From 737f3eeff1bae501c3e7f383364a23e3c169a575 Mon Sep 17 00:00:00 2001
From: Yusuf Simonson <simonson@gmail.com>
Date: Fri, 28 Nov 2025 11:45:35 -0500
Subject: [PATCH] Added lifetime to LlamaBatch to track underlying data from
 get_one constructor

---
 examples/mtmd/src/mtmd.rs      |  6 +++---
 llama-cpp-2/src/llama_batch.rs | 12 ++++++++----
 2 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/examples/mtmd/src/mtmd.rs b/examples/mtmd/src/mtmd.rs
index e0f52de9..a2af6328 100644
--- a/examples/mtmd/src/mtmd.rs
+++ b/examples/mtmd/src/mtmd.rs
@@ -72,11 +72,11 @@ pub struct MtmdCliParams {
 
 /// State of the MTMD CLI application.
 #[allow(missing_debug_implementations)]
-pub struct MtmdCliContext {
+pub struct MtmdCliContext<'a> {
     /// The MTMD context for multimodal processing.
     pub mtmd_ctx: MtmdContext,
     /// The batch used for processing tokens.
-    pub batch: LlamaBatch,
+    pub batch: LlamaBatch<'a>,
     /// The list of loaded bitmaps (images/audio).
     pub bitmaps: Vec<MtmdBitmap>,
     /// The number of past tokens processed.
@@ -87,7 +87,7 @@ pub struct MtmdCliContext {
     pub chat: Vec<LlamaChatMessage>,
 }
 
-impl MtmdCliContext {
+impl<'a> MtmdCliContext<'a> {
     /// Creates a new MTMD CLI context
     ///
     /// # Errors
diff --git a/llama-cpp-2/src/llama_batch.rs b/llama-cpp-2/src/llama_batch.rs
index b96588c7..acb7ecec 100644
--- a/llama-cpp-2/src/llama_batch.rs
+++ b/llama-cpp-2/src/llama_batch.rs
@@ -2,10 +2,11 @@
 
 use crate::token::LlamaToken;
 use llama_cpp_sys_2::{llama_batch, llama_batch_free, llama_batch_init, llama_pos, llama_seq_id};
+use std::marker::PhantomData;
 
 /// A safe wrapper around `llama_batch`.
 #[derive(Debug)]
-pub struct LlamaBatch {
+pub struct LlamaBatch<'a> {
     /// The number of tokens the batch was allocated with. they are safe to write to - but not necessarily read from as they are not necessarily initialized
     allocated: usize,
     /// The logits that are initialized. Used by [`LlamaContext`] to ensure that only initialized logits are accessed.
@@ -13,6 +14,7 @@ pub struct LlamaBatch {
     #[allow(clippy::doc_markdown)]
     /// The llama_cpp batch. always initialize by `llama_cpp_sys_2::llama_batch_init(allocated, <unknown>, <unknown>)`
     pub(crate) llama_batch: llama_batch,
+    phantom: PhantomData<&'a [LlamaToken]>,
 }
 
 /// Errors that can occur when adding a token to a batch.
@@ -26,7 +28,7 @@ pub enum BatchAddError {
     EmptyBuffer,
 }
 
-impl LlamaBatch {
+impl<'a> LlamaBatch<'a> {
     /// Clear the batch. This does not free the memory associated with the batch, but it does reset
     /// the number of tokens to 0.
     pub fn clear(&mut self) {
@@ -150,6 +152,7 @@ impl LlamaBatch {
             allocated: n_tokens,
             initialized_logits: vec![],
             llama_batch: batch,
+            phantom: PhantomData,
         }
     }
 
@@ -163,7 +166,7 @@ impl LlamaBatch {
     ///
     /// # Panics
     /// If the number of tokens in ``tokens`` exceeds [`i32::MAX`].
-    pub fn get_one(tokens: &[LlamaToken]) -> Result<Self, BatchAddError> {
+    pub fn get_one(tokens: &'a [LlamaToken]) -> Result<Self, BatchAddError> {
         if tokens.is_empty() {
             return Err(BatchAddError::EmptyBuffer);
         }
@@ -183,6 +186,7 @@ impl LlamaBatch {
                 .try_into()
                 .expect("number of tokens exceeds i32::MAX + 1")],
             llama_batch: batch,
+            phantom: PhantomData,
         };
         Ok(batch)
     }
@@ -194,7 +198,7 @@ impl LlamaBatch {
     }
 }
 
-impl Drop for LlamaBatch {
+impl<'a> Drop for LlamaBatch<'a> {
     /// Drops the `LlamaBatch`.
     ///
     /// ```