diff --git a/crates/g3-core/src/prompts.rs b/crates/g3-core/src/prompts.rs
index 48075c1..ec383af 100644
--- a/crates/g3-core/src/prompts.rs
+++ b/crates/g3-core/src/prompts.rs
@@ -275,10 +275,6 @@ This ensures the TODO list is tracked against the specific version of requiremen
 // ============================================================================
 
 /// System prompt for providers with native tool calling (Anthropic, OpenAI, etc.)
-/// Note: This is kept for backwards compatibility but the function is preferred
-pub const SYSTEM_PROMPT_FOR_NATIVE_TOOL_USE: &str = "";
-
-/// Generate system prompt for native tool calling providers
 pub fn get_system_prompt_for_native() -> String {
     format!(
         "{}\n\n{}\n\n{}\n\n{}\n\n{}\n\n{}",
@@ -292,10 +288,6 @@ pub fn get_system_prompt_for_native() -> String {
 }
 
 /// System prompt for providers without native tool calling (embedded models)
-/// Note: This is kept for backwards compatibility but the function is preferred
-pub const SYSTEM_PROMPT_FOR_NON_NATIVE_TOOL_USE: &str = "";
-
-/// Generate system prompt for non-native tool calling providers (embedded models)
 pub fn get_system_prompt_for_non_native() -> String {
     format!(
         "{}\n\n{}\n\n{}\n\n{}{}\n\n{}\n\n{}\n\n{}",
diff --git a/crates/g3-providers/src/embedded.rs b/crates/g3-providers/src/embedded.rs
index 8e47646..526e0fd 100644
--- a/crates/g3-providers/src/embedded.rs
+++ b/crates/g3-providers/src/embedded.rs
@@ -42,8 +42,6 @@ pub struct EmbeddedProvider {
     temperature: f32,
     /// Context window size
     context_length: u32,
-    /// Number of GPU layers
-    gpu_layers: u32,
     /// Number of threads
     threads: Option<u32>,
 }
@@ -138,7 +136,6 @@ impl EmbeddedProvider {
             max_tokens,
             temperature: temperature.unwrap_or(0.1),
             context_length: context_size,
-            gpu_layers: n_gpu_layers,
             threads,
         })
     }
@@ -341,21 +338,6 @@ impl EmbeddedProvider {
         }
     }
 
-    /// Clean stop sequences from generated text.
-    fn clean_stop_sequences(&self, text: &str) -> String {
-        let mut cleaned = text.to_string();
-        let stop_sequences = self.get_stop_sequences();
-
-        for stop_seq in &stop_sequences {
-            if let Some(pos) = cleaned.find(stop_seq) {
-                cleaned.truncate(pos);
-                break; // Only remove the first occurrence to avoid over-truncation
-            }
-        }
-
-        cleaned.trim().to_string()
-    }
-
     /// Get the effective max tokens for generation
     fn effective_max_tokens(&self) -> u32 {
         self.max_tokens