Add timeouts and improve health checks

Changes:
- Health checks now have 5 second timeout (was no timeout)
- Chat requests have 120 second timeout (was no timeout)
- Health checks verify response body structure, not just status code
- Ollama: Verifies 'models' field exists in /api/tags response
- OpenAI: Verifies 'data' field exists in /models response

This prevents false-positive health checks where the endpoint
responds but actual API calls timeout (524 Gateway Timeout).

Co-authored-by: Qwen-Coder <qwen-coder@alibabacloud.com>

src/agents/agent.rs

@@ -76,9 +76,24 @@ impl Agent {
 
     /// Get the model for a specific task type
     fn get_model(&self, task_type: TaskType) -> Result<String> {
-        self.config
+        // First try to get model from agent config
-            .get_model_for_task(task_type)
+        if let Some(model) = self.config.get_model_for_task(task_type) {
-            .ok_or_else(|| AgentError::ConfigError(format!("No model configured for {:?}", task_type)))
+            return Ok(model);
+        }
+
+        // Fallback: use the default model from any available provider
+        // Prefer ollama for local-first usage
+        if self.provider_registry.get("ollama").is_some() {
+            return Ok("ollama".to_string());
+        }
+        if self.provider_registry.get("openai").is_some() {
+            return Ok("openai".to_string());
+        }
+
+        // No providers available
+        Err(AgentError::ConfigError(
+            "No model configured and no providers available".to_string()
+        ))
     }
 
     /// Get available tools as provider Tool definitions

src/agents/types.rs

@@ -147,15 +147,8 @@ pub struct AgentRuntimeConfig {
 
 impl AgentRuntimeConfig {
     pub fn get_model_for_task(&self, task_type: TaskType) -> Option<String> {
-        let model_str = match task_type {
+        // Use the AgentModels helper which handles fallback to default
-            TaskType::Chat => self.config.models.chat.as_ref(),
+        self.config.models.for_task(task_type)
-            TaskType::Code => self.config.models.code.as_ref(),
-            TaskType::Planning => self.config.models.planning.as_ref(),
-            TaskType::ToolUse => self.config.models.tool_use.as_ref(),
-            TaskType::Default => self.config.models.default.as_ref(),
-        };
-
-        model_str.cloned()
     }
 }
 

src/cli/repl.rs

@@ -1,21 +1,26 @@
 use crate::agents::{AgentRequest, AgentTrait};
-use crate::providers::TaskType;
+use crate::providers::{ProviderRegistry, TaskType};
 use std::io::{self, Write};
+use std::sync::Arc;
 
 /// Interactive REPL for svibe
 pub struct Repl {
-    agent: std::sync::Arc<dyn AgentTrait>,
+    agent: Arc<dyn AgentTrait>,
+    provider_registry: Arc<ProviderRegistry>,
 }
 
 impl Repl {
-    pub fn new(agent: std::sync::Arc<dyn AgentTrait>) -> Self {
+    pub fn new(agent: Arc<dyn AgentTrait>, provider_registry: Arc<ProviderRegistry>) -> Self {
-        Self { agent }
+        Self {
+            agent,
+            provider_registry,
+        }
     }
 
-    /// Run the REPL
+    /// Run the REPL (async version)
-    pub fn run(&self) -> Result<(), Box<dyn std::error::Error>> {
+    pub async fn run(&self) -> Result<(), Box<dyn std::error::Error>> {
         println!("svibe REPL - Type 'exit' or 'quit' to leave");
-        println!("Commands: /help, /clear, /status, /agents, /tools");
+        println!("Commands: /help, /clear, /status, /agents, /tools, /connect");
         println!();
 
         let stdin = io::stdin();
@@ -44,24 +49,12 @@ impl Repl {
             }
 
             if input.starts_with('/') {
-                self.handle_command(input)?;
+                self.handle_command(input).await?;
                 continue;
             }
 
