dwani 0.1.3__py3-none-any.whl → 0.1.4__py3-none-any.whl

dwani/asr.py

@@ -4,7 +4,7 @@ def asr_transcribe(client, file_path, language):
     with open(file_path, "rb") as f:
         files = {"file": f}
         resp = requests.post(
-            f"{client.api_base}/transcribe/?language={language}",
+            f"{client.api_base}/v1/transcribe/?language={language}",
             headers=client._headers(),
             files=files
         )
@@ -18,20 +18,3 @@ class ASR:
         from . import _get_client
         return _get_client().transcribe(*args, **kwargs)
 
-
-'''
-from .docs import Documents
-
-class documents:
-    @staticmethod
-    def ocr(file_path, language=None):
-        return _get_client().document_ocr(file_path, language)
-
-    @staticmethod
-    def translate(file_path, src_lang, tgt_lang):
-        return _get_client().document_translate(file_path, src_lang, tgt_lang)
-
-    @staticmethod
-    def summarize(file_path, language=None):
-        return _get_client().document_summarize(file_path, language)
-'''

dwani/audio.py

@@ -1,16 +1,16 @@
 from .exceptions import DhwaniAPIError
 import requests
-def audio_speech(client, input, voice, model, response_format="mp3", output_file=None):
-    data = {
+
+def audio_speech(client, input, response_format="mp3", output_file=None):
+    params = {
         "input": input,
-        "voice": voice,
-        "model": model,
         "response_format": response_format
     }
     resp = requests.post(
         f"{client.api_base}/v1/audio/speech",
-        headers={**client._headers(), "Content-Type": "application/json"},
-        json=data,
+        headers={**client._headers(), "accept": "application/json"},
+        params=params,
+        data='',  # Empty body, as in the curl example
         stream=True
     )
     if resp.status_code != 200:

dwani/client.py

@@ -5,7 +5,7 @@ from .exceptions import DhwaniAPIError
 class DhwaniClient:
     def __init__(self, api_key=None, api_base=None):
         self.api_key = api_key or os.getenv("DWANI_API_KEY")
-        self.api_base = api_base or os.getenv("DWANI_API_BASE", "http://localhost:7860")
+        self.api_base = api_base or os.getenv("DWANI_API_BASE_URL", "http://localhost:7860")
         if not self.api_key:
             raise ValueError("DHWANI_API_KEY not set")
 
@@ -16,14 +16,13 @@ class DhwaniClient:
         from .chat import chat_create
         return chat_create(self, prompt, src_lang, tgt_lang, **kwargs)
 
-
     def speech(self, *args, **kwargs):
         from .audio import audio_speech
         return audio_speech(self, *args, **kwargs)
 
-    def caption(self, *args, **kwargs):
+    def caption(self, file_path, query="describe the image", src_lang="eng_Latn", tgt_lang="kan_Knda"):
         from .vision import vision_caption
-        return vision_caption(self, *args, **kwargs)
+        return vision_caption(self, file_path, query, src_lang, tgt_lang)
 
     def transcribe(self, *args, **kwargs):
         from .asr import asr_transcribe

dwani/vision.py

@@ -1,12 +1,22 @@
 from .exceptions import DhwaniAPIError
 import requests
-def vision_caption(client, file_path, length="short"):
+def vision_caption(client, file_path, query="describe the image", src_lang="eng_Latn", tgt_lang="kan_Knda"):
+    # Build the endpoint using the client's api_base
+    url = (
+        f"{client.api_base}/v1/indic_visual_query"
+        f"?src_lang={src_lang}&tgt_lang={tgt_lang}"
+    )
+    headers = {
+        **client._headers(),
+        "accept": "application/json"
+        # Note: 'Content-Type' will be set automatically by requests when using 'files'
+    }
     with open(file_path, "rb") as f:
-        files = {"file": f}
-        data = {"length": length}
+        files = {"file": (file_path, f, "image/png")}
+        data = {"query": query}
         resp = requests.post(
-            f"{client.api_base}/caption/",
-            headers=client._headers(),
+            url,
+            headers=headers,
             files=files,
             data=data
         )

{dwani-0.1.3.dist-info → dwani-0.1.4.dist-info}/METADATA

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: dwani
-Version: 0.1.3
-Summary: Multimodal AI server for Indian languages (speech, vision, LLMs, TTS, ASR, etc.)
+Version: 0.1.4
+Summary: Multimodal API for Indian languages (speech, vision, LLMs, TTS, ASR, etc.)
 Author-email: sachin <python@dwani.ai>
 License: MIT License
         
