组件进画布：修改处理细节 (baidubce#891)

* LLM-aPaaS-32435 组件进画布和多文件支持

* 修复单测

* 组件进画布，增加异常处理

* 组件进画布：增加细节处理

* 组件进画布：增加细节处理

* Revert "组件进画布：增加细节处理"

This reverts commit c1cb8c6.

---------

Co-authored-by: yepeiwen01 <yepeiwen01@bidu.com>

Author: peiwenYe

python/core/components/v2/general_ocr/component.py

@@ -25,6 +25,7 @@
 from appbuilder.core.component import Component
 from appbuilder.core.components.general_ocr.model import *
 from appbuilder.core.message import Message
+from appbuilder.core.constants import COMPONENT_SUPPORT_FILE_NUMBER
 from appbuilder.utils.trace.tracer_wrapper import components_run_trace, components_run_stream_trace
 
 
@@ -87,7 +88,7 @@ class GeneralOCR(Component):
                         "type": "object",
                         "description": "需要识别的PDF文件的对应页码，key为pdf_names中的文件名，value为对应的页码，当 pdf_file 参数有效时，识别传入页码的对应页面内容，若不传入，则默认识别第 1 页",
                         "additionalProperties": {
-                            "type": "integer"
+                            "type": "string"
                         },
                         "default": {}
                     },
@@ -272,37 +273,71 @@ def tool_eval(
             
         """
         traceid = kwargs.get("_sys_traceid", "")
+        if not img_names and not img_urls and not pdf_names and not pdf_urls:
+            raise InvalidRequestArgumentError(request_id=traceid,
+                                                message="img_names\img_urls\pdf_names\pdf_urls can not both be empty")
         sys_file_urls = kwargs.get("_sys_file_urls", {})
         support_pdf_type = ["pdf"]
         support_img_type = ["png", "jpg", "jpeg", "webp", "heic", "tif", "tiff", "dcm", "mha", "nii.gz"]
         img_map = {}
         pdf_map = {}
+        unsupported_files = []
+        unknown_files = []
         img_names = [os.path.basename(name) for name in img_names]
         pdf_names = [os.path.basename(name) for name in pdf_names]
-        for file_name, file_url in sys_file_urls.items():
-            if file_url in img_urls or file_name in img_names:
-                img_map[file_name] = file_url
-            elif file_name in pdf_names or file_url in pdf_urls:
-                pdf_map[file_name] = {"url": file_url, "page_num": pdf_file_num.get(file_name, "1")}
+
+        for img_name in img_names:
+            if len(img_map) >= COMPONENT_SUPPORT_FILE_NUMBER:
+                break
+            file_type = img_name.split(".")[-1].lower()
+            if img_name in sys_file_urls:
+                if file_type in support_img_type:
+                    img_map[img_name] = sys_file_urls.get(img_name, "")
+                else:
+                    unsupported_files.append(img_name)
             else:
-                file_type = file_name.split(".")[-1].lower()
+                unknown_files.append(img_name)
+
+        for pdf_name in pdf_names:
+            if len(img_map) + len(pdf_map) >= 10:
+                break
+            file_type = pdf_name.split(".")[-1].lower()
+            if pdf_name in sys_file_urls:
                 if file_type in support_pdf_type:
-                    pdf_map[file_name] = {"url": file_url, "page_num": pdf_file_num.get(file_name, "1")}
-                elif file_type in support_img_type:
-                    img_map[file_name] = file_url
+                    pdf_map[pdf_name] = {"url": sys_file_urls.get(pdf_name, ""), "page_num": pdf_file_num.get(pdf_name, "1")}
+                else:
+                    unsupported_files.append(pdf_name)
+            else:
+                unknown_files.append(pdf_name)
 
         for img_url in img_urls:
+            if len(img_map) + len(pdf_map) >= 10:
+                break
+            if img_url in list(sys_file_urls.values()):
+                continue
             file_name = img_url.split("/")[-1].split("?")[0]
             file_type = file_name.split(".")[-1].lower()
-            if file_type in support_img_type and file_name not in img_map:
-                img_map[file_name] = img_url 
+            if file_type in support_img_type:
+                img_map[img_url] = img_url 
+            else:
+                unsupported_files.append(img_url)
 
