Modified paths to be OS-independent

Author: david-yz-liu

CHANGELOG.md

@@ -3,3 +3,4 @@
 ## [unreleased]
 
 - Modified project structure and configuration to use uv
+- Modified paths to be OS-independent

src/markus_ai_server/server.py

@@ -6,6 +6,7 @@
 import logging
 import os
 import subprocess
+from pathlib import Path
 from typing import Optional
 
 import ollama
@@ -28,8 +29,8 @@
 DEFAULT_MODEL = os.getenv('DEFAULT_MODEL', 'deepseek-coder-v2:latest')
 
 # Llama.cpp configuration
-LLAMA_CPP_CLI = os.getenv('LLAMA_CPP_CLI', '/data1/llama.cpp/bin/llama-cli')
-GGUF_DIR = os.getenv('GGUF_DIR', '/data1/GGUF')
+LLAMA_CPP_CLI = os.getenv('LLAMA_CPP_CLI', str(Path('/data1/llama.cpp/bin/llama-cli')))
+GGUF_DIR = os.getenv('GGUF_DIR', str(Path('/data1/GGUF')))
 
 # Llama server configuration
 _llama_server_url = os.getenv('LLAMA_SERVER_URL')  # e.g., http://localhost:8080 or localhost:8080

test/test_cli_mode.py

@@ -1,5 +1,6 @@
 import os
 import subprocess
+from pathlib import Path
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -46,7 +47,7 @@ class TestLlamaCppCLI:
 
     def test_chat_with_llamacpp_success(self, mock_resolve_model_path, mock_subprocess):
         """Test successful CLI chat with llama.cpp."""
-        model_path = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+        model_path = str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'))
         mock_resolve_model_path.return_value = model_path
 
         mock_result = MagicMock()
@@ -61,7 +62,7 @@ def test_chat_with_llamacpp_success(self, mock_resolve_model_path, mock_subproce
         # Verify correct CLI command structure
         args, kwargs = mock_subprocess.call_args
         cmd = args[0]
-        assert '/data1/llama.cpp/bin/llama-cli' in cmd
+        assert str(Path('/data1/llama.cpp/bin/llama-cli')) in cmd
         assert '-m' in cmd and model_path in cmd
         assert '--n-gpu-layers' in cmd and '40' in cmd
         assert '--single-turn' in cmd
@@ -75,7 +76,9 @@ def test_chat_with_llamacpp_model_not_found(self, mock_resolve_model_path):
 
     def test_chat_with_llamacpp_subprocess_error(self, mock_resolve_model_path, mock_subprocess):
         """Test CLI chat when subprocess fails."""
-        mock_resolve_model_path.return_value = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+        mock_resolve_model_path.return_value = str(
+            Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf')
+        )
 
         error = subprocess.CalledProcessError(1, 'cmd')
         error.stderr = b'CUDA out of memory'
@@ -91,9 +94,11 @@ class TestCLIModeRouting:
     @pytest.fixture(autouse=True)
     def setup_routing_mocks(self):
         """Set up common mocks for routing tests."""
-        with patch('markus_ai_server.server.chat_with_llamacpp') as mock_chat_llamacpp, patch(
-            'markus_ai_server.server.is_llamacpp_available'
-        ) as mock_available, patch('markus_ai_server.server.chat_with_ollama') as mock_chat_ollama:
+        with (
+            patch('markus_ai_server.server.chat_with_llamacpp') as mock_chat_llamacpp,
+            patch('markus_ai_server.server.is_llamacpp_available') as mock_available,
+            patch('markus_ai_server.server.chat_with_ollama') as mock_chat_ollama,
+        ):
             self.mock_chat_llamacpp = mock_chat_llamacpp
             self.mock_available = mock_available
             self.mock_chat_ollama = mock_chat_ollama
@@ -177,7 +182,7 @@ class TestCLIModeIntegration:
 
     def test_complete_cli_flow_with_real_model(self, mock_glob, mock_subprocess):
         """Test complete CLI flow: model resolution → CLI execution."""
-        model_path = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+        model_path = str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'))
 
         mock_glob.return_value = [model_path]
         mock_result = MagicMock()
@@ -204,7 +209,7 @@ def test_complete_cli_fallback_flow_to_ollama(self, mock_glob, mock_ollama):
         result = chat_with_model(TEST_OLLAMA_MODEL, 'Fallback test', llama_mode='cli')
 
         assert result == "Ollama CLI fallback integration test successful!"
-        mock_glob.assert_called_once_with(f'/data1/GGUF/{TEST_OLLAMA_MODEL}/*.gguf')
+        mock_glob.assert_called_once_with(str(Path(f'/data1/GGUF/{TEST_OLLAMA_MODEL}/*.gguf')))
         mock_ollama.assert_called_once()
 
     def test_cli_mode_passes_json_schema_to_ollama(self, tmp_path):
@@ -215,9 +220,10 @@ def test_cli_mode_passes_json_schema_to_ollama(self, tmp_path):
         test_schema = {"schema": {"type": "object", "properties": {"answer": {"type": "string"}}}}
 