@@ -28,7 +28,7 @@ License: MIT License
 Project-URL: Homepage, https://github.com/dwani-ai/dwani-python
 Project-URL: Source, https://github.com/dwani-ai/dwani-python
 Project-URL: Issues, https://github.com/dwani-ai/dwani-python/issues
-Requires-Python: >=3.10
+Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: requests>=2.25.0
@@ -49,7 +49,7 @@ import os
 
 dwani.api_key = os.getenv("DWANI_API_KEY")
 
-dwani.api_base = os.getenv("DWANI_API_BASE")
+dwani.api_base = os.getenv("DWANI_API_BASE_URL")
 
 resp = dwani.Chat.create("Hello!", "eng_Latn", "kan_Knda")
 print(resp)

dwani-0.1.4.dist-info/RECORD

@@ -0,0 +1,13 @@
+dwani/__init__.py,sha256=P2pyHkZ7JHn6lHSEbCdV4hjYAwCOXHN3RbsNIU0F5PE,1084
+dwani/asr.py,sha256=Y5Mbv1KsvhfXNZacMZycUHPn79NRQC-XPH0j9SYPUSY,590
+dwani/audio.py,sha256=Q9vw4uBxGy1vQzmiZjZGrY8hkAEQNkGhjz5OcnpFEQQ,888
+dwani/chat.py,sha256=WFuEShNd4nd6KUbIZTkm3eyPGoP33GepOLJax86nNn8,720
+dwani/client.py,sha256=DmGF93Doibam-rPuV64XlyIwgiWrDJigXZNBCkPz-ho,1601
+dwani/docs.py,sha256=2B87KqshPl7-2gDJAJxdvcgmPWuC2IQ1FcsDIyeVJPg,2202
+dwani/exceptions.py,sha256=qEN5ukqlnN7v-kHNEnISWFMpPMt6uTft9mPsTXJ4LVA,227
+dwani/vision.py,sha256=JtMSS0hI-8gxtLugVP10__enZsPPy8jheMyWXvrGrdw,1015
+dwani-0.1.4.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
+dwani-0.1.4.dist-info/METADATA,sha256=MZunthMWfUpXPchujVFu20UZDSURcnSd7FRWLt7KCos,2158
+dwani-0.1.4.dist-info/WHEEL,sha256=DnLRTWE75wApRYVsjgc6wsVswC54sMSJhAEd4xhDpBk,91
+dwani-0.1.4.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
+dwani-0.1.4.dist-info/RECORD,,

dwani-0.1.3.dist-info/RECORD

@@ -1,13 +0,0 @@
-dwani/__init__.py,sha256=P2pyHkZ7JHn6lHSEbCdV4hjYAwCOXHN3RbsNIU0F5PE,1084
-dwani/asr.py,sha256=Xe62t8dNQy9gRkn4LyjaagOXjKZ90aexYEL2D-oVP2U,1042
-dwani/audio.py,sha256=lyFjYAv4EsDIHPbopv9Lf5LJIzuJEuoBO9H37TVGaUo,894
-dwani/chat.py,sha256=WFuEShNd4nd6KUbIZTkm3eyPGoP33GepOLJax86nNn8,720
-dwani/client.py,sha256=qtXDAv9t-Eu5HGvH4ubrcmIIH8VgrTHn0c_Ozm4-cmQ,1513
-dwani/docs.py,sha256=2B87KqshPl7-2gDJAJxdvcgmPWuC2IQ1FcsDIyeVJPg,2202
-dwani/exceptions.py,sha256=qEN5ukqlnN7v-kHNEnISWFMpPMt6uTft9mPsTXJ4LVA,227
-dwani/vision.py,sha256=xi9gVsL8MA8VQYXIV99uqKuxYfs1kWu4tvqzbe95F1Y,623
-dwani-0.1.3.dist-info/licenses/LICENSE,sha256=IAD8tbwWZbPWHXgYjabHoMv0aaUzZUYzYiEbfhTCisY,1070
-dwani-0.1.3.dist-info/METADATA,sha256=KALRQIuxWFjbHHsF8QAdHHs76DVcfSAUPoWsGBcOsFs,2161
-dwani-0.1.3.dist-info/WHEEL,sha256=DnLRTWE75wApRYVsjgc6wsVswC54sMSJhAEd4xhDpBk,91
-dwani-0.1.3.dist-info/top_level.txt,sha256=AM5EhkyuO_EXQFR9JIxEV6tAYMCCyc-a1dLifpCGBUk,6
-dwani-0.1.3.dist-info/RECORD,,