-        for pdf_url in pdf_urls: 
+        for pdf_url in pdf_urls:
+            if len(img_map) + len(pdf_map) >= 10:
+                break
+            if pdf_url in list(sys_file_urls.values()):
+                continue
             file_name = pdf_url.split("/")[-1].split("?")[0]
             file_type = file_name.split(".")[-1].lower()
-            if file_type in support_pdf_type and file_name not in pdf_map:
-                pdf_map[file_name] = {"url": pdf_url, "page_num": pdf_file_num.get(file_name, "1")}
-    
+            if file_type in support_pdf_type:
+                pdf_map[pdf_url] = {"url": pdf_url, "page_num": pdf_file_num.get(file_name, "1")}
+            else:
+                unsupported_files.append(pdf_url)
+
+        if not img_map and not pdf_map:
+            raise InvalidRequestArgumentError(
+                f"request format error, file url does not exist")
+
         if img_map:
             for img_name, img_url in img_map.items():
                 try:
@@ -313,9 +348,9 @@ def tool_eval(
                     result_response, raw_data = self._recognize(req, request_id=traceid)
                     result = proto.Message.to_dict(result_response)
                     results = {
-                        f"{img_name}识别结果": " \n".join(item["words"] for item in result["words_result"])
+                        f"{img_name}": " \n".join(item["words"] for item in result["words_result"])
                     }
-                    res = json.dumps(results, ensure_ascii=False, indent=4)
+                    res = json.dumps(results, ensure_ascii=False)
                     yield self.create_output(type="text", text=res, raw_data=raw_data, visible_scope="llm")
                     yield self.create_output(type="text", text="", raw_data=raw_data, visible_scope="user")
                 except Exception as e:
@@ -336,15 +371,51 @@ def tool_eval(
                     result_response, raw_data = self._recognize(req, request_id=traceid)
                     result = proto.Message.to_dict(result_response)
                     results = {
-                        f"{pdf_name}识别结果": " \n".join(item["words"] for item in result["words_result"])
+                        f"{pdf_name}": " \n".join(item["words"] for item in result["words_result"])
                     }
-                    res = json.dumps(results, ensure_ascii=False, indent=4)
+                    res = json.dumps(results, ensure_ascii=False)
                     yield self.create_output(type="text", text=res, raw_data=raw_data, visible_scope="llm")
                     yield self.create_output(type="text", text="", raw_data=raw_data, visible_scope="user")
                 except Exception as e:
                     logging.warning(f"{pdf_name} ocr failed with exception: {e}")
                     continue
-
-        if not img_map and not pdf_map:
-            raise InvalidRequestArgumentError(
-                f"request format error, file url does not exist")
+        
+        for file_name in unknown_files:
+            results = {
+                f"{file_name}": "无法获取url，请确认是否上传成功"
+            }
+            res = json.dumps(results, ensure_ascii=False)
+            llm_result = self.create_output(
+                type = "text",
+                visible_scope= "llm",
+                text=res,
+                name="llm_text"
+            )
+            yield llm_result
+            user_result = self.create_output(
+                type = "text",
+                visible_scope= "user",
+                text="",
+                name="user_text"
+            )
+            yield user_result
+        
+        for file_name in unsupported_files:
+            results = {
+                f"{file_name}": "不支持的文件类型，请确认是否为图片或者pdf文件"
+            }
+            res = json.dumps(results, ensure_ascii=False)
+            llm_result = self.create_output(
+                type = "text",
+                visible_scope= "llm",
+                text=res,
+                name="llm_text"
+            )
+            yield llm_result
+            user_result = self.create_output(
+                type = "text",
+                visible_scope= "user",
+                text="",
+                name="user_text"
+            )
+            yield user_result

python/core/components/v2/handwrite_ocr/component.py

@@ -12,6 +12,7 @@
 
 r"""手写文字识别组件"""
 import base64
+import json
 import logging
 from typing import Optional
 from appbuilder.core._exception import AppBuilderServerException, InvalidRequestArgumentError
@@ -20,6 +21,7 @@
 from appbuilder.core.message import Message
 from appbuilder.core._client import HTTPClient
 from appbuilder.core import utils
+from appbuilder.core.constants import COMPONENT_SUPPORT_FILE_NUMBER
 from appbuilder.utils.trace.tracer_wrapper import components_run_trace, components_run_stream_trace
 
