From bbd3047633f14213fbd409a36bd0c9cdb34c5fd7 Mon Sep 17 00:00:00 2001
From: Yusuf Simonson <simonson@gmail.com>
Date: Fri, 28 Nov 2025 11:32:31 -0500
Subject: [PATCH 1/2] Fix capitalization of LlamaCppError

---
 llama-cpp-2/src/lib.rs           |  4 ++--
 llama-cpp-2/src/llama_backend.rs |  8 ++++----
 llama-cpp-2/src/model/params.rs  | 10 +++++-----
 llama-cpp-sys-2/llama.cpp        |  2 +-
 4 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/llama-cpp-2/src/lib.rs b/llama-cpp-2/src/lib.rs
index 16f05e31..1daa7f9f 100644
--- a/llama-cpp-2/src/lib.rs
+++ b/llama-cpp-2/src/lib.rs
@@ -35,11 +35,11 @@ pub mod token;
 pub mod token_type;
 
 /// A failable result from a llama.cpp function.
-pub type Result<T> = std::result::Result<T, LLamaCppError>;
+pub type Result<T> = std::result::Result<T, LlamaCppError>;
 
 /// All errors that can occur in the llama-cpp crate.
 #[derive(Debug, Eq, PartialEq, thiserror::Error)]
-pub enum LLamaCppError {
+pub enum LlamaCppError {
     /// The backend was already initialized. This can generally be ignored as initializing the backend
     /// is idempotent.
     #[error("BackendAlreadyInitialized")]
diff --git a/llama-cpp-2/src/llama_backend.rs b/llama-cpp-2/src/llama_backend.rs
index 1cc3fa3d..7954e9c6 100644
--- a/llama-cpp-2/src/llama_backend.rs
+++ b/llama-cpp-2/src/llama_backend.rs
@@ -1,6 +1,6 @@
 //! Representation of an initialized llama backend
 
-use crate::LLamaCppError;
+use crate::LlamaCppError;
 use llama_cpp_sys_2::ggml_log_level;
 use std::sync::atomic::AtomicBool;
 use std::sync::atomic::Ordering::SeqCst;
@@ -18,7 +18,7 @@ impl LlamaBackend {
     fn mark_init() -> crate::Result<()> {
         match LLAMA_BACKEND_INITIALIZED.compare_exchange(false, true, SeqCst, SeqCst) {
             Ok(_) => Ok(()),
-            Err(_) => Err(LLamaCppError::BackendAlreadyInitialized),
+            Err(_) => Err(LlamaCppError::BackendAlreadyInitialized),
         }
     }
 
@@ -28,7 +28,7 @@ impl LlamaBackend {
     ///
     /// ```
     ///# use llama_cpp_2::llama_backend::LlamaBackend;
-    ///# use llama_cpp_2::LLamaCppError;
+    ///# use llama_cpp_2::LlamaCppError;
     ///# use std::error::Error;
     ///
     ///# fn main() -> Result<(), Box<dyn Error>> {
@@ -36,7 +36,7 @@ impl LlamaBackend {
     ///
     /// let backend = LlamaBackend::init()?;
     /// // the llama backend can only be initialized once
-    /// assert_eq!(Err(LLamaCppError::BackendAlreadyInitialized), LlamaBackend::init());
+    /// assert_eq!(Err(LlamaCppError::BackendAlreadyInitialized), LlamaBackend::init());
     ///
     ///# Ok(())
     ///# }
diff --git a/llama-cpp-2/src/model/params.rs b/llama-cpp-2/src/model/params.rs
index 47f4c257..a54a4485 100644
--- a/llama-cpp-2/src/model/params.rs
+++ b/llama-cpp-2/src/model/params.rs
@@ -1,7 +1,7 @@
 //! A safe wrapper around `llama_model_params`.
 
 use crate::model::params::kv_overrides::KvOverrides;
-use crate::LLamaCppError;
+use crate::LlamaCppError;
 use std::ffi::{c_char, CStr};
 use std::fmt::{Debug, Formatter};
 use std::pin::Pin;
@@ -375,19 +375,19 @@ impl LlamaModelParams {
     /// You don't need to specify CPU or ACCEL devices.
     ///
     /// # Errors
-    /// Returns `LLamaCppError::BackendDeviceNotFound` if any device index is invalid.
-    pub fn with_devices(mut self, devices: &[usize]) -> Result<Self, LLamaCppError> {
+    /// Returns `LlamaCppError::BackendDeviceNotFound` if any device index is invalid.
+    pub fn with_devices(mut self, devices: &[usize]) -> Result<Self, LlamaCppError> {
         for dev in self.devices.iter_mut() {
             *dev = std::ptr::null_mut();
         }
         // Check device count
         let max_devices = crate::max_devices().min(LLAMA_CPP_MAX_DEVICES);
         if devices.len() > max_devices {
-            return Err(LLamaCppError::MaxDevicesExceeded(max_devices));
+            return Err(LlamaCppError::MaxDevicesExceeded(max_devices));
         }
         for (i, &dev) in devices.iter().enumerate() {
             if dev >= unsafe { llama_cpp_sys_2::ggml_backend_dev_count() } {
-                return Err(LLamaCppError::BackendDeviceNotFound(dev));
+                return Err(LlamaCppError::BackendDeviceNotFound(dev));
             }
             let backend_dev = unsafe { llama_cpp_sys_2::ggml_backend_dev_get(dev) };
             self.devices[i] = backend_dev;
diff --git a/llama-cpp-sys-2/llama.cpp b/llama-cpp-sys-2/llama.cpp
index d7395115..4c91f263 160000
--- a/llama-cpp-sys-2/llama.cpp
+++ b/llama-cpp-sys-2/llama.cpp
@@ -1 +1 @@
-Subproject commit d7395115baf395b75a73a17b0b796e746e468da9
+Subproject commit 4c91f2633f29a51ac2dcaa1c462483ea0ef6de8a

From de7db85f3c3752c6883786708eff8e253d96d6b9 Mon Sep 17 00:00:00 2001
From: Yusuf Simonson <simonson@gmail.com>
Date: Fri, 28 Nov 2025 11:48:03 -0500
Subject: [PATCH 2/2] Revert back to baseline llama.cpp

---
 llama-cpp-sys-2/llama.cpp | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/llama-cpp-sys-2/llama.cpp b/llama-cpp-sys-2/llama.cpp
index 4c91f263..d7395115 160000
--- a/llama-cpp-sys-2/llama.cpp
+++ b/llama-cpp-sys-2/llama.cpp
@@ -1 +1 @@
-Subproject commit 4c91f2633f29a51ac2dcaa1c462483ea0ef6de8a
+Subproject commit d7395115baf395b75a73a17b0b796e746e468da9