diff --git a/src/openai/pipelines/llm_engine.rs b/src/openai/pipelines/llm_engine.rs
index 4cffbdd..0ba766c 100644
--- a/src/openai/pipelines/llm_engine.rs
+++ b/src/openai/pipelines/llm_engine.rs
@@ -370,7 +370,12 @@ impl<'a> LLMEngine<'a> {
                     }
 
                     let block_number = if i / self.cache_config.block_size >= table.len() {
-                        table.get(table.len() - 1).unwrap() //position exceed! use last position
+                        panic!(
+                            "Block table is too small (prompt)! i={} block_size={} table_len={}",
+                            i,
+                            self.cache_config.block_size,
+                            table.len()
+                        );
                     } else {
                         table.get(i / self.cache_config.block_size).unwrap()
                     };
@@ -460,7 +465,7 @@ impl<'a> LLMEngine<'a> {
                     .collect::<Vec<_>>();
 
                 let block_number = if position / self.cache_config.block_size >= table.len() {
-                    table.get(table.len() - 1).unwrap() //position exceed! use last position; TODO (bug fix)
+                    panic!("Block table is too small (completion)! start_pos={} block_size={} table_len={}", position, self.cache_config.block_size, table.len());
                 } else {
                     table.get(position / self.cache_config.block_size).unwrap()
                 };
diff --git a/src/scheduler/block_engine.rs b/src/scheduler/block_engine.rs
index 1db9507..a9ac7a3 100644
--- a/src/scheduler/block_engine.rs
+++ b/src/scheduler/block_engine.rs
@@ -27,6 +27,10 @@ impl LogicalTokenBlock {
         self.num_tokens == self.block_size
     }
 
+    pub fn is_empty(&self) -> bool {
+        self.num_tokens == 0
+    }
+
     pub fn append_token_id(&mut self, token: usize) {
         assert!(!self.is_full());
         self.tokens[self.num_tokens] = token;
diff --git a/src/scheduler/sequence.rs b/src/scheduler/sequence.rs
index 6e2db88..b1fc0db 100644
--- a/src/scheduler/sequence.rs
+++ b/src/scheduler/sequence.rs
@@ -76,7 +76,7 @@ impl _Sequence {
 
     pub fn blocks_to_add_new_tok(&self) -> usize {
         let last = self.logical_token_blocks.last();
-        if !last.is_some_and(|last| last.is_full()) {
+        if !last.is_some_and(|last| last.is_full() || last.is_empty()) {
             // If we have space
             0
         } else {
@@ -168,17 +168,22 @@ impl _Sequence {
 
     fn append_token_to_blocks(&mut self, token: usize) {
         let last = self.logical_token_blocks.last_mut();
-        if last.is_some() && !last.as_ref().is_some_and(|last| last.is_full()) {
-            // If we have space
-            let last = last.unwrap();
-            last.append_token_id(token);
-        } else {
-            self.logical_token_blocks
-                .push(LogicalTokenBlock::new(self.block_size));
+        match last {
+            Some(last) => {
+                last.append_token_id(tok);
+            }
+            None => {
+                self.logical_token_blocks
+                    .push(LogicalTokenBlock::new(self.block_size));
+                self.logical_token_blocks
+                    .last_mut()
+                    .unwrap()
+                    .append_token_id(tok);
+            }
+        }
+        if self.logical_token_blocks.last().as_ref().unwrap().is_full() {
             self.logical_token_blocks
-                .last_mut()
-                .unwrap()
-                .append_token_id(token);
+                .push(LogicalTokenBlock::new(*block_size));
         }
     }
 }