feat: support multi-language description

pdfix · Nov 6, 2024 · f9bb758 · f9bb758
1 parent fadea41
commit f9bb758
Show file tree

Hide file tree

Showing 4 changed files with 40 additions and 6 deletions.
diff --git a/config.json b/config.json
@@ -10,11 +10,26 @@
       "icon": "add_photo_alternate",
       "category": "Tags",
       "configurations": [
+        {
+          "program": "docker run --platform linux/amd64 -v \"${working_directory}:/data\" -w /data --rm pdfix/alt-text-openai:latest --name \"${license_name}\" --key \"${license_key}\" detect -i \"${input_pdf}\" -o \"${output_pdf}\" --openai \"${openai_key}\" --lang \"${lang}\"",
+          "name": "Alt Text with OpenAI",
+          "desc": "Generate and applies alternative image descriptions to PDF files using PDFix SDK and OpenAI"
+        },
+        {
+          "program": "docker run --platform linux/amd64 -v \"${working_directory}:/data\" -w /data --rm pdfix/alt-text-openai:latest --name \"${license_name}\" --key \"${license_key}\" detect -i \"${input_pdf}\" -o \"${output_pdf}\" --openai \"${openai_key}\" --lang \"${lang}\" --overwrite",
+          "name": "Alt Text with OpenAI",
+          "desc": "Generate and applies alternative image descriptions to PDF files using PDFix SDK and OpenAI"
+        },       
         {
           "program": "docker run --platform linux/amd64 -v \"${working_directory}:/data\" -w /data --rm pdfix/alt-text-openai:latest --name \"${license_name}\" --key \"${license_key}\" detect -i \"${input_pdf}\" -o \"${output_pdf}\" --openai \"${openai_key}\"",
           "name": "Alt Text with OpenAI",
           "desc": "Generate and applies alternative image descriptions to PDF files using PDFix SDK and OpenAI"
-        }   
+        },
+        {
+          "program": "docker run --platform linux/amd64 -v \"${working_directory}:/data\" -w /data --rm pdfix/alt-text-openai:latest --name \"${license_name}\" --key \"${license_key}\" detect -i \"${input_pdf}\" -o \"${output_pdf}\" --openai \"${openai_key}\" --overwrite",
+          "name": "Alt Text with OpenAI",
+          "desc": "Generate and applies alternative image descriptions to PDF files using PDFix SDK and OpenAI"
+        },
       ],
        "args": [
         {
@@ -39,6 +54,13 @@
           "desc": "OpenAI API Key",
           "type": "string",
           "value": ""
+        },
+        {
+          "title": "Alternate description language",
+          "name": "lang",
+          "desc": "Alternate description language",
+          "type": "string",
+          "value": ""
         }
       ]
     }

diff --git a/src/ai.py b/src/ai.py
@@ -9,7 +9,7 @@ def encode_image(image_path: str):
         return base64.b64encode(image_file.read()).decode("utf-8")
 
 
-def alt_description(img_path: str, api_key: str):
+def alt_description(img_path: str, api_key: str, lang: str):
     # Getting the base64 string
     base64_image = encode_image(img_path)
     # print("Using api key: {}".format(api_key))
@@ -23,7 +23,7 @@ def alt_description(img_path: str, api_key: str):
                 "content": [
                     {
                         "type": "text",
-                        "text": "Generate alternate description for the image",
+                        "text": f"Generate alternate description for the image in {lang} language",
                     },
                     {
                         "type": "image_url",

diff --git a/src/main.py b/src/main.py
@@ -64,6 +64,13 @@ def main():
         default=False,
         help="Overwrite alternate text if already present in the tag",
     )
+    pars_detect.add_argument(
+        "--lang",
+        type=str,
+        required=False,
+        default="en",
+        help="Alternate description laguage",
+    )
 
     try:
         args = parser.parse_args()
@@ -100,6 +107,7 @@ def main():
                     args.key,
                     args.openai,
                     args.overwrite,
+                    args.lang,
                 )
                 # print(desc)
             except Exception as e:

diff --git a/src/process_pdf.py b/src/process_pdf.py
@@ -111,6 +111,7 @@ def browse_figure_tags(
     doc: PdfDoc,
     api_key: str,
     overwrite: bool,
+    lang: str,
 ) -> None:
     count = parent.GetNumChildren()
     struct_tree = doc.GetStructTree()
@@ -120,9 +121,9 @@ def browse_figure_tags(
         child_elem = struct_tree.GetStructElementFromObject(parent.GetChildObject(i))
         if child_elem.GetType(True) == "Figure":
             # process figure element
-            update_image_alt(child_elem, doc, api_key, overwrite)
+            update_image_alt(child_elem, doc, api_key, overwrite, lang)
         else:
-            browse_figure_tags(child_elem, doc, api_key, overwrite)
+            browse_figure_tags(child_elem, doc, api_key, overwrite, lang)
 
 
 def alt_text(
@@ -132,6 +133,7 @@ def alt_text(
     license_key: str,
     api_key: str,
     overwrite: bool,
+    lang: str,
 ) -> None:
     """Run OpenAI for alternate text description.
 
@@ -149,6 +151,8 @@ def alt_text(
         OpenAI API key.
     overwrite : bool
         Ovewrite alternate text if already present.
+    lang : str
+        Alternate description language.
 
     """
     pdfix = GetPdfix()
@@ -172,7 +176,7 @@ def alt_text(
 
     child_elem = struct_tree.GetStructElementFromObject(struct_tree.GetChildObject(0))
     try:
-        browse_figure_tags(child_elem, doc, api_key, overwrite)
+        browse_figure_tags(child_elem, doc, api_key, overwrite, lang)
     except Exception as e:
         raise e