fix: Enable vector_db by default with health check and fallback to local LLM

- Add vector_db_health_check() function to verify Qdrant availability
- Add wait loop for vector_db startup in bootstrap (15 seconds)
- Fallback to local LLM when external URL configured but no API key provided
- Prevent external LLM (api.z.ai) usage without authentication key

This fixes the production issues:
- Qdrant vector database not available at https://localhost:6333
- External LLM being used instead of local when no key is configured
- Ensures vector_db is properly started and ready before use

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

src/core/bootstrap/bootstrap_manager.rs

@@ -1,6 +1,6 @@
 // Bootstrap manager implementation
 use crate::core::bootstrap::bootstrap_types::{BootstrapManager, BootstrapProgress};
-use crate::core::bootstrap::bootstrap_utils::{cache_health_check, safe_pkill, vault_health_check};
+use crate::core::bootstrap::bootstrap_utils::{cache_health_check, safe_pkill, vault_health_check, vector_db_health_check};
 use crate::core::config::AppConfig;
 use crate::core::package_manager::{InstallMode, PackageManager};
 use log::{info, warn};
@@ -79,13 +79,29 @@ impl BootstrapManager {
         }
 
         if pm.is_installed("vector_db") {
-            info!("Starting Vector database...");
-            match pm.start("vector_db") {
-                Ok(_child) => {
-                    info!("Vector database started");
-                }
-                Err(e) => {
-                    warn!("Failed to start Vector database: {}", e);
+            let vector_db_already_running = vector_db_health_check();
+            if vector_db_already_running {
+                info!("Vector database (Qdrant) is already running");
+            } else {
+                info!("Starting Vector database (Qdrant)...");
+                match pm.start("vector_db") {
+                    Ok(_child) => {
+                        info!("Vector database process started, waiting for readiness...");
+                        // Wait for vector_db to be ready
+                        for i in 0..15 {
+                            sleep(Duration::from_secs(1)).await;
+                            if vector_db_health_check() {
+                                info!("Vector database (Qdrant) is responding");
+                                break;
+                            }
+                            if i == 14 {
+                                warn!("Vector database did not respond after 15 seconds");
+                            }
+                        }
+                    }
+                    Err(e) => {
+                        warn!("Failed to start Vector database: {}", e);
+                    }
                 }
             }
         }

src/core/bootstrap/bootstrap_utils.rs

@@ -146,6 +146,40 @@ pub fn cache_health_check() -> bool {
     }
 }
 
+/// Check if Qdrant vector database is healthy
+pub fn vector_db_health_check() -> bool {
+    // Qdrant has a /healthz endpoint, use curl to check
+    // Try both HTTP and HTTPS
+    let urls = [
+        "http://localhost:6333/healthz",
+        "https://localhost:6333/healthz",
+    ];
+
+    for url in &urls {
+        if let Ok(output) = Command::new("curl")
+            .args(["-f", "-s", "--connect-timeout", "2", "-k", url])
+            .output()
+        {
+            if output.status.success() {
+                // Qdrant healthz returns "OK" or JSON with status
+                let response = String::from_utf8_lossy(&output.stdout);
+                if response.contains("OK") || response.contains("\"status\":\"ok\"") {
+                    return true;
+                }
+            }
+        }
+    }
+
+    // Fallback: just check if port 6333 is listening
+    match Command::new("nc")
+        .args(["-z", "-w", "1", "127.0.0.1", "6333"])
+        .output()
+    {
+        Ok(output) => output.status.success(),
+        Err(_) => false,
+    }
+}
+
 /// Get current user safely
 pub fn safe_fuser() -> String {
     // Return shell command that uses $USER environment variable

src/llm/local.rs

@@ -81,18 +81,22 @@ pub async fn ensure_llama_servers_running(
     };
 
     let llm_model = if llm_model.is_empty() {
-        info!("No LLM model configured, using default: ../../../../data/llm/DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_M.gguf");
-        "../../../../data/llm/DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_M.gguf".to_string()
+        info!("No LLM model configured, using default: DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_M.gguf");
+        "DeepSeek-R1-Distill-Qwen-1.5B-Q3_K_M.gguf".to_string()
     } else {
         llm_model
     };
 
     let embedding_model = if embedding_model.is_empty() {
-        info!("No embedding model configured, using default: ../../../../data/llm/bge-small-en-v1.5-f32.gguf");
-        "../../../../data/llm/bge-small-en-v1.5-f32.gguf".to_string()
+        info!("No embedding model configured, using default: bge-small-en-v1.5-f32.gguf");
+        "bge-small-en-v1.5-f32.gguf".to_string()
     } else {
         embedding_model
     };
+
+    // For llama-server startup, we need the full path
+    let llm_model_path = format!("{}/../../../../data/llm/{}", llm_server_path, llm_model);
+    let embedding_model_path = format!("{}/../../../../data/llm/{}", llm_server_path, embedding_model);
     if !llm_server_enabled {
         info!("Local LLM server management disabled (llm-server=false). Using external endpoints.");
         info!("  LLM URL: {llm_url}");
@@ -160,13 +164,13 @@ pub async fn ensure_llama_servers_running(
         info!("Starting LLM server...");
         let app_state_clone = Arc::clone(&app_state);
         let llm_server_path_clone = llm_server_path.clone();
-        let llm_model_clone = llm_model.clone();
+        let llm_model_path_clone = llm_model_path.clone();
         let llm_url_clone = llm_url.clone();
         tasks.push(tokio::spawn(async move {
             start_llm_server(
                 app_state_clone,
                 llm_server_path_clone,
-                llm_model_clone,
+                llm_model_path_clone,
                 llm_url_clone,
             )
         }));
@@ -177,7 +181,7 @@ pub async fn ensure_llama_servers_running(
         info!("Starting Embedding server...");
         tasks.push(tokio::spawn(start_embedding_server(
             llm_server_path.clone(),
-            embedding_model.clone(),
+            embedding_model_path.clone(),
             embedding_url.clone(),
         )));
     } else if embedding_model.is_empty() {

src/main_module/bootstrap.rs

@@ -432,7 +432,7 @@ pub async fn create_app_state(
         info!("LLM Model: {}", llm_model);
     }
 
-    let _llm_key = std::env::var("LLM_KEY")
+    let llm_key = std::env::var("LLM_KEY")
         .or_else(|_| std::env::var("OPENAI_API_KEY"))
         .or_else(|_| {
             config_manager
@@ -441,6 +441,18 @@ pub async fn create_app_state(
         })
         .unwrap_or_default();
 
+    // If llm-url points to external API but no key is configured, fall back to local LLM
+    let llm_url = if llm_key.is_empty()
+        && !llm_url.contains("localhost")
+        && !llm_url.contains("127.0.0.1")
+        && (llm_url.contains("api.z.ai") || llm_url.contains("openai.com") || llm_url.contains("anthropic.com"))
+    {
+        warn!("External LLM URL configured ({}), but no API key provided. Falling back to local LLM at http://localhost:8081", llm_url);
+        "http://localhost:8081".to_string()
+    } else {
+        llm_url
+    };
+
     // LLM endpoint path configuration
     let llm_endpoint_path = config_manager
         .get_config(