dwani 0.1.1__tar.gz

dwani-0.1.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2025 Sachin Shetty
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

dwani-0.1.1/PKG-INFO

@@ -0,0 +1,193 @@
+Metadata-Version: 2.4
+Name: dwani
+Version: 0.1.1
+Summary: Multimodal AI server for Indian languages (speech, vision, LLMs, TTS, ASR, etc.)
+Author-email: sachin <python@dwani.ai>
+License: MIT License
+        
+        Copyright (c) 2025 Sachin Shetty
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+        
+Project-URL: Homepage, https://github.com/dwani-ai/dwani-server
+Project-URL: Source, https://github.com/dwani-ai/dwani-server
+Project-URL: Issues, https://github.com/dwani-ai/dwani-server/issues
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: fastapi>=0.95.0
+Requires-Dist: uvicorn[standard]>=0.22.0
+Requires-Dist: pydantic>=2.0.0
+Requires-Dist: requests>=2.25.0
+Requires-Dist: python-multipart>=0.0.5
+Requires-Dist: pydantic-settings>=2.0.0
+Dynamic: license-file
+
+# Dhwani Server
+
+Dhwani API is a FastAPI-based application providing AI-powered services for Indian languages, including text-to-speech (TTS), language model (LLM) chat, vision-language model (VLM) capabilities, and automatic speech recognition (ASR). It supports lazy loading of models for fast startup and includes endpoints for various tasks.
+
+## Features
+- **Text-to-Speech (TTS)**: Generate audio from text in Indian languages using Parler TTS.
+- **Chat**: Process Kannada prompts and respond in Kannada via translation and LLM.
+- **Vision-Language Model (VLM)**: Caption images, answer visual queries, detect, and point objects.
+- **Automatic Speech Recognition (ASR)**: Transcribe audio files in multiple Indian languages.
+- **Lazy Loading**: Models load on-demand or via an explicit endpoint for fast startup.
+
+## Prerequisites
+- **System Requirements - User **:
+    - **Python**: 3.10
+    - Ubuntu 22.04
+    - git 
+    - vscode
+- **System Requirements - Server **:
+  - Ubuntu with sufficient RAM (16GB+ recommended for models).
+  - Optional: NVIDIA GPU with CUDA support for faster inference.
+- **FFmpeg**: Required for audio processing (ASR).
+
+- Server Setup
+```bash
+export HF_HOME=/home/ubuntu/data-dhwani-models
+export HF_TOKEN='YOur-HF-token'
+python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+```
+## Installation
+
+1. **Clone the Repository**:
+   ```bash
+   git clone https://github.com/slabstech/dhwani-server
+   cd dhwani-server
+   ```
+
+2. Install Libraries:
+    - On Ubuntu: ```sudo apt-get install ffmpeg build-essential```
+
+3. Set Up Virtual Environment:
+    ```bash
+    python -m venv venv
+    source venv/bin/activate 
+    ```
+4. Install Dependencies:
+    ```bash
+    sudo apt-get install -y ffmpeg build-essential
+    curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --no-modify-path --profile minimal
+    . "$HOME/.cargo/env"
+    export CC=/usr/bin/gcc
+    export ENV CXX=/usr/bin/g++
+    ```
+    ```bash
+    pip install --no-cache-dir --upgrade pip setuptools psutil setuptools-rust torch==2.6.0
+    pip install --no-cache-dir flash-attn --no-build-isolation
+    ```
+
+    ```bash
+    pip install -r requirements.txt
+    ```
+    
+4. Set Environment Variable:
+    Create a .env file in the root directory and add your API key:
+    plaintext
+    ```bash
+    API_KEY=your_secret_key
+    ```
+
+5. Running the Server
+    - Start the Server:
+    ```bash
