File size: 13,637 Bytes
c2f9396
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
import pytest
import asyncio
from unittest.mock import Mock, patch, AsyncMock
from app.models import ChatMessage, ChatRequest
from app.llm_manager import LLMManager


class TestLLMManager:
    """Test the LLM manager functionality."""

    @pytest.fixture
    def llm_manager(self):
        """Create a fresh LLM manager instance for each test."""
        return LLMManager()

    @pytest.fixture
    def sample_request(self):
        """Create a sample chat request."""
        messages = [
            ChatMessage(role="system", content="You are helpful."),
            ChatMessage(role="user", content="Hello!"),
        ]
        return ChatRequest(messages=messages, max_tokens=50)

    def test_initialization(self, llm_manager):
        """Test LLM manager initialization."""
        assert llm_manager.model_path is not None
        assert llm_manager.model is None
        assert llm_manager.tokenizer is None
        assert llm_manager.model_type == "llama_cpp"
        assert llm_manager.context_window == 2048
        assert llm_manager.is_loaded is False
        assert len(llm_manager.mock_responses) > 0

    def test_custom_model_path(self):
        """Test LLM manager with custom model path."""
        custom_path = "/custom/path/model.gguf"
        llm_manager = LLMManager(model_path=custom_path)
        assert llm_manager.model_path == custom_path

    @pytest.mark.asyncio
    async def test_load_model_mock_fallback(self, llm_manager):
        """Test model loading falls back to mock when no models available."""
        with patch("app.llm_manager.LLAMA_AVAILABLE", False):
            with patch("app.llm_manager.TRANSFORMERS_AVAILABLE", False):
                with patch("app.llm_manager.Path") as mock_path:
                    mock_path.return_value.exists.return_value = False
                    success = await llm_manager.load_model()
                    assert success is True
                    assert llm_manager.is_loaded is True
                    assert llm_manager.model_type == "mock"

    @pytest.mark.asyncio
    async def test_load_llama_model(self, llm_manager):
        """Test loading model with llama-cpp-python."""
        mock_llama = Mock()

        with patch("app.llm_manager.LLAMA_AVAILABLE", True):
            with patch("app.llm_manager.Path") as mock_path:
                mock_path.return_value.exists.return_value = True
                with patch("app.llm_manager.Llama", return_value=mock_llama):
                    with patch("os.cpu_count", return_value=4):
                        success = await llm_manager.load_model()

                        assert success is True
                        assert llm_manager.is_loaded is True
                        assert llm_manager.model_type == "llama_cpp"
                        assert llm_manager.model == mock_llama

    @pytest.mark.asyncio
    async def test_load_transformers_model(self, llm_manager):
        """Test loading model with transformers."""
        mock_tokenizer = Mock()
        mock_model = Mock()

        with patch("app.llm_manager.LLAMA_AVAILABLE", False):
            with patch("app.llm_manager.TRANSFORMERS_AVAILABLE", True):
                with patch(
                    "app.llm_manager.AutoTokenizer.from_pretrained",
                    return_value=mock_tokenizer,
                ):
                    with patch(
                        "app.llm_manager.AutoModelForCausalLM.from_pretrained",
                        return_value=mock_model,
                    ):
                        with patch(
                            "app.llm_manager.torch.cuda.is_available",
                            return_value=False,
                        ):
                            success = await llm_manager.load_model()

                            assert success is True
                            assert llm_manager.is_loaded is True
                            assert llm_manager.model_type == "transformers"
                            assert llm_manager.tokenizer == mock_tokenizer
                            assert llm_manager.model == mock_model