         #  Prepare mocks
-        with patch('markus_ai_server.server.is_llamacpp_available', return_value=False), patch(
-            'markus_ai_server.server.chat_with_ollama'
-        ) as mock_ollama:
+        with (
+            patch('markus_ai_server.server.is_llamacpp_available', return_value=False),
+            patch('markus_ai_server.server.chat_with_ollama') as mock_ollama,
+        ):
             mock_ollama.return_value = "schema-aware response"
 
             result = chat_with_model(TEST_OLLAMA_MODEL, "Give me an answer", llama_mode='cli', json_schema=test_schema)

test/test_core.py

@@ -1,4 +1,5 @@
 import os
+from pathlib import Path
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -35,13 +36,13 @@ class TestModelResolution:
 
     def test_resolve_model_path_found(self, mock_glob):
         """Test model path resolution when model exists."""
-        model_path = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+        model_path = str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'))
         mock_glob.return_value = [model_path]
 
         result = resolve_model_path(TEST_LLAMACPP_MODEL)
 
         assert result == model_path
-        mock_glob.assert_called_once_with(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/*.gguf')
+        mock_glob.assert_called_once_with(str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/*.gguf')))
 
     def test_resolve_model_path_not_found(self, mock_glob):
         """Test model path resolution when model doesn't exist."""
@@ -54,7 +55,7 @@ def test_resolve_model_path_not_found(self, mock_glob):
     def test_is_llamacpp_available_true(self):
         """Test model availability check when model exists."""
         with patch('markus_ai_server.server.resolve_model_path') as mock_resolve:
-            mock_resolve.return_value = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+            mock_resolve.return_value = str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'))
 
             result = is_llamacpp_available(TEST_LLAMACPP_MODEL)
 

test/test_server_mode.py

@@ -1,4 +1,5 @@
 import os
+from pathlib import Path
 from unittest.mock import MagicMock, patch
 
 import pytest
@@ -92,10 +93,11 @@ class TestServerModeRouting:
     @pytest.fixture(autouse=True)
     def setup_routing_mocks(self):
         """Set up common mocks for routing tests."""
-        with patch('markus_ai_server.server.chat_with_llama_server_http') as mock_chat_server, patch(
-            'markus_ai_server.server.is_llamacpp_available'
-        ) as mock_available, patch('markus_ai_server.server.chat_with_ollama') as mock_chat_ollama, patch(
-            'markus_ai_server.server.LLAMA_SERVER_URL', 'http://localhost:8080'
+        with (
+            patch('markus_ai_server.server.chat_with_llama_server_http') as mock_chat_server,
+            patch('markus_ai_server.server.is_llamacpp_available') as mock_available,
+            patch('markus_ai_server.server.chat_with_ollama') as mock_chat_ollama,
+            patch('markus_ai_server.server.LLAMA_SERVER_URL', 'http://localhost:8080'),
         ):
             self.mock_chat_server = mock_chat_server
             self.mock_available = mock_available
@@ -177,7 +179,7 @@ class TestServerModeIntegration:
 
     def test_complete_server_flow_with_real_model(self, mock_glob, mock_requests_post, mock_llama_server_url):
         """Test complete server flow: model resolution → HTTP API call."""
-        model_path = f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'
+        model_path = str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/{TEST_LLAMACPP_MODEL}.gguf'))
 
         # Mock model found (only checked once for availability in server mode)
         mock_glob.return_value = [model_path]
@@ -192,7 +194,7 @@ def test_complete_server_flow_with_real_model(self, mock_glob, mock_requests_pos
 
         assert result == "Server integration test successful!"
         # In server mode, glob.glob only called once for is_llamacpp_available
-        mock_glob.assert_called_once_with(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/*.gguf')
+        mock_glob.assert_called_once_with(str(Path(f'/data1/GGUF/{TEST_LLAMACPP_MODEL}/*.gguf')))
         mock_requests_post.assert_called_once()
 
     def test_complete_server_fallback_flow_to_ollama(self, mock_glob, mock_ollama, mock_llama_server_url):
@@ -208,7 +210,7 @@ def test_complete_server_fallback_flow_to_ollama(self, mock_glob, mock_ollama, m
         result = chat_with_model(TEST_OLLAMA_MODEL, 'Fallback test', llama_mode='server')
 
         assert result == "Ollama server fallback integration test successful!"
-        mock_glob.assert_called_once_with(f'/data1/GGUF/{TEST_OLLAMA_MODEL}/*.gguf')
+        mock_glob.assert_called_once_with(str(Path(f'/data1/GGUF/{TEST_OLLAMA_MODEL}/*.gguf')))
         mock_ollama.assert_called_once()
 
     def test_server_mode_passes_json_schema_to_llama_server(self, tmp_path, mock_requests_post, mock_llama_server_url):