feat: strong typing for complete endpoints

src/dto/api.rs

@@ -3,21 +3,6 @@ use utoipa::ToSchema;
 
 use crate::errors::OllamaError;
 
-#[derive(Serialize, Deserialize, ToSchema)]
-pub struct ChatRequest {
-    pub model: String,
-    pub prompt: Option<String>,
-    pub messages: Option<Vec<Message>>,
-    #[serde(default)]
-    pub stream: bool,
-    pub temperature: Option<f32>,
-    pub top_p: Option<f32>,
-    pub max_tokens: Option<u32>,
-    pub stop: Option<Vec<String>>,
-    pub system: Option<String>,
-    pub keep_alive: Option<String>,
-}
-
 #[derive(Debug, Serialize, Deserialize, ToSchema)]
 #[serde(rename_all = "lowercase")]
 pub enum Role {
@@ -37,6 +22,8 @@ pub struct Message {
     pub content: String,
 }
 
+// ---------------------------
+
 #[derive(Debug, Serialize, Deserialize, ToSchema)]
 pub struct ModelsResponse {
     pub models: Vec<ModelInfo>,
@@ -91,3 +78,80 @@ pub struct UnloadModelResponse {
     pub model: String,
     pub status: String,
 }
+
+#[derive(Debug, Deserialize, Serialize, ToSchema)]
+pub struct BaseLLMRequest {
+    pub model: String,
+
+    #[serde(default)]
+    pub stream: bool,
+
+    pub temperature: Option<f32>,
+    pub top_p: Option<f32>,
+
+    // Ollama-native
+    pub top_k: Option<u32>,
+    pub repeat_penalty: Option<f32>,
+    pub seed: Option<i64>,
+
+    pub num_ctx: Option<u32>,
+    pub num_predict: Option<u32>,
+
+    pub stop: Option<Vec<String>>,
+
+    pub keep_alive: Option<String>,
+}
+
+#[derive(Debug, Deserialize, Serialize, ToSchema)]
+pub struct CompletionRequest {
+    #[serde(flatten)]
+    pub base: BaseLLMRequest,
+
+    pub prompt: String,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum CompletionObject {
+    TextCompletion,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+#[serde(rename_all = "snake_case")]
+pub enum FinishReason {
+    Stop,
+    Length,
+    ContentFilter,
+    ToolCalls,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+pub struct CompletionResponse {
+    pub id: String,
+    pub object: CompletionObject,
+    pub created: u64,
+    pub model: String,
+    pub choices: Vec<Choice>,
+    pub usage: Usage,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+pub struct Choice {
+    pub text: String,
+    pub index: u32,
+    pub finish_reason: FinishReason,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+pub struct Usage {
+    pub prompt_tokens: u32,
+    pub completion_tokens: u32,
+    pub total_tokens: u32,
+}
+
+#[derive(Debug, Serialize, Deserialize, ToSchema)]
+pub struct CompletionChunk {
+    pub id: String,
+    pub object: String,
+    pub choices: Vec<Choice>,
+}