    @pytest.mark.asyncio
    async def test_load_model_failure(self, llm_manager):
        """Test model loading failure handling."""
        with patch("app.llm_manager.LLAMA_AVAILABLE", False):
            with patch("app.llm_manager.TRANSFORMERS_AVAILABLE", False):
                with patch("app.llm_manager.Path") as mock_path:
                    mock_path.return_value.exists.return_value = False
                    # Force an exception in the mock fallback
                    with patch.object(
                        llm_manager,
                        "_load_transformers_model",
                        side_effect=Exception("Load failed"),
                    ):
                        success = await llm_manager.load_model()
                        assert (
                            success is True
                        )  # Should still succeed with mock fallback
                        assert llm_manager.is_loaded is True

    def test_format_messages(self, llm_manager):
        """Test message formatting."""
        messages = [
            ChatMessage(role="system", content="You are helpful."),
            ChatMessage(role="user", content="Hello!"),
        ]

        result = llm_manager.format_messages(messages)
        expected = "<|system|>\nYou are helpful.\n<|/system|>\n<|user|>\nHello!\n<|/user|>\n<|assistant|>"
        assert result == expected

    def test_truncate_context_no_tokenizer(self, llm_manager):
        """Test context truncation when no tokenizer is available."""
        prompt = "This is a test prompt"
        result = llm_manager.truncate_context(prompt, 100)
        assert result == prompt

    def test_truncate_context_with_tokenizer(self, llm_manager):
        """Test context truncation with tokenizer."""
        mock_tokenizer = Mock()
        mock_tokenizer.encode.return_value = [1, 2, 3, 4, 5] * 500  # Long token list
        mock_tokenizer.decode.return_value = "truncated prompt"
        llm_manager.tokenizer = mock_tokenizer

        prompt = "This is a test prompt"
        result = llm_manager.truncate_context(prompt, 100)

        assert result == "truncated prompt"
        mock_tokenizer.encode.assert_called_once_with(prompt)

    @pytest.mark.asyncio
    async def test_generate_stream_not_loaded(self, llm_manager, sample_request):
        """Test that generate_stream raises error when model not loaded."""
        with pytest.raises(RuntimeError, match="Model not loaded"):
            async for _ in llm_manager.generate_stream(sample_request):
                pass

    @pytest.mark.asyncio
    async def test_generate_mock_stream(self, llm_manager, sample_request):
        """Test mock streaming generation."""
        llm_manager.is_loaded = True
        llm_manager.model_type = "mock"

        chunks = []
        async for chunk in llm_manager.generate_stream(sample_request):
            chunks.append(chunk)

        # Should have multiple chunks (words) plus completion signal
        assert len(chunks) > 1

        # Check structure of chunks
        for chunk in chunks[:-1]:  # All except last
            assert "id" in chunk
            assert "object" in chunk
            assert chunk["object"] == "chat.completion.chunk"
            assert "choices" in chunk
            assert len(chunk["choices"]) == 1
            assert "delta" in chunk["choices"][0]
            assert "content" in chunk["choices"][0]["delta"]

        # Check completion signal
        last_chunk = chunks[-1]
        assert last_chunk["choices"][0]["finish_reason"] == "stop"

    @pytest.mark.asyncio
    async def test_generate_llama_stream(self, llm_manager, sample_request):
        """Test llama-cpp streaming generation."""
        llm_manager.is_loaded = True
        llm_manager.model_type = "llama_cpp"
        llm_manager.model = Mock()

        # Mock llama response
        mock_response = [
            {"choices": [{"delta": {"content": "Hello"}, "finish_reason": None}]},
            {"choices": [{"delta": {"content": " world"}, "finish_reason": None}]},
            {"choices": [{"delta": {}, "finish_reason": "stop"}]},
        ]
        llm_manager.model.return_value = mock_response

        chunks = []
        async for chunk in llm_manager.generate_stream(sample_request):
            chunks.append(chunk)

        # Should have chunks for each token plus completion
        assert len(chunks) >= 2

        # Check that llama model was called correctly
        llm_manager.model.assert_called_once()
        call_args = llm_manager.model.call_args
        assert call_args[1]["stream"] is True
        assert call_args[1]["max_tokens"] == 50

    @pytest.mark.asyncio
    async def test_generate_transformers_stream(self, llm_manager, sample_request):
        """Test transformers streaming generation."""
        llm_manager.is_loaded = True
        llm_manager.model_type = "transformers"
        llm_manager.tokenizer = Mock()
        llm_manager.model = Mock()

