feat: add Vision LLM integration (CLIP + Qwen3-VL cascade)

- Add Qwen3-VL dynamic management (start/stop/status CLI)
- Add CLIP + Qwen3-VL cascade detection strategy
- Add Vision CLI commands (vision start/stop/status, detect)
- Add cascade_vision processor module
- Add clip processor module
- Add qwen_vl_manager module

Changes:
- scripts/start_qwen3vl.sh, stop_qwen3vl.sh: Qwen3-VL management scripts
- src/core/vision/: Qwen3-VL manager module
- src/core/processor/cascade_vision.rs: CLIP + Qwen3-VL cascade logic
- src/core/processor/clip.rs: CLIP classification and detection
- src/api/clip_api.rs: CLIP API endpoints
- src/cli/vision.rs: Vision CLI implementation
- src/cli/args.rs: Add Vision and Detect commands
- src/main.rs: Integrate Vision CLI
- src/core/mod.rs: Add vision module
- src/core/processor/mod.rs: Add cascade_vision module

src/api/agent_api.rs

@@ -1,8 +1,8 @@
 use axum::{extract::State, http::StatusCode, response::Json, routing::post, Router};
-use reqwest::Client;
 use serde::{Deserialize, Serialize};
 
 use crate::api::types::AppState;
+use crate::core::llm::function_calling::LLM_CLIENT;
 
 pub fn agent_routes() -> Router<AppState> {
     Router::new().route("/api/v1/agents/translate", post(translate_text))
@@ -42,7 +42,6 @@ async fn translate_text(
     );
 
     // Call LLM via configurable endpoint
-    let client = Client::new();
     let llm_url = crate::core::config::llm::CHAT_URL.as_str();
     let model = crate::core::config::llm::CHAT_MODEL.as_str();
 
@@ -57,7 +56,7 @@ async fn translate_text(
         "temperature": 0.1
     });
 
-    let response = client.post(llm_url).json(&body).send().await.map_err(|e| {
+    let response = LLM_CLIENT.post(llm_url).json(&body).send().await.map_err(|e| {
         (
             StatusCode::INTERNAL_SERVER_ERROR,
             format!("Failed to call LLM: {}", e),