dwani 0.1.19__tar.gz → 0.1.21__tar.gz

{dwani-0.1.19 → dwani-0.1.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dwani
-Version: 0.1.19
+Version: 0.1.21
 Summary: Multimodal API for Indian + European languages (Chat, Vision, TTS, ASR, Translate, Docs)
 Author-email: sachin <python@dwani.ai>
 License: MIT License
@@ -78,7 +78,7 @@ dwani.api_base = os.getenv("DWANI_API_BASE_URL")
 
 #### Document - OCR
 ```python
-result = dwani.Documents.run_ocr_number(file_path="dwani-workshop.pdf", page_number=1, model="gemma3")
+result = dwani.Documents.run_ocr_page(file_path="dwani-workshop.pdf", page_number=1, model="gemma3")
 print(result)
 ```
 ```json
@@ -86,6 +86,17 @@ print(result)
 ```
 
 
+#### Document - Summary
+
+```python
+result = dwani.Documents.summarize_all(
+            file_path="dwani-workshop.pdf", model="gemma3" , tgt_lang="english"  
+    )
+
+print("Document Query Response: gemma3- ", result["summary"])
+```
+
+
 ### Text Query 
 ---
 - gemma3 (default)
@@ -168,4 +179,16 @@ python -m build
 
 python -m twine upload dist/*
 
+-->
+
+<!--
+Without Batch  
+2025-07-14 13:39:50,330 - dwani_api - INFO - Request to /indic-summarize-pdf-all took 245.381 seconds
+INFO:dwani_api:Request to /indic-summarize-pdf-all took 245.381 seconds
+
+With Batch
+
+vllm serve google/gemma-3-4b-it --served-model-name gemma3 --host 0.0.0.0 --port 9000 --gpu-memory-utilization 0.8 --tensor-parallel-size 1 --max-model-len 65536     --dtype bfloat16 
+
+
 -->

{dwani-0.1.19 → dwani-0.1.21}/README.md

@@ -41,7 +41,7 @@ dwani.api_base = os.getenv("DWANI_API_BASE_URL")
 
 #### Document - OCR
 ```python
-result = dwani.Documents.run_ocr_number(file_path="dwani-workshop.pdf", page_number=1, model="gemma3")
+result = dwani.Documents.run_ocr_page(file_path="dwani-workshop.pdf", page_number=1, model="gemma3")
 print(result)
 ```
 ```json
@@ -49,6 +49,17 @@ print(result)
 ```
 
 
+#### Document - Summary
+
+```python
+result = dwani.Documents.summarize_all(
+            file_path="dwani-workshop.pdf", model="gemma3" , tgt_lang="english"  
+    )
+
+print("Document Query Response: gemma3- ", result["summary"])
+```
+
+
 ### Text Query 
 ---
 - gemma3 (default)
@@ -131,4 +142,16 @@ python -m build
 
 python -m twine upload dist/*
 
+-->
+
+<!--
+Without Batch  
+2025-07-14 13:39:50,330 - dwani_api - INFO - Request to /indic-summarize-pdf-all took 245.381 seconds
+INFO:dwani_api:Request to /indic-summarize-pdf-all took 245.381 seconds
+
+With Batch
+
+vllm serve google/gemma-3-4b-it --served-model-name gemma3 --host 0.0.0.0 --port 9000 --gpu-memory-utilization 0.8 --tensor-parallel-size 1 --max-model-len 65536     --dtype bfloat16 
+
+
 -->

{dwani-0.1.19 → dwani-0.1.21}/dwani/__init__.py

@@ -53,25 +53,36 @@ class translate:
 
 class document:
     @staticmethod
-    def run_ocr_number(file_path, page_number=1, model="gemma3"):
-        return _get_client().document_ocr_number(file_path, page_number, model)
+    def run_ocr_page(file_path, page_number=1, model="gemma3"):
+        return _get_client().document_ocr_page(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"):
-        return _get_client().document_summarize(file_path, page_number, src_lang, tgt_lang, model)
+    def run_summarize_page(file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
+        return _get_client().document_summarize_page(file_path, page_number, tgt_lang, model)
+    
+
+    @staticmethod
+    def run_summarize_all(file_path,  tgt_lang="kan_Knda", model="gemma3"):
+        return _get_client().document_summarize_all(file_path, tgt_lang, model)
     
     @staticmethod
-    def run_extract(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
-        return _get_client().extract(file_path, page_number, src_lang, tgt_lang, model)
+    def run_extract(file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
+        return _get_client().extract(file_path, page_number, tgt_lang, model)
+
     