+        python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+    ```
+
+    - The server starts with models loaded on start
+    - Access the interactive API docs at http://localhost:7860/docs.
+    
+    - (Optional) Load All Models:
+    Preload all models (LLM, Translation, TTS, VLM, ASR) with:
+    - 
+    ```bash
+    curl -X POST "http://localhost:7860/load_all_models" -H "X-API-Key: your_secret_key"
+    ```
+
+- Usage
+    - Endpoints
+        - All endpoints require the X-API-Key header with the value from your .env file.
+
+    -    Health Check: GET /health
+    ```bash
+        curl "http://localhost:7860/health"
+    ```
+    - Response: 
+        ```bash 
+        {"status": "healthy", "model": "Qwen/Qwen2.5-3B-Instruct"}
+        ```
+    
+    - Text-to-Speech: POST /v1/audio/speech
+    ```  bash
+    curl -X POST "http://localhost:7860/v1/audio/speech" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"input": "ನಮಸ್ಕಾರ", "voice": "Female voice", "model": "ai4bharat/indic-parler-tts", "response_format": "mp3"}' --output speech.mp3
+    ```
+    - Chat: POST /chat
+    ``` bash
+        curl -X POST "http://localhost:7860/chat" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"prompt": "ನೀವು ಹೇಗಿದ್ದೀರಿ?"}'
+    ```
+
+    - Response: 
+            ```{"response": "<Kannada response>"}```
+    - Image Captioning: POST /caption/
+    ```bash
+        curl -X POST "http://localhost:7860/caption/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "length=short"
+    ```
+    - Response:``` {"caption": "<short caption>"}```
+    - Visual Query: POST /visual_query/
+    ```bash
+        curl -X POST "http://localhost:7860/visual_query/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "query=What is this?"
+    ```
+    - Response: ```{"answer": "<answer>"}```
+    - Object Detection: POST /detect/
+    ```bash
+        curl -X POST "http://localhost:7860/detect/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=face"
+    ```
+    - Response: ```{"objects": [<list of detected objects>]}```
+    - Object Pointing: POST /point/
+    ```bash
+
+        curl -X POST "http://localhost:7860/point/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=person"
+    ```
+    - Response: ```{"points": [<list of points>]}```
+    - Transcription: POST /transcribe/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe/?language=kannada" -H "X-API-Key: your_secret_key" -F "file=@audio.wav"
+    ```
+    - Response: ```{"text": "<transcribed text>"}```
+    - Batch Transcription: POST /transcribe_batch/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe_batch/?language=kannada" -H "X-API-Key: your_secret_key" -F "files=@audio1.wav" -F "files=@audio2.mp3"
+    ```
+    - Response: ```{"transcriptions": ["<text1>", "<text2>"]}```
+
+- Notes
+    - Lazy Loading: Models load on first use or via /load_all_models. Expect a delay on the first request for each model type.
+    Supported Languages: ASR supports multiple Indian languages (e.g., kannada, hindi, tamil); see models/asr.py for the full list.
+    Logs: Check dhwani_api.log for detailed logs (rotated at 10MB, 5 backups).
+    Performance: Use a GPU with flash-attn installed for faster TTS and ASR inference.
+
+- Troubleshooting
+
+    - Module Errors: Ensure all dependencies are installed. Re-run pip install if needed.
+    FFmpeg Not Found: Install FFmpeg and ensure it’s in your PATH.
+    Permission Denied: Run with sudo if accessing restricted ports (e.g., < 1024).
+