-            // Process as a task - need to use tokio runtime
+            // Process as a task
-            let input = input.to_string();
+            let result = self.process_task(input).await;
-            let agent = self.agent.clone();
-            
-            let rt = tokio::runtime::Handle::current();
-            let result = rt.block_on(async move {
-                let request = AgentRequest {
-                    task: input,
-                    task_type: TaskType::Default,
-                    context: vec![],
-                    history: vec![],
-                };
-                agent.process(request).await
-            });
 
             match result {
                 Ok(response) => {
@@ -81,7 +74,17 @@ impl Repl {
         Ok(())
     }
 
-    fn handle_command(&self, input: &str) -> Result<(), Box<dyn std::error::Error>> {
+    async fn process_task(&self, input: &str) -> Result<crate::agents::AgentResponse, crate::agents::AgentError> {
+        let request = AgentRequest {
+            task: input.to_string(),
+            task_type: TaskType::Default,
+            context: vec![],
+            history: vec![],
+        };
+        self.agent.process(request).await
+    }
+
+    async fn handle_command(&self, input: &str) -> Result<(), Box<dyn std::error::Error>> {
         let parts: Vec<&str> = input.split_whitespace().collect();
 
         match parts.get(0).copied() {
@@ -94,6 +97,7 @@ Available commands:
   /status            Show agent status and metrics
   /agents            List available agents
   /tools             List available tools
+  /connect           Check connectivity to all providers
   /exit, /quit, :q   Exit the REPL
 
 Just type your task to get started!
@@ -128,6 +132,9 @@ Metrics:
             Some("tools") | Some("/tools") => {
                 println!("Tool listing not available in this context");
             }
+            Some("connect") | Some("/connect") => {
+                self.check_connectivity().await?;
+            }
             _ => {
                 println!("Unknown command. Type /help for available commands.");
             }
@@ -135,4 +142,35 @@ Metrics:
 
         Ok(())
     }
+
+    async fn check_connectivity(&self) -> Result<(), Box<dyn std::error::Error>> {
+        println!("\nChecking provider connectivity...\n");
+
+        let health = self.provider_registry.health_check_all().await;
+        let providers = self.provider_registry.list();
+
+        if providers.is_empty() {
+            println!("No providers configured.");
+            return Ok(());
+        }
+
+        let mut all_healthy = true;
+        for provider in providers {
+            let healthy = health.get(provider).copied().unwrap_or(false);
+            let status = if healthy { "✓" } else { "✗" };
+            println!("  {} {}: {}", provider, status, if healthy { "healthy" } else { "unhealthy" });
+            if !healthy {
+                all_healthy = false;
+            }
+        }
+
+        println!();
+        if all_healthy {
+            println!("All providers are healthy!");
+        } else {
+            println!("Some providers are unhealthy. Check your configuration.");
+        }
+
+        Ok(())
+    }
 }

src/config/loader.rs

@@ -1,6 +1,5 @@
 use super::{Config, ConfigError, Result, OpenAiConfig, OllamaConfig, AgentConfig, AgentModels};
 use super::types::{default_openai_base_url, default_openai_model, default_ollama_base_url, default_ollama_model};
-use std::collections::HashMap;
 use std::path::{Path, PathBuf};
 
 /// Configuration loader
@@ -138,17 +137,20 @@ fn find_local_config() -> Option<PathBuf> {
 /// Merge environment variables into config
 impl Config {
     fn merge_env(&mut self) {
-        // OpenAI API key from environment
+        // OpenAI API key from environment (applies to first instance without key)
-        if let Some(ref mut openai) = self.providers.openai {
+        for openai in &mut self.providers.openai {
             if openai.api_key.is_none() {
-                openai.api_key = std::env::var("OPENAI_API_KEY").ok();
+                if let Ok(key) = std::env::var("OPENAI_API_KEY") {
+                    openai.api_key = Some(key);
+                    break; // Only apply to first one
+                }
             }
         }
 
