[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)

* init

* init

* fix format

* add

* add files

* add ut

* fix some

* add ut

* add more

* add

* fix pre-commit

* fix pre-commit

* fix cover

* skip long seq

* add

* add

* fix

* remove not need

* fix set attr

* fix comments

* fix comments

* fix failed tests

---------

Co-authored-by: gongweibao <gognweibao@baidu.com>

tests/input/test_preprocess.py

@@ -0,0 +1,87 @@
+# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Tests for InputPreprocessor.create_processor().
+
+Why mock:
+  - ModelConfig, ReasoningParserManager, ToolParserManager, and concrete processor
+    classes all depend on model files or external resources not available in tests.
+    We mock them at the import boundary to test InputPreprocessor's routing logic.
+"""
+
+import unittest
+from types import SimpleNamespace
+from unittest.mock import MagicMock, patch
+
+
+def _make_model_config(arch, enable_mm=False):
+    cfg = SimpleNamespace(
+        model="test_model",
+        architectures=[arch],
+        enable_mm=enable_mm,
+    )
+    return cfg
+
+
+class TestInputPreprocessorBranching(unittest.TestCase):
+    """Test that create_processor picks the right processor class based on architecture and flags."""
+
+    def test_init_stores_params(self):
+        from fastdeploy.input.preprocess import InputPreprocessor
+
+        config = _make_model_config("LlamaForCausalLM")
+        pp = InputPreprocessor(
+            model_config=config,
+            reasoning_parser="qwen3",
+            tool_parser="ernie_x1",
+            limit_mm_per_prompt={"image": 2},
+        )
+        self.assertEqual(pp.model_name_or_path, "test_model")
+        self.assertEqual(pp.reasoning_parser, "qwen3")
+        self.assertEqual(pp.tool_parser, "ernie_x1")
+        self.assertEqual(pp.limit_mm_per_prompt, {"image": 2})
+
+    def test_create_processor_text_normal_path(self):
+        """Normal path: non-Ernie, non-MM arch creates a text DataProcessor."""
+        from fastdeploy.input.preprocess import InputPreprocessor
+
+        config = _make_model_config("LlamaForCausalLM", enable_mm=False)
+        pp = InputPreprocessor(model_config=config)
+
+        mock_dp = MagicMock()
+        with (
+            patch.dict("sys.modules", {"fastdeploy.plugins": None, "fastdeploy.plugins.input_processor": None}),
+            patch("fastdeploy.input.preprocess.envs") as mock_envs,
+            patch("fastdeploy.input.text_processor.DataProcessor", return_value=mock_dp),
+        ):
+            mock_envs.ENABLE_V1_DATA_PROCESSOR = False
+            pp.create_processor()
+
+        self.assertIs(pp.processor, mock_dp)
+
+    def test_unsupported_mm_arch_raises(self):
+        """When enable_mm=True and arch is unrecognized, should raise ValueError."""
+        from fastdeploy.input.preprocess import InputPreprocessor
+
+        config = _make_model_config("UnknownMMArch", enable_mm=True)
+        pp = InputPreprocessor(model_config=config)
+
+        with patch.dict("sys.modules", {"fastdeploy.plugins": None, "fastdeploy.plugins.input_processor": None}):
+            with self.assertRaises(ValueError):
+                pp.create_processor()
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/input/test_tokenizer_client.py

@@ -99,3 +99,63 @@ async def test_encode_timeout():
 
     with pytest.raises(TimeoutError):
         await client.encode_image(request)
+
+
+@pytest.mark.asyncio
+async def test_encode_invalid_type():
+    """Test invalid encode type raises ValueError (line 130).
+    NOTE: Public methods hardcode the type param, so we test the private method directly
+    to verify the validation boundary."""
+    base_url = "http://testserver"
+    client = AsyncTokenizerClient(base_url=base_url)
+
+    request = ImageEncodeRequest(
+        version="v1", req_id="req_invalid", is_gen=False, resolution=256, image_url="http://example.com/image.jpg"
+    )
+
+    with pytest.raises(ValueError, match="Invalid encode type"):
+        await client._async_encode_request("invalid_type", request.model_dump())
+
+
+@pytest.mark.asyncio
+async def test_decode_invalid_type():
+    """Test invalid decode type raises ValueError (line 186).
+    NOTE: Public methods hardcode the type param, so we test the private method directly
+    to verify the validation boundary."""
+    base_url = "http://testserver"
+    client = AsyncTokenizerClient(base_url=base_url)
+
+    with pytest.raises(ValueError, match="Invalid decode type"):
+        await client._async_decode_request("invalid_type", {})
+
+
+@pytest.mark.asyncio
+@respx.mock
+async def test_encode_network_error_continues_polling():
+    """Test network error during polling is caught and logged (line 164)."""
+    base_url = "http://testserver"
+    client = AsyncTokenizerClient(base_url=base_url, max_wait=2, poll_interval=0.1)
+
+    # Mock create task
+    respx.post(f"{base_url}/image/encode").mock(
+        return_value=httpx.Response(200, json={"code": 0, "task_tag": "task_network_error"})
+    )
+
+    # First poll fails with network error, second succeeds
+    call_count = 0
+
+    def side_effect(request):
+        nonlocal call_count
+        call_count += 1
+        if call_count == 1:
+            raise httpx.RequestError("Network error")
+        return httpx.Response(200, json={"state": "Finished", "result": {"key": "value"}})
+
+    respx.get(f"{base_url}/encode/get").mock(side_effect=side_effect)
+
+    request = ImageEncodeRequest(
+        version="v1", req_id="req_network", is_gen=False, resolution=256, image_url="http://example.com/image.jpg"
+    )
+
+    result = await client.encode_image(request)
+    assert result["key"] == "value"