     @staticmethod
-    def run_doc_query(file_path, page_number=1, prompt="list the key points", src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
-        return _get_client().doc_query(file_path, page_number, prompt, src_lang, tgt_lang, model)
+    def query_page(file_path, page_number=1,prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+        return _get_client().query_page(file_path, page_number, prompt, query_lang, tgt_lang, model)
+    
+
+    @staticmethod
+    def query_all(file_path, prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+        return _get_client().query_all(file_path, prompt, query_lang, tgt_lang, model)
     
     @staticmethod
-    def run_doc_query_kannada(file_path, page_number=1, prompt="list key points", src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def run_doc_query_kannada(file_path, page_number=1, prompt="list key points", src_lang="kan_Latn", tgt_lang="kan_Knda", model="gemma3"):
         return _get_client().doc_query_kannada(file_path, page_number, prompt, src_lang, tgt_lang, model)

{dwani-0.1.19 → dwani-0.1.21}/dwani/client.py

@@ -51,17 +51,22 @@ class DwaniClient:
         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"):
+    def document_summarize(self, file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
         from .docs import document_summarize
-        return document_summarize(self, file_path, page_number, src_lang, tgt_lang, model)
+        return document_summarize(self, file_path, page_number, tgt_lang, model)
 
-    def extract(self, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def extract(self, file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
         from .docs import extract
-        return extract(self, file_path=file_path, page_number=page_number, src_lang=src_lang, tgt_lang=tgt_lang, model=model)
+        return extract(self, file_path=file_path, page_number=page_number, tgt_lang=tgt_lang, model=model)
+
+    def query_page(self, file_path, page_number=1, prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+        from .docs import query_page
+        return query_page(self, file_path, page_number=page_number, prompt=prompt, query_lang=query_lang, tgt_lang=tgt_lang, model=model)
+
+    def query_all(self, file_path, prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+        from .docs import query_all
+        return query_all(self, file_path, prompt=prompt, query_lang=query_lang, tgt_lang=tgt_lang, model=model)
 
-    def doc_query(self, file_path, page_number=1, prompt="list the key points", src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
-        from .docs import doc_query
-        return doc_query(self, file_path, page_number=page_number, prompt=prompt, src_lang=src_lang, tgt_lang=tgt_lang, model=model)
 
     def doc_query_kannada(self, file_path, page_number=1, prompt="list key points", src_lang="eng_Latn", language="kan_Knda", model="gemma3"):
         from .docs import doc_query_kannada

{dwani-0.1.19 → dwani-0.1.21}/dwani/docs.py

@@ -67,7 +67,7 @@ def document_ocr_all(client, file_path, model="gemma3"):
     return resp.json()
 
 
-def document_ocr_number(client, file_path, page_number, model="gemma3"):
+def document_ocr_page(client, file_path, page_number, 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)
@@ -94,9 +94,10 @@ def document_ocr_number(client, file_path, page_number, model="gemma3"):
     
     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"):
+
+def document_summarize_page(client, file_path, page_number=1, 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}")
+    logger.debug(f"Calling document_summarize: file_path={file_path}, page_number={page_number}, tgt_lang={tgt_lang}, model={model}")
     validate_model(model)
     
     if not file_path.lower().endswith('.pdf'):
@@ -104,7 +105,6 @@ def document_summarize(client, file_path, page_number=1, src_lang="eng_Latn", tg
     if page_number < 1:
         raise ValueError("Page number must be at least 1")
     
-    src_lang_code = normalize_language(src_lang)
     tgt_lang_code = normalize_language(tgt_lang)
     
     url = f"{client.api_base}/v1/indic-summarize-pdf"
@@ -113,7 +113,6 @@ def document_summarize(client, file_path, page_number=1, src_lang="eng_Latn", tg
         files = {"file": (file_path, f, "application/pdf")}
         data = {
             "page_number": str(page_number),
-            "src_lang": src_lang_code,
             "tgt_lang": tgt_lang_code,
             "model": model
         }
@@ -135,9 +134,47 @@ def document_summarize(client, file_path, page_number=1, src_lang="eng_Latn", tg
 
     return resp.json()
 