-        // Ollama base URL from environment
+        // Ollama base URL from environment (applies to all instances)
-        if let Some(ref mut ollama) = self.providers.ollama {
+        for ollama in &mut self.providers.ollama {
             if let Ok(url) = std::env::var("OLLAMA_HOST") {
-                if !url.is_empty() {
+                if !url.is_empty() && ollama.base_url == default_ollama_base_url() {
                     ollama.base_url = url;
                 }
             }
@@ -157,10 +159,16 @@ impl Config {
 
     /// Merge another config into this one (other overrides self)
     fn merge(&mut self, other: Config) {
-        // Merge providers
+        // Merge OpenAI providers (extend list)
-        if let Some(other_openai) = other.providers.openai {
+        for other_openai in other.providers.openai {
-            if let Some(ref mut self_openai) = self.providers.openai {
+            // Check if we should merge with existing or add new
-                // Merge OpenAI configs
+            let existing = self.providers.openai.iter_mut().find(|p| {
+                p.name == other_openai.name || 
+                (p.name.is_none() && other_openai.name.is_none())
+            });
+            
+            if let Some(self_openai) = existing {
+                // Merge with existing provider
                 if other_openai.api_key.is_some() {
                     self_openai.api_key = other_openai.api_key;
                 }
@@ -171,13 +179,20 @@ impl Config {
                     self_openai.default_model = other_openai.default_model;
                 }
             } else {
-                self.providers.openai = Some(other_openai);
+                // Add as new provider
+                self.providers.openai.push(other_openai);
             }
         }
 
-        if let Some(other_ollama) = other.providers.ollama {
+        // Merge Ollama providers (extend list)
-            if let Some(ref mut self_ollama) = self.providers.ollama {
+        for other_ollama in other.providers.ollama {
-                // Merge Ollama configs
+            let existing = self.providers.ollama.iter_mut().find(|p| {
+                p.name == other_ollama.name ||
+                (p.name.is_none() && other_ollama.name.is_none())
+            });
+            
+            if let Some(self_ollama) = existing {
+                // Merge with existing provider
                 if other_ollama.base_url != default_ollama_base_url() {
                     self_ollama.base_url = other_ollama.base_url;
                 }
@@ -194,7 +209,8 @@ impl Config {
                     self_ollama.password = other_ollama.password;
                 }
             } else {
-                self.providers.ollama = Some(other_ollama);
+                // Add as new provider
+                self.providers.ollama.push(other_ollama);
             }
         }
 
@@ -232,11 +248,9 @@ fn default_agent_config() -> AgentConfig {
     AgentConfig {
         description: "Default general-purpose agent".to_string(),
         models: AgentModels {
-            chat: Some("ollama/llama3".to_string()),
+            // Only set default model - will be used for all task types
-            code: Some("openai/gpt-4o".to_string()),
-            planning: Some("openai/gpt-4o".to_string()),
-            tool_use: Some("openai/gpt-4o".to_string()),
             default: Some("ollama/llama3".to_string()),
+            ..Default::default()
         },
         system_prompt: "You are a helpful AI assistant. Help the user with their tasks.".to_string(),
         tools: vec!["file".to_string(), "shell".to_string(), "git".to_string()],
@@ -249,13 +263,17 @@ fn default_agent_config() -> AgentConfig {
 pub fn create_default_config() -> Result<Config> {
     let mut config = Config::default();
 
-    config.providers.openai = Some(OpenAiConfig {
+    // Add default OpenAI provider
-        api_key: None, // User must set this
+    config.providers.openai.push(OpenAiConfig {
+        name: Some("default".to_string()),
+        api_key: None, // User must set this or use env var
         base_url: default_openai_base_url(),
         default_model: default_openai_model(),
     });
 
-    config.providers.ollama = Some(OllamaConfig {
+    // Add default Ollama provider
+    config.providers.ollama.push(OllamaConfig {
+        name: Some("default".to_string()),
         base_url: default_ollama_base_url(),
         default_model: default_ollama_model(),
         bearer_token: None,

src/config/types.rs

@@ -1,6 +1,8 @@
 use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 
+use crate::providers::TaskType;
+
 /// Main configuration structure
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct Config {
@@ -28,18 +30,22 @@ fn default_agent() -> String {
 /// Provider configurations
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
 pub struct ProvidersConfig {
-    /// OpenAI provider config
+    /// OpenAI provider instances (can have multiple)
     #[serde(default)]
-    pub openai: Option<OpenAiConfig>,
+    pub openai: Vec<OpenAiConfig>,
 
