Update app.py

app.py

@@ -100,37 +100,12 @@ def start_llama_server():
 
     binary_path = build_latest_llama_server()
 
-    from huggingface_hub import hf_hub_download, list_repo_files
+    from huggingface_hub import hf_hub_download
 
-    repo_id = "unsloth/GLM-5.3-Flash-GGUF"
-    quant = "UD-IQ2_XXS"
-
-    # Discover the real filenames instead of guessing them.
-    all_files = list_repo_files(repo_id)
-
-    model_files = sorted(
-        f for f in all_files
-        if f.endswith(".gguf") and quant in f and "mmproj" not in f.lower()
-    )
-    if not model_files:
-        raise RuntimeError(
-            f"No .gguf files matching '{quant}' found in {repo_id}. "
-            f"Available files: {all_files}"
-        )
-    print(f"Downloading {len(model_files)} model file(s): {model_files}")
-
-    # Download every shard; llama.cpp loads the rest automatically when
-    # pointed at the first one (…-00001-of-0000N.gguf).
-    local_paths = [hf_hub_download(repo_id=repo_id, filename=f) for f in model_files]
-    model_path = local_paths[0]
-
-    # Vision projector (optional): prefer BF16, then F16, then anything.
-    mmproj_files = [f for f in all_files if "mmproj" in f.lower() and f.endswith(".gguf")]
-    mmproj_file = next(
-        (f for tag in ("BF16", "F16") for f in mmproj_files if tag in f.upper()),
-        mmproj_files[0] if mmproj_files else None,
+    model_path = hf_hub_download(
+        repo_id="bartowski/moonshotai_Kimi-Linear-48B-A3B-Instruct-GGUF",
+        filename="moonshotai_Kimi-Linear-48B-A3B-Instruct-Q4_K_M.gguf",
     )
-    mmproj_path = hf_hub_download(repo_id=repo_id, filename=mmproj_file) if mmproj_file else None
 
     binary_dir = os.path.dirname(os.path.abspath(binary_path))
     env = os.environ.copy()
@@ -139,7 +114,6 @@ def start_llama_server():
     cmd = [
         binary_path,
         "-m", model_path,
-        *(["-mm", mmproj_path] if mmproj_path else []),
         "--port", "8000",
         "--host", "127.0.0.1",
         "-t", "16",