Merge pull request #477 from rhatdan/llama.cpp

Revert "Switch to llama-simple-chat"

ramalama/model.py

@@ -262,8 +262,19 @@ def run(self, args):
         if not args.container:
             exec_model_path = model_path
 
-        exec_args = ["llama-simple-chat", "-m", exec_model_path]
-        exec_args += self.common_params
+        exec_args = ["llama-cli", "-m", exec_model_path, "--in-prefix", "", "--in-suffix", ""]
+
+        if not args.debug:
+            exec_args += ["--no-display-prompt"]
+
+        exec_args += [
+            "-p",
+            prompt,
+        ] + self.common_params
+
+        if not args.ARGS and sys.stdin.isatty():
+            exec_args.append("-cnv")
+
         if args.gpu:
             exec_args.extend(self.gpu_args())
 

test/system/030-run.bats

@@ -29,7 +29,7 @@ load helpers
 	is "$output" ".*${image} /bin/sh -c" "verify image name"
     else
 	run_ramalama --dryrun run ${model}
-	is "$output" 'llama-simple-chat -m /path/to/model.*' "dryrun correct"
+	is "$output" 'llama-cli -m /path/to/model --in-prefix --in-suffix --no-display-prompt -p.*' "dryrun correct"
 
 	run_ramalama 1 run --name foobar tiny
 	is "${lines[0]}"  "Error: --nocontainer and --name options conflict. --name requires a container." "conflict between nocontainer and --name line"