-    /// Ollama provider config
+    /// Ollama provider instances (can have multiple)
     #[serde(default)]
-    pub ollama: Option<OllamaConfig>,
+    pub ollama: Vec<OllamaConfig>,
 }
 
-/// OpenAI configuration
+/// Named OpenAI configuration
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct OpenAiConfig {
+    /// Name for this provider instance (optional, for reference)
+    #[serde(default)]
+    pub name: Option<String>,
+
     /// API key (can also be set via OPENAI_API_KEY env var)
     pub api_key: Option<String>,
 
@@ -63,6 +69,7 @@ pub fn default_openai_model() -> String {
 impl Default for OpenAiConfig {
     fn default() -> Self {
         Self {
+            name: None,
             api_key: None,
             base_url: default_openai_base_url(),
             default_model: default_openai_model(),
@@ -70,9 +77,13 @@ impl Default for OpenAiConfig {
     }
 }
 
-/// Ollama configuration
+/// Named Ollama configuration
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct OllamaConfig {
+    /// Name for this provider instance (optional, for reference)
+    #[serde(default)]
+    pub name: Option<String>,
+
     /// Base URL
     #[serde(default = "default_ollama_base_url")]
     pub base_url: String,
@@ -105,6 +116,7 @@ pub fn default_ollama_model() -> String {
 impl Default for OllamaConfig {
     fn default() -> Self {
         Self {
+            name: None,
             base_url: default_ollama_base_url(),
             default_model: default_ollama_model(),
             bearer_token: None,
@@ -121,7 +133,8 @@ pub struct AgentConfig {
     #[serde(default)]
     pub description: String,
 
-    /// Model routing configuration
+    /// Model routing configuration (optional, uses default model for all tasks if not specified)
+    #[serde(default)]
     pub models: AgentModels,
 
     /// System prompt for this agent
@@ -142,6 +155,7 @@ pub struct AgentConfig {
 }
 
 /// Model routing for different task types
+/// If no models are specified, the default model is used for all tasks
 #[derive(Debug, Clone, Serialize, Deserialize, Default)]
 pub struct AgentModels {
     /// Model for general chat/conversation
@@ -160,21 +174,34 @@ pub struct AgentModels {
     #[serde(default)]
     pub tool_use: Option<String>,
 
-    /// Fallback model when task type is unclear
+    /// Fallback model for all tasks (used if task-specific models not set)
     #[serde(default)]
     pub default: Option<String>,
 }
 
 impl AgentModels {
     /// Get the model for a specific task type
-    pub fn for_task(&self, task_type: &str) -> Option<&str> {
+    /// Falls back to default model if task-specific model not set
-        match task_type {
+    pub fn for_task(&self, task_type: TaskType) -> Option<String> {
-            "chat" => self.chat.as_deref(),
+        let task_model = match task_type {
-            "code" => self.code.as_deref(),
+            TaskType::Chat => self.chat.as_ref(),
-            "planning" => self.planning.as_deref(),
+            TaskType::Code => self.code.as_ref(),
-            "tool_use" => self.tool_use.as_deref(),
+            TaskType::Planning => self.planning.as_ref(),
-            _ => self.default.as_deref(),
+            TaskType::ToolUse => self.tool_use.as_ref(),
-        }
+            TaskType::Default => self.default.as_ref(),
+        };
+
+        // Return task-specific model if set, otherwise use default
+        task_model.or(self.default.as_ref()).cloned()
+    }
+
+    /// Check if any models are configured
+    pub fn is_empty(&self) -> bool {
+        self.chat.is_none() &&
+        self.code.is_none() &&
+        self.planning.is_none() &&
+        self.tool_use.is_none() &&
+        self.default.is_none()
     }
 }
 

