
+    #[serde(rename = "medium")]
+    Medium,
+    #[serde(rename = "high")]
+    High,
 }
 
 /// OpenAI chat message

tests/integration/openai.rs

@@ -11,18 +11,20 @@ async fn test_openai_generate() {
     let client = Inference::new();
 
     let mut request = GenerateRequest::new(
-        "gpt-3.5-turbo",
+        "gpt-5-mini-2025-08-07",
         vec![Message {
             role: Role::User,
             content: "Say 'Hello, World!' and nothing else".into(),
             name: None,
         }],
     );
     request.options.temperature = Some(0.0);
-    request.options.max_tokens = Some(10);
+    request.options.max_tokens = Some(5000);
 
     let response = client.generate(&request).await;
 
+    println!("Response: {:#?}", response);
+
     assert!(response.is_ok(), "Request failed: {:?}", response.err());
     let response = response.unwrap();
 
@@ -36,15 +38,15 @@ async fn test_openai_streaming() {
     let client = Inference::new();
 
     let mut request = GenerateRequest::new(
-        "gpt-3.5-turbo",
+        "gpt-5-nano-2025-08-07",
         vec![Message {
             role: Role::User,
             content: "Count from 1 to 3".into(),
             name: None,
         }],
     );
     request.options.temperature = Some(0.0);
-    request.options.max_tokens = Some(20);
+    request.options.max_tokens = Some(5000);
 
     let stream = client.stream(&request).await;
     assert!(stream.is_ok(), "Stream creation failed: {:?}", stream.err());