-def extract(client, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+
+def document_summarize_all(client, file_path, tgt_lang="kan_Knda", model="gemma3"):
+    """Summarize a PDF document with language """
+    logger.debug(f"Calling document_summarize: file_path={file_path}, tgt_lang={tgt_lang}, model={model}")
+    validate_model(model)
+    
+    if not file_path.lower().endswith('.pdf'):
+        raise ValueError("File must be a PDF")
+    
+    tgt_lang_code = normalize_language(tgt_lang)
+    
+    url = f"{client.api_base}/v1/indic-summarize-pdf-all"
+    headers = client._headers()
+    with open(file_path, "rb") as f:
+        files = {"file": (file_path, f, "application/pdf")}
+        data = {
+            "tgt_lang": tgt_lang_code,
+            "model": model
+        }
+
+        try:
+            resp = requests.post(
+                url,
+                headers=headers,
+                files=files,
+                data=data,
+                timeout=90
+            )
+            resp.raise_for_status()
+        except requests.RequestException as e:
+            logger.error(f"Summarize request failed: {str(e)}")
+            raise DwaniAPIError(resp) if 'resp' in locals() else DwaniAPIError.from_exception(e)
+    
+    logger.debug(f"Summarize response: {resp.status_code}")
+
+    return resp.json()
+
+
+def extract(client, file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
     """Extract and translate text from a PDF document using form data."""
-    logger.debug(f"Calling extract: file_path={file_path}, page_number={page_number}, src_lang={src_lang}, tgt_lang={tgt_lang}, model={model}")
+    logger.debug(f"Calling extract: file_path={file_path}, page_number={page_number}, tgt_lang={tgt_lang}, model={model}")
     validate_model(model)
     
     if not file_path.lower().endswith('.pdf'):
@@ -145,7 +182,6 @@ def extract(client, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan
     if page_number < 1:
         raise ValueError("Page number must be at least 1")
     
-    src_lang_code = normalize_language(src_lang)
     tgt_lang_code = normalize_language(tgt_lang)
     
     url = f"{client.api_base}/v1/indic-extract-text/"
@@ -155,7 +191,6 @@ def extract(client, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan
 
         data = {
             "page_number": str(page_number),
-            "src_lang": src_lang_code,
             "tgt_lang": tgt_lang_code,
             "model": model
         }
@@ -176,17 +211,17 @@ def extract(client, file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan
 
     return resp.json()
 
-def doc_query(
+def query_page(
     client,
     file_path,
     page_number=1,
     prompt="list the key points",
-    src_lang="eng_Latn",
     tgt_lang="kan_Knda",
+    query_lang="eng_Latn",
     model="gemma3"
 ):
     """Query a document with a custom prompt and language options."""
-    logger.debug(f"Calling doc_query: file_path={file_path}, page_number={page_number}, prompt={prompt}, src_lang={src_lang}, tgt_lang={tgt_lang}, model={model}")
+    logger.debug(f"Calling doc_query: file_path={file_path}, page_number={page_number}, prompt={prompt}, tgt_lang={tgt_lang}, model={model}")
     validate_model(model)
     
     if not file_path.lower().endswith('.pdf'):
@@ -196,9 +231,11 @@ def doc_query(
     if not prompt.strip():
         raise ValueError("Prompt cannot be empty")
     
-    src_lang_code = normalize_language(src_lang)
     tgt_lang_code = normalize_language(tgt_lang)
     
+    query_lang_code = normalize_language(query_lang)
+
+
     url = f"{client.api_base}/v1/indic-custom-prompt-pdf"
     headers = client._headers()
     with open(file_path, "rb") as f:
@@ -206,8 +243,58 @@ def doc_query(
         data = {
             "page_number": str(page_number),
             "prompt": prompt,
-            "src_lang": src_lang_code,
             "tgt_lang": tgt_lang_code,
+            "query_lang": query_lang_code,
+            "model": model
+        }
+
+        try:
+            resp = requests.post(
+                url,
+                headers=headers,
+                files=files,
+                data=data,
+                #params=params,
+                timeout=90
+            )
+            resp.raise_for_status()
+        except requests.RequestException as e:
+            logger.error(f"Doc query request failed: {str(e)}")
+            raise DwaniAPIError(resp) if 'resp' in locals() else DwaniAPIError.from_exception(e)
+    
+    logger.debug(f"Doc query response: {resp.status_code}")
+
+    return resp.json()
+
+def query_all(
+    client,
+    file_path,
+    prompt="list the key points",
+    tgt_lang="kan_Knda",
+    query_lang="eng_Latn",
+    model="gemma3"
+):
+    """Query a document with a custom prompt and language options."""
+    logger.debug(f"Calling doc_query: file_path={file_path}, prompt={prompt}, tgt_lang={tgt_lang}, model={model}")
+    validate_model(model)
+    
+    if not file_path.lower().endswith('.pdf'):
+        raise ValueError("File must be a PDF")
+    if not prompt.strip():
+        raise ValueError("Prompt cannot be empty")
+    
+    tgt_lang_code = normalize_language(tgt_lang)
+    
+    query_lang_code = normalize_language(query_lang)
+
+    url = f"{client.api_base}/v1/indic-custom-prompt-pdf-all"
+    headers = client._headers()
+    with open(file_path, "rb") as f:
+        files = {"file": (file_path, f, "application/pdf")}
+        data = {
+            "prompt": prompt,
+            "tgt_lang": tgt_lang_code,
+            "query_lang": query_lang_code,
             "model": model
         }
 