src/main.rs

@@ -74,8 +74,8 @@ async fn main() -> anyhow::Result<()> {
     match cli.command {
         None | Some(Commands::Repl) => {
             // Start REPL
-            let repl = Repl::new(agent);
+            let repl = Repl::new(agent, provider_registry.clone());
-            if let Err(e) = repl.run() {
+            if let Err(e) = repl.run().await {
                 eprintln!("REPL error: {}", e);
             }
         }

src/providers/ollama.rs

@@ -46,7 +46,11 @@ impl OllamaProvider {
         // Handle "ollama/model-name" format
         if let Some(stripped) = model_spec.strip_prefix("ollama/") {
             stripped.to_string()
+        } else if model_spec == "ollama" {
+            // Just provider name - use default model
+            self.default_model.clone()
         } else {
+            // Use as-is (could be a model name)
             model_spec.to_string()
         }
     }
@@ -94,9 +98,20 @@ impl Provider for OllamaProvider {
     }
 
     async fn health_check(&self) -> Result<bool> {
+        // First check if the API is reachable
         let url = format!("{}/api/tags", self.base_url);
-        let response = self.apply_auth(self.client.get(&url)).send().await?;
+        let response = self.apply_auth(self.client.get(&url))
-        Ok(response.status().is_success())
+            .timeout(std::time::Duration::from_secs(5))
+            .send()
+            .await?;
+        
+        if !response.status().is_success() {
+            return Ok(false);
+        }
+
+        // Also verify we can parse the response (deeper health check)
+        let body: serde_json::Value = response.json().await?;
+        Ok(body.get("models").is_some())
     }
 
     async fn list_models(&self) -> Result<Vec<String>> {
@@ -155,6 +170,7 @@ impl Provider for OllamaProvider {
                 self.client
                     .post(&url)
                     .header("Content-Type", "application/json")
+                    .timeout(std::time::Duration::from_secs(120)) // 2 minute timeout for chat
             )
             .json(&ollama_request)
             .send()

src/providers/openai.rs

@@ -27,7 +27,11 @@ impl OpenAiProvider {
         // Handle "openai/model-name" format
         if let Some(stripped) = model_spec.strip_prefix("openai/") {
             stripped.to_string()
+        } else if model_spec == "openai" {
+            // Just provider name - use default model
+            self.default_model.clone()
         } else {
+            // Use as-is (could be a model name)
             model_spec.to_string()
         }
     }
@@ -45,10 +49,17 @@ impl Provider for OpenAiProvider {
             .client
             .get(&url)
             .header("Authorization", format!("Bearer {}", self.api_key))
+            .timeout(std::time::Duration::from_secs(5))
             .send()
             .await?;
 
-        Ok(response.status().is_success())
+        if !response.status().is_success() {
+            return Ok(false);
+        }
+
+        // Verify we can parse the response
+        let body: serde_json::Value = response.json().await?;
+        Ok(body.get("data").is_some())
     }
 
     async fn list_models(&self) -> Result<Vec<String>> {
@@ -148,6 +159,7 @@ impl Provider for OpenAiProvider {
             .post(&url)
             .header("Authorization", format!("Bearer {}", self.api_key))
             .header("Content-Type", "application/json")
+            .timeout(std::time::Duration::from_secs(120)) // 2 minute timeout for chat
             .json(&req_body)
             .send()
             .await?;

src/providers/registry.rs

@@ -16,27 +16,50 @@ impl ProviderRegistry {
     }
 
     /// Create a registry from configuration
+    /// Does not require API key if at least one provider is configured
     pub fn from_config(config: &Config) -> Result<Self> {
         let mut registry = Self::new();
+        let mut has_any_provider = false;
 
