dwani 0.1.14__py3-none-any.whl → 0.1.17__py3-none-any.whl

dwani/__init__.py

@@ -53,8 +53,12 @@ class translate:
 
 class document:
     @staticmethod
-    def run_ocr(file_path, language="eng_Latn", model="gemma3"):
-        return _get_client().document_ocr(file_path, language, model)
+    def run_ocr_number(file_path, page_number=1, model="gemma3"):
+        return _get_client().document_ocr_number(file_path, page_number, model)
+    
+    @staticmethod
+    def run_ocr_all(file_path, model="gemma3"):
+        return _get_client().document_ocr_all(file_path, model)
     
     @staticmethod
     def run_summarize(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):

dwani/client.py

@@ -43,9 +43,13 @@ class DwaniClient:
         from .asr import asr_transcribe
         return asr_transcribe(self, file_path=file_path, language=language)
     
-    def document_ocr(self, file_path, language=None, model="gemma3"):
-        from .docs import document_ocr
-        return document_ocr(self, file_path=file_path, language=language, model=model)
+    def document_ocr_number(self, file_path, page_number=1,model="gemma3"):
+        from .docs import document_ocr_number
+        return document_ocr_number(self, file_path=file_path, page_number=page_number, model=model)
+
+    def document_ocr_all(self, file_path,model="gemma3"):
+        from .docs import document_ocr_all
+        return document_ocr_all(self, file_path=file_path, model=model)
 
     def document_summarize(self, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
         from .docs import document_summarize

dwani/docs.py

@@ -40,21 +40,19 @@ def validate_model(model):
         raise ValueError(f"Unsupported model: {model}. Supported models: {VALID_MODELS}")
     return model
 
-def document_ocr(client, file_path, language=None, model="gemma3"):
+def document_ocr_all(client, file_path, model="gemma3"):
     """OCR a document (image/PDF) and return extracted text."""
-    logger.debug(f"Calling document_ocr: file_path={file_path}, language={language}, model={model}")
+    logger.debug(f"Calling document_ocr: file_path={file_path}, model={model}")
     validate_model(model)
     
     data = {"model": model}
-    if language:
-        data["language"] = normalize_language(language)
-    
+
     with open(file_path, "rb") as f:
         mime_type = "application/pdf" if file_path.lower().endswith('.pdf') else "image/png"
         files = {"file": (file_path, f, mime_type)}
         try:
             resp = requests.post(
-                f"{client.api_base}/v1/document/ocr",
+                f"{client.api_base}/v1/extract-text-all",
                 headers=client._headers(),
                 files=files,
                 data=data,
@@ -68,6 +66,33 @@ def document_ocr(client, file_path, language=None, model="gemma3"):
     logger.debug(f"OCR response: {resp.status_code}")
     return resp.json()
 
+
+def document_ocr_number(client, file_path, page_number=1, model="gemma3"):
+    """OCR a document (image/PDF) and return extracted text."""
+    logger.debug(f"Calling document_ocr: file_path={file_path}, model={model}")
+    validate_model(model)
+    
+    data = {"model": model,
+            "page_number": str(page_number)}
+
+    with open(file_path, "rb") as f:
+        mime_type = "application/pdf" if file_path.lower().endswith('.pdf') else "image/png"
+        files = {"file": (file_path, f, mime_type)}
+        try:
+            resp = requests.post(
+                f"{client.api_base}/v1/extract-text",
+                headers=client._headers(),
+                files=files,
+                data=data,
+                timeout=60
+            )
+            resp.raise_for_status()
+        except requests.RequestException as e:
+            logger.error(f"OCR request failed: {str(e)}")
+            raise DwaniAPIError(resp) if 'resp' in locals() else DwaniAPIError.from_exception(e)
+    
+    logger.debug(f"OCR response: {resp.status_code}")
+    return resp.json()
 def document_summarize(client, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
     """Summarize a PDF document with language and page number options."""
     logger.debug(f"Calling document_summarize: file_path={file_path}, page_number={page_number}, src_lang={src_lang}, tgt_lang={tgt_lang}, model={model}")
@@ -256,10 +281,15 @@ def doc_query_kannada(
 
 class Documents:
     @staticmethod
-    def ocr(file_path, language=None, model="gemma3"):
+    def run_ocr_number(file_path, page_number=1,model="gemma3"):
+        from .client import DwaniClient
+        client = DwaniClient()
+        return document_ocr_number(client, file_path, page_number=page_number, model=model)
+    @staticmethod
+    def run_ocr_all(file_path, model="gemma3"):
         from .client import DwaniClient
         client = DwaniClient()
-        return document_ocr(client, file_path, language, model)
+        return document_ocr_all(client, file_path, model)
     
     @staticmethod
     def summarize(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):

