From dd2d52b5e35e00c4eec1df803bafade6d9b4ea90 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Wed, 15 Apr 2026 19:34:58 +0000
Subject: [PATCH] Fix -m None passed to llama-server when cached model not
 found

When LLAMA_CACHED_MODEL is set but the model isn't present in the cache,
find_cached.py was printing Python's None as the string "None", causing
start.sh to pass "-m None" to llama-server.

- find_cached.py: print an error to stderr and exit 1 when the model path
  cannot be resolved, instead of printing "None"
- start.sh: capture find_cached.py output into a local variable, check
  the exit code and guard against empty output before constructing
  CACHED_LLAMA_ARGS; also quote the env-var expansions to handle spaces

https://claude.ai/code/session_011ny5CFYnrzPbRneSzio5CR
---
 src/find_cached.py | 7 +++++++
 src/start.sh       | 8 +++++++-
 2 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/src/find_cached.py b/src/find_cached.py
index 97a5b95..5dfd6b2 100644
--- a/src/find_cached.py
+++ b/src/find_cached.py
@@ -3,6 +3,7 @@ Finds the full LLM GGUF path from the Hugging Face cache.
 """
 
 import os
+import sys
 import argparse
 
 CACHE_DIR = "/runpod-volume/huggingface-cache/hub"
@@ -52,6 +53,12 @@ def main():
     args = parser.parse_args()
 
     model_path = find_model_path(args.model, args.path)
+    if model_path is None:
+        print(
+            f"Error: Cached model not found. Model='{args.model}', GGUF='{args.path}', Cache dir='{CACHE_DIR}'",
+            file=sys.stderr,
+        )
+        sys.exit(1)
     print(model_path, end="")
 
 
diff --git a/src/start.sh b/src/start.sh
index d57f7d3..52d5cb0 100644
--- a/src/start.sh
+++ b/src/start.sh
@@ -17,7 +17,13 @@ cleanup() {
 CACHED_LLAMA_ARGS=""
 
 find_cached_path() {
-    CACHED_LLAMA_ARGS="-m $(python ./find_cached.py $LLAMA_CACHED_MODEL $LLAMA_CACHED_GGUF_PATH)"
+    local model_path
+    model_path=$(python ./find_cached.py "$LLAMA_CACHED_MODEL" "$LLAMA_CACHED_GGUF_PATH")
+    if [ $? -ne 0 ] || [ -z "$model_path" ]; then
+        echo "start.sh: Error: Could not resolve cached model path. Check that LLAMA_CACHED_MODEL and LLAMA_CACHED_GGUF_PATH are correct and the model is fully cached on the network volume."
+        exit 1
+    fi
+    CACHED_LLAMA_ARGS="-m $model_path"
 }
 
 # check if $LLAMA_CACHED_MODEL is set and not empty