feat: add unload endpoint

src/providers/ollama.rs

@@ -107,6 +107,31 @@ impl OllamaProvider {
         }))
     }
 
+    pub async fn unload_model(&self, model: &str) -> Result<Value, OllamaError> {
+        self.validate_model(model).await?;
+
+        let payload = json!({
+            "model":      model,
+            "prompt":     "",
+            "keep_alive": "0",
+            "stream":     false,
+        });
+
+        let res = self
+            .client
+            .post(format!("{}/api/generate", self.base_url))
+            .json(&payload)
+            .send()
+            .await?
+            .json::<Value>()
+            .await?;
+
+        Ok(json!({
+            "model":  res.get("model"),
+            "status": "unloaded",
+        }))
+    }
+
     pub async fn completions(&self, body: Value) -> Result<Value, OllamaError> {
         let prompt = body
             .get("prompt")

src/routes/v1/mod.rs

@@ -11,5 +11,6 @@ pub fn router() -> Router<AppState> {
         .route("/completions", post(chat::completions))
         .route("/chat/completions", post(chat::chat_completions))
         .route("/models/{model}/load", post(models::load_model))
+        .route("/models/{model}/unload", post(models::unload_model))
         .layer(middleware::from_fn(auth_middleware))
 }

src/routes/v1/models.rs

@@ -36,6 +36,17 @@ pub async fn load_model(
     }
 }
 
+pub async fn unload_model(
+    State(state): State<AppState>,
+    Path(model): Path<String>,
+) -> Result<Json<Value>, (axum::http::StatusCode, String)> {
+    match state.ollama.unload_model(&model).await {
+        // ← correct method
+        Ok(response) => Ok(Json(response)),
+        Err(e) => Err(ollama_err(e)),
+    }
+}
+
 fn ollama_err(e: OllamaError) -> (axum::http::StatusCode, String) {
     match e {
         OllamaError::ModelNotFound(m) => (