Pass model via args.

1 files changed, 10 insertions, 10 deletions

diff --git a/main.c b/main.c
index 414a463..266bc34 100644
--- a/main.c
+++ b/main.c
@@ -7,8 +7,6 @@
 #include "mem.h"
 #include "llama.h"
 
-#define MODEL_PATH "Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf"
-
 #define SYSTEM_PROMPT \
     "You are a lexicographer. Define the target word strictly in the context provided.\n\n" \
     "Output format strictly as follows:\n" \
@@ -192,8 +190,13 @@ int main(int argc , char *argv[])
 	struct llama_model *model;
 	struct llama_model_params mparams;
 
-	if (unveil(MODEL_PATH, "r") == -1)
-		err(1, "unveil %s failed", MODEL_PATH);
+	if (argc < 3)
+		errx(1, "usage: %s [model] [prompt]", argv[0]);
+
+	const char *model_path = argv[1];
+
+	if (unveil(model_path, "r") == -1)
+		err(1, "unveil %s failed", model_path);
 
 	if (unveil(NULL, NULL) == -1)
 		err(1, "unveil lock failed");
@@ -201,23 +204,20 @@ int main(int argc , char *argv[])
 	if (pledge("stdio rpath", NULL) == -1)
 		err(1, "initial pledge failed");
 
-	if (argc < 2)
-		errx(1, "usage: %s [prompt]", argv[0]);
-
 	llama_backend_init();
 
 	mparams = llama_model_default_params();
 	mparams.n_gpu_layers = 0; /* force all layers onto CPU */
 	mparams.load_mode = LLAMA_LOAD_MODE_MMAP;
 
-	model = llama_model_load_from_file(MODEL_PATH, mparams);
+	model = llama_model_load_from_file(model_path, mparams);
 	if (!model)
-		errx(1, "failed to load model from file %s", MODEL_PATH);
+		errx(1, "failed to load model from file %s", model_path);
 
 	if (pledge("stdio", NULL) == -1)
 		err(1, "secondary pledge failed");
 
-	const char *prompt = argv[1];
+	const char *prompt = argv[2];
 	process_request(model, prompt);
 
 	llama_model_free(model);