dwani 0.1.14__py3-none-any.whl → 0.1.16__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
dwani/__init__.py CHANGED
@@ -53,8 +53,8 @@ class translate:
53
53
 
54
54
  class document:
55
55
  @staticmethod
56
- def run_ocr(file_path, language="eng_Latn", model="gemma3"):
57
- return _get_client().document_ocr(file_path, language, model)
56
+ def run_ocr(file_path, model="gemma3"):
57
+ return _get_client().document_ocr(file_path, model)
58
58
 
59
59
  @staticmethod
60
60
  def run_summarize(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
dwani/client.py CHANGED
@@ -43,9 +43,9 @@ class DwaniClient:
43
43
  from .asr import asr_transcribe
44
44
  return asr_transcribe(self, file_path=file_path, language=language)
45
45
 
46
- def document_ocr(self, file_path, language=None, model="gemma3"):
46
+ def document_ocr(self, file_path, model="gemma3"):
47
47
  from .docs import document_ocr
48
- return document_ocr(self, file_path=file_path, language=language, model=model)
48
+ return document_ocr(self, file_path=file_path, model=model)
49
49
 
50
50
  def document_summarize(self, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
51
51
  from .docs import document_summarize
dwani/docs.py CHANGED
@@ -40,21 +40,19 @@ def validate_model(model):
40
40
  raise ValueError(f"Unsupported model: {model}. Supported models: {VALID_MODELS}")
41
41
  return model
42
42
 
43
- def document_ocr(client, file_path, language=None, model="gemma3"):
43
+ def document_ocr(client, file_path, model="gemma3"):
44
44
  """OCR a document (image/PDF) and return extracted text."""
45
- logger.debug(f"Calling document_ocr: file_path={file_path}, language={language}, model={model}")
45
+ logger.debug(f"Calling document_ocr: file_path={file_path}, model={model}")
46
46
  validate_model(model)
47
47
 
48
48
  data = {"model": model}
49
- if language:
50
- data["language"] = normalize_language(language)
51
-
49
+
52
50
  with open(file_path, "rb") as f:
53
51
  mime_type = "application/pdf" if file_path.lower().endswith('.pdf') else "image/png"
54
52
  files = {"file": (file_path, f, mime_type)}
55
53
  try:
56
54
  resp = requests.post(
57
- f"{client.api_base}/v1/document/ocr",
55
+ f"{client.api_base}/v1/extract-text",
58
56
  headers=client._headers(),
59
57
  files=files,
60
58
  data=data,
@@ -256,10 +254,10 @@ def doc_query_kannada(
256
254
 
257
255
  class Documents:
258
256
  @staticmethod
259
- def ocr(file_path, language=None, model="gemma3"):
257
+ def run_ocr(file_path, model="gemma3"):
260
258
  from .client import DwaniClient
261
259
  client = DwaniClient()
262
- return document_ocr(client, file_path, language, model)
260
+ return document_ocr(client, file_path, model)
263
261
 