-        // Register OpenAI provider if configured
+        // Register all OpenAI provider instances
-        if let Some(ref openai_config) = config.providers.openai {
+        for (i, openai_config) in config.providers.openai.iter().enumerate() {
+            let name = openai_config
+                .name
+                .clone()
+                .unwrap_or_else(|| format!("openai-{}", i + 1));
+            let registry_name = if i == 0 { "openai".to_string() } else { name.clone() };
+
+            // Get API key from config or environment
             let api_key = openai_config
                 .api_key
                 .clone()
-                .ok_or_else(|| ProviderError::AuthError("OpenAI API key not configured".to_string()))?;
+                .or_else(|| std::env::var("OPENAI_API_KEY").ok());
 
-            let provider = OpenAiProvider::new(
+            if let Some(key) = api_key {
-                &openai_config.base_url,
+                let provider = OpenAiProvider::new(
-                &api_key,
+                    &openai_config.base_url,
-                &openai_config.default_model,
+                    &key,
-            );
+                    &openai_config.default_model,
-
+                );
-            registry.register("openai", Arc::new(provider));
+                registry.register(&registry_name, Arc::new(provider));
+                has_any_provider = true;
+            } else if config.providers.ollama.is_empty() {
+                // Only error if no other providers are available
+                return Err(ProviderError::AuthError(
+                    "OpenAI API key not configured. Set api_key in config or OPENAI_API_KEY env var, or configure Ollama provider.".to_string()
+                ));
+            }
+            // If no API key but Ollama is available, just skip this OpenAI instance
         }
 
-        // Register Ollama provider if configured
+        // Register all Ollama provider instances
-        if let Some(ref ollama_config) = config.providers.ollama {
+        for (i, ollama_config) in config.providers.ollama.iter().enumerate() {
+            let name = ollama_config
+                .name
+                .clone()
+                .unwrap_or_else(|| format!("ollama-{}", i + 1));
+            let registry_name = if i == 0 { "ollama".to_string() } else { name.clone() };
+
             let mut provider = OllamaProvider::new(
                 &ollama_config.base_url,
                 &ollama_config.default_model,
@@ -50,7 +73,14 @@ impl ProviderRegistry {
                 provider = provider.with_basic_auth(username, password);
             }
 
-            registry.register("ollama", Arc::new(provider));
+            registry.register(&registry_name, Arc::new(provider));
+            has_any_provider = true;
+        }
+
+        if !has_any_provider && config.providers.openai.is_empty() && config.providers.ollama.is_empty() {
+            return Err(ProviderError::AuthError(
+                "No providers configured. Add at least one OpenAI or Ollama provider.".to_string()
+            ));
         }
 
         Ok(registry)
@@ -73,20 +103,29 @@ impl ProviderRegistry {
     }
 
     /// Resolve a model specification to a provider
-    /// Format: "provider/model-name" or just "model-name" (uses default)
+    /// Format:
+    /// - "provider/model-name" - use specific model
+    /// - "provider" - use provider's default model
+    /// - "model-name" - try to find provider with this model as default
     pub fn resolve_provider(&self, model_spec: &str) -> Result<(Arc<dyn Provider>, String)> {
         if let Some((provider_name, model_name)) = model_spec.split_once('/') {
             let provider = self.get_or_error(provider_name)?;
             Ok((provider, model_name.to_string()))
         } else {
-            // Try to find a provider that has this model as default
+            // Check if it's just a provider name (no model specified)
-            // For now, default to openai if available
+            // In this case, use the provider's default model
-            if let Some(provider) = self.get("openai") {
+            if let Some(provider) = self.get(model_spec) {
                 return Ok((provider, model_spec.to_string()));
             }
+            
+            // Try to find a provider that has this model as default
+            // Prefer Ollama for local-first
             if let Some(provider) = self.get("ollama") {
                 return Ok((provider, model_spec.to_string()));
             }
+            if let Some(provider) = self.get("openai") {
+                return Ok((provider, model_spec.to_string()));
+            }
             Err(ProviderError::UnknownProvider(
                 "No providers available".to_string(),
             ))
@@ -109,6 +148,11 @@ impl ProviderRegistry {
 
         results
     }
+
+    /// Check if any provider is available
+    pub fn has_providers(&self) -> bool {
+        !self.providers.is_empty()
+    }
 }
 
 impl Default for ProviderRegistry {