[DataProcessor] Remove legacy vl_processor directories and deprecated files (#7422)

* delete code

* add unit test

* fix review

* delete code & fix unit test

* fix unit test

Author: luukunn

docs/usage/code_overview.md

@@ -169,17 +169,17 @@ The main entry file `fastdeploy/__init__.py` exports core classes:
 | File | Function | Development Guide |
 |------|----------|-------------------|
 | `text_processor.py` | `BaseDataProcessor` text processor base class | Input processing extension |
-| `ernie4_5_processor.py` | ERNIE 4.5 input processor | Baidu model input processing |
+| `multimodal_processor.py` | Unified multimodal processor | Multimodal input processing |
 | `ernie4_5_tokenizer.py` | ERNIE 4.5 tokenizer | Tokenization logic modification |
 | `preprocess.py` | Input preprocessing utilities | Preprocessing flow |
 
 **Multimodal Processing Subdirectories**:
 
 | Directory | Function |
 |-----------|----------|
-| `ernie4_5_vl_processor/` | ERNIE 4.5 VL image/video processing |
-| `qwen_vl_processor/` | Qwen VL multimodal processing |
-| `paddleocr_vl_processor/` | PaddleOCR VL processing |
+| `encodings/` | Model-specific encoding strategies (Ernie, Qwen, PaddleOCR) |
+| `image_processors/` | Image preprocessing (Adaptive, Qwen, Qwen3, PaddleOCR) |
+| `multimodal_processor.py` | Unified multimodal processor |
 
 ---
 

docs/zh/usage/code_overview.md

@@ -169,17 +169,17 @@ FastDeploy/
 | 文件 | 功能 | 开发指引 |
 |------|------|----------|
 | `text_processor.py` | `BaseDataProcessor` 文本处理器基类 | 输入处理扩展 |
-| `ernie4_5_processor.py` | ERNIE 4.5 输入处理器 | 百度模型输入处理 |
+| `multimodal_processor.py` | 统一多模态处理器 | 多模态输入处理 |
 | `ernie4_5_tokenizer.py` | ERNIE 4.5 分词器 | 分词逻辑修改 |
 | `preprocess.py` | 输入预处理工具 | 预处理流程 |
 
 **多模态处理子目录**:
 
 | 目录 | 功能 |
 |------|------|
-| `ernie4_5_vl_processor/` | ERNIE 4.5 VL 图像/视频处理 |
-| `qwen_vl_processor/` | Qwen VL 多模态处理 |
-| `paddleocr_vl_processor/` | PaddleOCR VL 处理 |
+| `encodings/` | 模型特定编码策略 (Ernie, Qwen, PaddleOCR) |
+| `image_processors/` | 图像预处理 (Adaptive, Qwen, Qwen3, PaddleOCR) |
+| `multimodal_processor.py` | 统一多模态处理器 |
 
 ---
 

fastdeploy/input/base_processor.py

@@ -15,11 +15,11 @@
 """Abstract base class for all data processors.
 
 Provides unified response-processing logic (ids2tokens, process_response_dict*,
-update_stop_seq, update_bad_words, pad_batch_data, …) extracted from the two
-existing concrete processors:
+update_stop_seq, update_bad_words, pad_batch_data, …) shared by all concrete
+processors:
 
-    DataProcessor      (fastdeploy/input/text_processor.py)
-    Ernie4_5Processor  (fastdeploy/input/ernie4_5_processor.py)
+    TextProcessor       (fastdeploy/input/text_processor.py)
+    MultiModalProcessor (fastdeploy/input/multimodal_processor.py)
 
 Key design decisions
 --------------------
@@ -28,16 +28,12 @@
   of each subclass.  Subclasses that do not call ``super().__init__()`` must
   initialise those three attributes themselves.
 
-* ``process_response_dict`` reads ``stream`` from ``kwargs`` (DataProcessor
-  convention).  Callers that previously passed ``stream`` as a positional
-  argument (ERNIE convention) must be updated to use ``stream=`` keyword.
+* ``process_response_dict`` reads ``stream`` from ``kwargs`` (default: True).
 
-* EOS removal uses ``in self.eos_token_ids`` (list membership).  ERNIE's
-  ``eos_token_ids`` contains exactly one element, so this is equivalent to the
-  ``==`` check it currently uses.
+* EOS removal uses ``in self.eos_token_ids`` (list membership).
 
 * tool_parser result never updates ``outputs["text"]``; only ``tool_calls`` is
-  set.  This matches DataProcessor behaviour.
+  set.
 
 * ``ids2tokens`` always returns a three-tuple
   ``(delta_text, previous_token_ids, previous_texts)``.  The HF-tokeniser