feat: configurable LLM request timeout via LLM_REQUEST_TIMEOUT_SECS

.env.example

@@ -5,6 +5,7 @@ DATABASE_POOL_SIZE=10
 # LLM Provider
 # LLM_BACKEND=nearai           # default
 # Possible values: nearai, ollama, openai_compatible, openai, anthropic, tinfoil
+# LLM_REQUEST_TIMEOUT_SECS=120  # Increase for local LLMs (Ollama, vLLM, LM Studio)
 
 # === Anthropic Direct ===
 # Two auth modes:

src/config/llm.rs

@@ -103,6 +103,10 @@ pub struct LlmConfig {
     /// Resolved provider config for registry-based providers.
     /// `None` when backend is "nearai".
     pub provider: Option<RegistryProviderConfig>,
+    /// HTTP request timeout in seconds for LLM API calls.
+    /// Default: 120. Increase for local LLMs (Ollama, vLLM, LM Studio) that
+    /// need more time for prompt evaluation on consumer hardware.
+    pub request_timeout_secs: u64,
 }
 
 /// NEAR AI configuration.
@@ -165,6 +169,7 @@ impl LlmConfig {
                 smart_routing_cascade: false,
             },
             provider: None,
+            request_timeout_secs: 120,
         }
     }
 
@@ -254,6 +259,8 @@ impl LlmConfig {
             )?)
         };
 
+        let request_timeout_secs = parse_optional_env("LLM_REQUEST_TIMEOUT_SECS", 120)?;
+
         Ok(Self {
             backend: if is_nearai {
                 "nearai".to_string()
@@ -265,6 +272,7 @@ impl LlmConfig {
             session,
             nearai,
             provider,
+            request_timeout_secs,
         })
     }
 
@@ -1016,4 +1024,30 @@ mod tests {
             assert_eq!(parsed, variant, "round-trip failed for {s}");
         }
     }
+
+    #[test]
+    fn test_request_timeout_defaults_to_120() {
+        let _guard = ENV_MUTEX.lock().expect("env mutex poisoned");
+        // SAFETY: Under ENV_MUTEX.
+        unsafe {
+            std::env::remove_var("LLM_REQUEST_TIMEOUT_SECS");
+        }
+        let config = LlmConfig::resolve(&Settings::default()).expect("resolve");
+        assert_eq!(config.request_timeout_secs, 120);
+    }
+
+    #[test]
+    fn test_request_timeout_configurable() {
+        let _guard = ENV_MUTEX.lock().expect("env mutex poisoned");
+        // SAFETY: Under ENV_MUTEX.
+        unsafe {
+            std::env::set_var("LLM_REQUEST_TIMEOUT_SECS", "300");
+        }
+        let config = LlmConfig::resolve(&Settings::default()).expect("resolve");
+        assert_eq!(config.request_timeout_secs, 300);
+        // SAFETY: Cleanup
+        unsafe {
+            std::env::remove_var("LLM_REQUEST_TIMEOUT_SECS");
+        }
+    }
 }

src/llm/mod.rs

@@ -58,8 +58,10 @@ pub fn create_llm_provider(
     config: &LlmConfig,
     session: Arc<SessionManager>,
 ) -> Result<Arc<dyn LlmProvider>, LlmError> {
+    let timeout = config.request_timeout_secs;
+
     if config.backend == "nearai" || config.backend == "near_ai" || config.backend == "near" {
-        return create_llm_provider_with_config(&config.nearai, session);
+        return create_llm_provider_with_config(&config.nearai, session, timeout);
     }
 
     let reg_config = config
@@ -79,6 +81,7 @@ pub fn create_llm_provider(
 pub fn create_llm_provider_with_config(
     config: &NearAiConfig,
     session: Arc<SessionManager>,
+    request_timeout_secs: u64,
 ) -> Result<Arc<dyn LlmProvider>, LlmError> {
     let auth_mode = if config.api_key.is_some() {
         "API key"
@@ -89,9 +92,14 @@ pub fn create_llm_provider_with_config(
         model = %config.model,
         base_url = %config.base_url,
         auth = auth_mode,
+        timeout_secs = request_timeout_secs,
         "Using NEAR AI (Chat Completions API)"
     );
-    Ok(Arc::new(NearAiChatProvider::new(config.clone(), session)?))
+    Ok(Arc::new(NearAiChatProvider::new_with_timeout(
+        config.clone(),
+        session,
+        request_timeout_secs,
+    )?))
 }
 
 /// Create a provider from a registry-resolved config.
@@ -365,7 +373,11 @@ pub fn build_provider_chain(
     let llm: Arc<dyn LlmProvider> = if let Some(ref cheap_model) = config.nearai.cheap_model {
         let mut cheap_config = config.nearai.clone();
         cheap_config.model = cheap_model.clone();
-        let cheap = create_llm_provider_with_config(&cheap_config, session.clone())?;
+        let cheap = create_llm_provider_with_config(
+            &cheap_config,
+            session.clone(),
+            config.request_timeout_secs,
+        )?;
         let cheap: Arc<dyn LlmProvider> = if retry_config.max_retries > 0 {
             Arc::new(RetryProvider::new(cheap, retry_config.clone()))
         } else {
@@ -397,7 +409,11 @@ pub fn build_provider_chain(
         }
         let mut fallback_config = config.nearai.clone();
         fallback_config.model = fallback_model.clone();
-        let fallback = create_llm_provider_with_config(&fallback_config, session.clone())?;
+        let fallback = create_llm_provider_with_config(
+            &fallback_config,
+            session.clone(),
+            config.request_timeout_secs,
+        )?;
         tracing::info!(
             primary = %llm.model_name(),
             fallback = %fallback.model_name(),
@@ -503,6 +519,7 @@ mod tests {
             session: SessionConfig::default(),
             nearai: test_nearai_config(),
             provider: None,
+            request_timeout_secs: 120,
         }
     }
 

src/llm/nearai_chat.rs

@@ -58,17 +58,28 @@ impl NearAiChatProvider {
     /// By default this enables tool-message flattening for compatibility with
     /// providers that reject `role: "tool"` messages.
     pub fn new(config: NearAiConfig, session: Arc<SessionManager>) -> Result<Self, LlmError> {
-        Self::new_with_flatten(config, session, true)
+        Self::new_with_options(config, session, true, 120)
     }
 
-    /// Create a chat completions provider with configurable tool-message flattening.
-    pub fn new_with_flatten(
+    /// Create a new provider with a custom request timeout.
+    pub fn new_with_timeout(
+        config: NearAiConfig,
+        session: Arc<SessionManager>,
+        request_timeout_secs: u64,
+    ) -> Result<Self, LlmError> {
+        Self::new_with_options(config, session, true, request_timeout_secs)
+    }
+
+    /// Create a chat completions provider with configurable tool-message flattening
+    /// and request timeout.
+    pub fn new_with_options(
         config: NearAiConfig,
         session: Arc<SessionManager>,
         flatten_tool_messages: bool,
+        request_timeout_secs: u64,
     ) -> Result<Self, LlmError> {
         let client = Client::builder()
-            .timeout(std::time::Duration::from_secs(120))
+            .timeout(std::time::Duration::from_secs(request_timeout_secs))
             .build()
             .map_err(|e| LlmError::RequestFailed {
                 provider: "nearai_chat".to_string(),

src/setup/wizard.rs

@@ -1476,6 +1476,7 @@ impl SetupWizard {
                 smart_routing_cascade: true,
             },
             provider: None,
+            request_timeout_secs: 120,
         };
 
         match create_llm_provider(&config, session) {