264
262
  @staticmethod
265
263
  def summarize(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: dwani
3
- Version: 0.1.14
3
+ Version: 0.1.16
4
4
  Summary: Multimodal API for Indian + European languages (Chat, Vision, TTS, ASR, Translate, Docs)
5
5
  Author-email: sachin <python@dwani.ai>
6
6
  License: MIT License
@@ -64,13 +64,26 @@ dwani.api_key = os.getenv("DWANI_API_KEY")
64
64
  dwani.api_base = os.getenv("DWANI_API_BASE_URL")
65
65
  ```
66
66
 
67
- - Check examples folder for detailed use cases
67
+
68
+ ---
69
+
70
+ - Source Code : [https://github.com/dwani-ai/dwani-python-sdk](https://github.com/dwani-ai/dwani-python-sdk)
71
+ - Check examples folder for detailed use cases
72
+
68
73
  - [examples/chat.py](examples/chat.py)
69
74
  - [examples/vision.py](examples/vision.py)
70
75
  - [examples/docs.py](examples/docs.py)
71
76
  - [examples/speech.py](examples/speech.py)
72
77
  - [examples/asr.py](examples/asr.py)
73
78
 
79
+ #### Document - OCR
80
+ ```python
81
+ result = dwani.Documents.run_ocr(file_path="dwani-workshop.pdf", model="gemma3")
82
+ print(result)
83
+ ```
84
+ ```json
85
+ {'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):**"}
86
+ ```
74
87
 
75
88
 
76
89
  ### Text Query
@@ -0,0 +1,14 @@
1
+ dwani/__init__.py,sha256=2WxVFPYpwyZ68yqbzNOEpmwSsi7ksvw8-pZRmReUZCQ,3009
2
+ dwani/asr.py,sha256=BAdqivQd57NJZX1dSY-J6EFi8TDdyuhf_AyCPcQ0M7w,1719
3
+ dwani/audio.py,sha256=CFQrYU-KLwO7pCh_R7c1SSDJ6bugE5_av7lV8XTl-dY,936
4
+ dwani/chat.py,sha256=Tui52XBhUyDyN2rOFoLme4oB0Q8fkD9_0tFDAnRzoaU,2979
5
+ dwani/client.py,sha256=UvzmXShctntMmm1rIONVssv1c8HNgzBMZLOjxrCbp-4,3360
6
+ dwani/docs.py,sha256=KSqmbVoImEFI_HK102iJwlemN3XQii2Mo7WOob2kFQE,10464
7
+ dwani/exceptions.py,sha256=n06dPmR20rS4T3sJBWHQhGxzg4SJKzird9Hx0YTwwo0,226
8
+ dwani/translate.py,sha256=c03N8-tN49IBcTA6GMOkrJ3MaVzZ12RnYdLQwRbEeoQ,2794
9
+ dwani/vision.py,sha256=FviGewoV936CSv_K-latw0t3ZhSSCOF5LaGaq1oE4uA,3607
10
+ dwani-0.1.16.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
11
+ dwani-0.1.16.dist-info/METADATA,sha256=UKmC0j_PK9K8XcCqZZUuUbzPWbRi_bPlkO0BR6E9Xk8,6008
12
+ dwani-0.1.16.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
13
+ dwani-0.1.16.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
14
+ dwani-0.1.16.dist-info/RECORD,,
@@ -1,14 +0,0 @@
1
- dwani/__init__.py,sha256=rV93n_1nOed4ak7NOcwtDjxsT7JbDIivhtRRcrDR80A,3040
2
- dwani/asr.py,sha256=BAdqivQd57NJZX1dSY-J6EFi8TDdyuhf_AyCPcQ0M7w,1719
3
- dwani/audio.py,sha256=CFQrYU-KLwO7pCh_R7c1SSDJ6bugE5_av7lV8XTl-dY,936
4
- dwani/chat.py,sha256=Tui52XBhUyDyN2rOFoLme4oB0Q8fkD9_0tFDAnRzoaU,2979
5
- dwani/client.py,sha256=fGtnY9L0u_mymY7xUeFDhCBX7P7fd4WNko7WTDgrDV4,3394
6
- dwani/docs.py,sha256=Cp0Gtudug79GH25toB-Npl35ZFA0TM32oZF2xH1VmNY,10598
7
- dwani/exceptions.py,sha256=n06dPmR20rS4T3sJBWHQhGxzg4SJKzird9Hx0YTwwo0,226
8
- dwani/translate.py,sha256=c03N8-tN49IBcTA6GMOkrJ3MaVzZ12RnYdLQwRbEeoQ,2794
9
- dwani/vision.py,sha256=FviGewoV936CSv_K-latw0t3ZhSSCOF5LaGaq1oE4uA,3607
10
- dwani-0.1.14.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
11
- dwani-0.1.14.dist-info/METADATA,sha256=dHauJTsE8ujnm8Ak8QU0hJSiANxTjL6LoZIKZ3QPNCg,5543
12
- dwani-0.1.14.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
13
- dwani-0.1.14.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
14
- dwani-0.1.14.dist-info/RECORD,,
File without changes