@@ -228,17 +315,18 @@ def doc_query(
 
     return resp.json()
 
+
 def doc_query_kannada(
     client,
     file_path,
     page_number=1,
     prompt="list key points",
-    src_lang="eng_Latn",
     tgt_lang="kan_Knda",
+    src_lang="kan_Knda",
     model="gemma3"
 ):
     """Query a document with a custom prompt, outputting in Kannada."""
-    logger.debug(f"Calling doc_query_kannada: file_path={file_path}, page_number={page_number}, prompt={prompt}, src_lang={src_lang}, tgt_lang={tgt_lang}, model={model}")
+    logger.debug(f"Calling doc_query_kannada: file_path={file_path}, page_number={page_number}, prompt={prompt}, tgt_lang={tgt_lang}, model={model}")
     validate_model(model)
     
     if not file_path.lower().endswith('.pdf'):
@@ -248,8 +336,10 @@ def doc_query_kannada(
     if not prompt.strip():
         raise ValueError("Prompt cannot be empty")
     
-    src_lang_code = normalize_language(src_lang)
     tgt_lang_code = normalize_language(tgt_lang) if tgt_lang else "kan_Knda"
+
+    src_lang_code = normalize_language(src_lang) 
+    
     
     url = f"{client.api_base}/v1/indic-custom-prompt-pdf"
     headers = client._headers()
@@ -259,8 +349,8 @@ def doc_query_kannada(
         data = {
             "page_number": str(page_number),
             "prompt": prompt,
-            "src_lang": src_lang_code,
             "tgt_lang": tgt_lang_code,
+            "src_lang": src_lang_code,
             "model": model
         }
         try:
@@ -282,36 +372,50 @@ def doc_query_kannada(
 
 class Documents:
     @staticmethod
-    def run_ocr_number(file_path, page_number=2,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, model)
+        return document_ocr_page(client, file_path=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_all(client, file_path, model)
+        return document_ocr_all(client, file_path=file_path, model=model)
     
     @staticmethod
-    def summarize(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def summarize_page(file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
+        from .client import DwaniClient
+        client = DwaniClient()
+        return document_summarize_page(client, file_path=file_path, page_number=page_number, tgt_lang=tgt_lang, model=model)
+
+
+    @staticmethod
+    def summarize_all(file_path, tgt_lang="kan_Knda", model="gemma3"):
+        from .client import DwaniClient
+        client = DwaniClient()
+        return document_summarize_all(client, file_path=file_path, tgt_lang=tgt_lang, model=model)
+
+    @staticmethod
+    def run_extract(file_path, page_number=1, tgt_lang="kan_Knda", model="gemma3"):
         from .client import DwaniClient
         client = DwaniClient()
-        return document_summarize(client, file_path, page_number, src_lang, tgt_lang, model)
+        return extract(client, file_path=file_path, page_number=page_number, tgt_lang=tgt_lang, model=model)
     
     @staticmethod
-    def run_extract(file_path, page_number=1, src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def query_page(file_path, page_number=1, prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
         from .client import DwaniClient
         client = DwaniClient()
-        return extract(client, file_path, page_number, src_lang, tgt_lang, model)
+        return query_page(client, file_path=file_path, page_number=page_number, prompt=prompt, query_lang=query_lang, tgt_lang=tgt_lang, model=model)
     
     @staticmethod
-    def run_doc_query(file_path, page_number=1, prompt="list the key points", src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def query_all(file_path, prompt="list the key points", query_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
         from .client import DwaniClient
         client = DwaniClient()
-        return doc_query(client, file_path, page_number, prompt, src_lang, tgt_lang, model)
+        return query_all(client, file_path=file_path, prompt=prompt, query_lang=query_lang, tgt_lang=tgt_lang, model=model)
     
+
     @staticmethod
-    def run_doc_query_kannada(file_path, page_number=1, prompt="list key points", src_lang="eng_Latn", tgt_lang="kan_Knda", model="gemma3"):
+    def run_doc_query_kannada(file_path, page_number=1, prompt="list key points", tgt_lang="kan_Knda", model="gemma3"):
         from .client import DwaniClient
         client = DwaniClient()
-        return doc_query_kannada(client, file_path, page_number, prompt, src_lang, tgt_lang, model)
+        return doc_query_kannada(client, file_path=file_path, page_number=page_number, prompt=prompt, tgt_lang=tgt_lang, model=model)

{dwani-0.1.19 → dwani-0.1.21}/pyproject.toml

@@ -5,7 +5,7 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "dwani"
 
-version = "0.1.19"
+version = "0.1.21"
 description = "Multimodal API for Indian + European languages (Chat, Vision, TTS, ASR, Translate, Docs)"
 authors = [
     { name="sachin", email="python@dwani.ai" }