        # Mock tokenizer and model
        llm_manager.tokenizer.encode.return_value = [1, 2, 3]
        llm_manager.tokenizer.decode.return_value = "test"
        llm_manager.tokenizer.eos_token_id = 0

        mock_tensor = Mock()
        mock_tensor.unsqueeze.return_value = mock_tensor
        llm_manager.model.generate.return_value = mock_tensor

        with patch("app.llm_manager.torch") as mock_torch:
            mock_torch.cuda.is_available.return_value = False
            mock_torch.cat.return_value = mock_tensor

            chunks = []
            async for chunk in llm_manager.generate_stream(sample_request):
                chunks.append(chunk)
                if len(chunks) >= 3:  # Limit to avoid infinite loop
                    break

            # Should have some chunks
            assert len(chunks) > 0

    @pytest.mark.asyncio
    async def test_generate_stream_error_handling(self, llm_manager, sample_request):
        """Test error handling in streaming generation."""
        llm_manager.is_loaded = True
        llm_manager.model_type = "llama_cpp"
        llm_manager.model = Mock()

        # Mock llama to raise exception
        llm_manager.model.side_effect = Exception("Generation failed")

        chunks = []
        async for chunk in llm_manager.generate_stream(sample_request):
            chunks.append(chunk)

        # Should have error chunk
        assert len(chunks) == 1
        assert "error" in chunks[0]
        assert chunks[0]["error"]["type"] == "generation_error"

    def test_get_model_info(self, llm_manager):
        """Test getting model information."""
        llm_manager.is_loaded = True
        llm_manager.model_type = "llama_cpp"

        info = llm_manager.get_model_info()

        assert info["id"] == "llama-2-7b-chat"
        assert info["object"] == "model"
        assert info["owned_by"] == "huggingface"
        assert info["type"] == "llama_cpp"
        assert info["context_window"] == 2048
        assert info["is_loaded"] is True

    def test_get_model_info_not_loaded(self, llm_manager):
        """Test getting model info when not loaded."""
        info = llm_manager.get_model_info()
        assert info["is_loaded"] is False


class TestLLMManagerIntegration:
    """Integration tests for LLM manager."""

    @pytest.mark.asyncio
    async def test_full_workflow_mock(self):
        """Test full workflow with mock model."""
        llm_manager = LLMManager()

        # Force mock mode
        llm_manager.is_loaded = True
        llm_manager.model_type = "mock"

        # Create request
        messages = [ChatMessage(role="user", content="Hello, how are you?")]
        request = ChatRequest(messages=messages, max_tokens=20)

        # Generate response
        chunks = []
        async for chunk in llm_manager.generate_stream(request):
            chunks.append(chunk)

        # Verify response
        assert len(chunks) > 1
        assert all("choices" in chunk for chunk in chunks[:-1])
        assert chunks[-1]["choices"][0]["finish_reason"] == "stop"

    @pytest.mark.asyncio
    async def test_context_truncation_integration(self):
        """Test context truncation in full workflow."""
        llm_manager = LLMManager()
        await llm_manager.load_model()

        # Create very long messages
        long_message = "x" * 10000
        messages = [
            ChatMessage(role="system", content="You are helpful."),
            ChatMessage(role="user", content=long_message),
            ChatMessage(role="assistant", content=long_message),
            ChatMessage(role="user", content="Short message"),
        ]

        request = ChatRequest(messages=messages, max_tokens=50)

        # Should not raise exception due to truncation
        chunks = []
        async for chunk in llm_manager.generate_stream(request):
            chunks.append(chunk)

        assert len(chunks) > 0

    @pytest.mark.asyncio
    async def test_different_model_types(self):
        """Test different model type configurations."""
        llm_manager = LLMManager()

        # Test llama_cpp type
        llm_manager.model_type = "llama_cpp"
        info = llm_manager.get_model_info()
        assert info["type"] == "llama_cpp"

        # Test transformers type
        llm_manager.model_type = "transformers"
        info = llm_manager.get_model_info()
        assert info["type"] == "transformers"

        # Test mock type
        llm_manager.model_type = "mock"
        info = llm_manager.get_model_info()
        assert info["type"] == "mock"