 class HandwriteOCR(Component):
@@ -140,27 +142,41 @@ def tool_eval(self,
         Yields:
             Generator[Output]: 生成器，每次迭代产生一个输出对象
         """
+        if not file_names and not file_urls:
+            raise InvalidRequestArgumentError(request_id=traceid,
+                                                message="file_names and file_urls cannot both be empty")
         supported_file_type = ["png", "jpg", "jpeg", "webp", "heic", "tif", "tiff", "dcm", "mha", "nii.gz"]
         traceid = kwargs.get("_sys_traceid", "")
-        result = ""
 
         sys_file_urls = kwargs.get('_sys_file_urls', {})
         available_img_urls = {}
-        for file_name, file_url in sys_file_urls.items():
-            file_type = file_name.split(".")[-1].lower()
-            if file_type in supported_file_type:
-                available_img_urls[file_name] = file_url
-        
+        unsupported_files  = []
+        unknown_file_name = []
+
         if file_names:
             for file_name in file_names:
+                if len(available_img_urls) >= COMPONENT_SUPPORT_FILE_NUMBER:
+                    break
+                file_type = file_name.split(".")[-1].lower()
                 if file_name in sys_file_urls:
-                    available_img_urls[file_name] = sys_file_urls.get(file_name, "")
-                    
+                    if file_type in supported_file_type:
+                        available_img_urls[file_name] = sys_file_urls.get(file_name, "")
+                    else: #不支持的文件类型
+                        unsupported_files.append(file_name)
+                else: #url未知
+                    unknown_file_name.append(file_name)
+
         for img_url in file_urls:
+            if len(available_img_urls) >= COMPONENT_SUPPORT_FILE_NUMBER:
+                break
+            if img_url in list(sys_file_urls.values()):  #只考虑用户手动传的file_url
+                continue
             file_name = img_url.split("/")[-1].split("?")[0]
             file_type = file_name.split(".")[-1].lower()
-            if file_type in supported_file_type and file_name not in available_img_urls :
-                available_img_urls[file_name] = img_url
+            if file_type in supported_file_type:
+                available_img_urls[img_url] = img_url
+            else:   #不支持的文件类型
+                unsupported_files.append(img_url)
 
         for file_name, file_url in available_img_urls.items():
             try:
@@ -172,12 +188,14 @@ def tool_eval(self,
                 req.detect_alteration = "true"
                 response = self._recognize(req, request_id=traceid)
                 text = "".join([w.words for w in response.words_result])
-                result = f"{file_name}的手写识别结果是：{text} "
-
+                results = {
+                    file_name: text
+                }
+                res = json.dumps(results, ensure_ascii=False)
                 llm_result = self.create_output(
                     type = "text",
                     visible_scope= "llm",
-                    text=result,
+                    text=res,
                     name="llm_text"
                 )
                 yield llm_result
@@ -192,6 +210,46 @@ def tool_eval(self,
             except Exception as e:
                 logging.warning(f"{file_name} ocr failed with exception: {e}")
                 continue
+        
+        for file_name in unknown_file_name:
+            results = {
+                file_name: "无法获取url，请确认是否上传成功"
+            }
+            res = json.dumps(results, ensure_ascii=False)
+            llm_result = self.create_output(
+                type = "text",
+                visible_scope= "llm",
+                text=res,
+                name="llm_text"
+            )
+            yield llm_result
+            user_result = self.create_output(
+                type = "text",
+                visible_scope= "user",
+                text="",
+                name="user_text"
+            )
+            yield user_result
+        
+        for file_name in unsupported_files:
+            results = {
+                file_name: "不支持的文件类型，请确认是否为图片"
+            }
+            res = json.dumps(results, ensure_ascii=False)
+            llm_result = self.create_output(
+                type = "text",
+                visible_scope= "llm",
+                text=res,
+                name="llm_text"
+            )
+            yield llm_result
+            user_result = self.create_output(
+                type = "text",
+                visible_scope= "user",
+                text="",
+                name="user_text"
+            )
+            yield user_result
 
     def _recognize(
         self,