{dwani-0.1.14.dist-info → dwani-0.1.17.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dwani
-Version: 0.1.14
+Version: 0.1.17
 Summary: Multimodal API for Indian + European languages (Chat, Vision, TTS, ASR, Translate, Docs)
 Author-email: sachin <python@dwani.ai>
 License: MIT License
@@ -64,13 +64,26 @@ dwani.api_key = os.getenv("DWANI_API_KEY")
 dwani.api_base = os.getenv("DWANI_API_BASE_URL")
 ```
 
-- Check examples folder for detailed use cases
+
+---
+
+- Source Code : [https://github.com/dwani-ai/dwani-python-sdk](https://github.com/dwani-ai/dwani-python-sdk)
+- Check examples folder for detailed use cases 
+
   - [examples/chat.py](examples/chat.py)
   - [examples/vision.py](examples/vision.py)
   - [examples/docs.py](examples/docs.py)
   - [examples/speech.py](examples/speech.py)
   - [examples/asr.py](examples/asr.py)
 
+#### Document - OCR
+```python
+result = dwani.Documents.run_ocr_number(file_path="dwani-workshop.pdf", page_number=1, model="gemma3")
+print(result)
+```
+```json
+{'page_content': "Here's the plain text extracted from the image:\n\ndwani's Goals\n\nTo integrate and enhance the following models and services for Kannada:\n\n*   **Automatic Speech Recognition (ASR):**"}
+```
 
 
 ### Text Query 

dwani-0.1.17.dist-info/RECORD

@@ -0,0 +1,14 @@
+dwani/__init__.py,sha256=TCAqgbvZjztYG0qzFQfafUG9R8mf1bqNWSdoVeopR1M,3186
+dwani/asr.py,sha256=BAdqivQd57NJZX1dSY-J6EFi8TDdyuhf_AyCPcQ0M7w,1719
+dwani/audio.py,sha256=CFQrYU-KLwO7pCh_R7c1SSDJ6bugE5_av7lV8XTl-dY,936
+dwani/chat.py,sha256=Tui52XBhUyDyN2rOFoLme4oB0Q8fkD9_0tFDAnRzoaU,2979
+dwani/client.py,sha256=xV6TpzMV9bR9goPg1tnrDopfxN_N7e-7W3MAUPwCNVs,3594
+dwani/docs.py,sha256=uEYFaR9U8gEtvzr9Ke80Z1T5DwYLom6uk22wBOsVAtQ,11794
+dwani/exceptions.py,sha256=n06dPmR20rS4T3sJBWHQhGxzg4SJKzird9Hx0YTwwo0,226
+dwani/translate.py,sha256=c03N8-tN49IBcTA6GMOkrJ3MaVzZ12RnYdLQwRbEeoQ,2794
+dwani/vision.py,sha256=FviGewoV936CSv_K-latw0t3ZhSSCOF5LaGaq1oE4uA,3607
+dwani-0.1.17.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
+dwani-0.1.17.dist-info/METADATA,sha256=4rKh-XX9JYElwkli9wUrh5xUmrC6fRXN7Q934bggjKA,6030
+dwani-0.1.17.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+dwani-0.1.17.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
+dwani-0.1.17.dist-info/RECORD,,

dwani-0.1.14.dist-info/RECORD

@@ -1,14 +0,0 @@
-dwani/__init__.py,sha256=rV93n_1nOed4ak7NOcwtDjxsT7JbDIivhtRRcrDR80A,3040
-dwani/asr.py,sha256=BAdqivQd57NJZX1dSY-J6EFi8TDdyuhf_AyCPcQ0M7w,1719
-dwani/audio.py,sha256=CFQrYU-KLwO7pCh_R7c1SSDJ6bugE5_av7lV8XTl-dY,936
-dwani/chat.py,sha256=Tui52XBhUyDyN2rOFoLme4oB0Q8fkD9_0tFDAnRzoaU,2979
-dwani/client.py,sha256=fGtnY9L0u_mymY7xUeFDhCBX7P7fd4WNko7WTDgrDV4,3394
-dwani/docs.py,sha256=Cp0Gtudug79GH25toB-Npl35ZFA0TM32oZF2xH1VmNY,10598
-dwani/exceptions.py,sha256=n06dPmR20rS4T3sJBWHQhGxzg4SJKzird9Hx0YTwwo0,226
-dwani/translate.py,sha256=c03N8-tN49IBcTA6GMOkrJ3MaVzZ12RnYdLQwRbEeoQ,2794
-dwani/vision.py,sha256=FviGewoV936CSv_K-latw0t3ZhSSCOF5LaGaq1oE4uA,3607
-dwani-0.1.14.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
-dwani-0.1.14.dist-info/METADATA,sha256=dHauJTsE8ujnm8Ak8QU0hJSiANxTjL6LoZIKZ3QPNCg,5543
-dwani-0.1.14.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-dwani-0.1.14.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
-dwani-0.1.14.dist-info/RECORD,,