dwani-0.1.1/README.md

@@ -0,0 +1,152 @@
+# Dhwani Server
+
+Dhwani API is a FastAPI-based application providing AI-powered services for Indian languages, including text-to-speech (TTS), language model (LLM) chat, vision-language model (VLM) capabilities, and automatic speech recognition (ASR). It supports lazy loading of models for fast startup and includes endpoints for various tasks.
+
+## Features
+- **Text-to-Speech (TTS)**: Generate audio from text in Indian languages using Parler TTS.
+- **Chat**: Process Kannada prompts and respond in Kannada via translation and LLM.
+- **Vision-Language Model (VLM)**: Caption images, answer visual queries, detect, and point objects.
+- **Automatic Speech Recognition (ASR)**: Transcribe audio files in multiple Indian languages.
+- **Lazy Loading**: Models load on-demand or via an explicit endpoint for fast startup.
+
+## Prerequisites
+- **System Requirements - User **:
+    - **Python**: 3.10
+    - Ubuntu 22.04
+    - git 
+    - vscode
+- **System Requirements - Server **:
+  - Ubuntu with sufficient RAM (16GB+ recommended for models).
+  - Optional: NVIDIA GPU with CUDA support for faster inference.
+- **FFmpeg**: Required for audio processing (ASR).
+
+- Server Setup
+```bash
+export HF_HOME=/home/ubuntu/data-dhwani-models
+export HF_TOKEN='YOur-HF-token'
+python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+```
+## Installation
+
+1. **Clone the Repository**:
+   ```bash
+   git clone https://github.com/slabstech/dhwani-server
+   cd dhwani-server
+   ```
+
+2. Install Libraries:
+    - On Ubuntu: ```sudo apt-get install ffmpeg build-essential```
+
+3. Set Up Virtual Environment:
+    ```bash
+    python -m venv venv
+    source venv/bin/activate 
+    ```
+4. Install Dependencies:
+    ```bash
+    sudo apt-get install -y ffmpeg build-essential
+    curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --no-modify-path --profile minimal
+    . "$HOME/.cargo/env"
+    export CC=/usr/bin/gcc
+    export ENV CXX=/usr/bin/g++
+    ```
+    ```bash
+    pip install --no-cache-dir --upgrade pip setuptools psutil setuptools-rust torch==2.6.0
+    pip install --no-cache-dir flash-attn --no-build-isolation
+    ```
+
+    ```bash
+    pip install -r requirements.txt
+    ```
+    
+4. Set Environment Variable:
+    Create a .env file in the root directory and add your API key:
+    plaintext
+    ```bash
+    API_KEY=your_secret_key
+    ```
+
+5. Running the Server
+    - Start the Server:
+    ```bash
+        python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+    ```
+
+    - The server starts with models loaded on start
+    - Access the interactive API docs at http://localhost:7860/docs.
+    
+    - (Optional) Load All Models:
+    Preload all models (LLM, Translation, TTS, VLM, ASR) with:
+    - 
+    ```bash
+    curl -X POST "http://localhost:7860/load_all_models" -H "X-API-Key: your_secret_key"
+    ```
+
+- Usage
+    - Endpoints
+        - All endpoints require the X-API-Key header with the value from your .env file.
+
+    -    Health Check: GET /health
+    ```bash
+        curl "http://localhost:7860/health"
+    ```
+    - Response: 
+        ```bash 
+        {"status": "healthy", "model": "Qwen/Qwen2.5-3B-Instruct"}
+        ```
+    
+    - Text-to-Speech: POST /v1/audio/speech
+    ```  bash
+    curl -X POST "http://localhost:7860/v1/audio/speech" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"input": "ನಮಸ್ಕಾರ", "voice": "Female voice", "model": "ai4bharat/indic-parler-tts", "response_format": "mp3"}' --output speech.mp3
+    ```
+    - Chat: POST /chat
+    ``` bash
+        curl -X POST "http://localhost:7860/chat" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"prompt": "ನೀವು ಹೇಗಿದ್ದೀರಿ?"}'
+    ```
+
+    - Response: 
+            ```{"response": "<Kannada response>"}```
+    - Image Captioning: POST /caption/
+    ```bash
+        curl -X POST "http://localhost:7860/caption/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "length=short"
+    ```
+    - Response:``` {"caption": "<short caption>"}```
+    - Visual Query: POST /visual_query/
+    ```bash
+        curl -X POST "http://localhost:7860/visual_query/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "query=What is this?"
+    ```
+    - Response: ```{"answer": "<answer>"}```
+    - Object Detection: POST /detect/
+    ```bash
+        curl -X POST "http://localhost:7860/detect/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=face"
+    ```
+    - Response: ```{"objects": [<list of detected objects>]}```
+    - Object Pointing: POST /point/
+    ```bash
+
+        curl -X POST "http://localhost:7860/point/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=person"
+    ```
+    - Response: ```{"points": [<list of points>]}```
+    - Transcription: POST /transcribe/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe/?language=kannada" -H "X-API-Key: your_secret_key" -F "file=@audio.wav"
+    ```
+    - Response: ```{"text": "<transcribed text>"}```
+    - Batch Transcription: POST /transcribe_batch/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe_batch/?language=kannada" -H "X-API-Key: your_secret_key" -F "files=@audio1.wav" -F "files=@audio2.mp3"
+    ```
+    - Response: ```{"transcriptions": ["<text1>", "<text2>"]}```
+
+- Notes
+    - Lazy Loading: Models load on first use or via /load_all_models. Expect a delay on the first request for each model type.
+    Supported Languages: ASR supports multiple Indian languages (e.g., kannada, hindi, tamil); see models/asr.py for the full list.
+    Logs: Check dhwani_api.log for detailed logs (rotated at 10MB, 5 backups).
+    Performance: Use a GPU with flash-attn installed for faster TTS and ASR inference.
+
+- Troubleshooting
+
+    - Module Errors: Ensure all dependencies are installed. Re-run pip install if needed.
+    FFmpeg Not Found: Install FFmpeg and ensure it’s in your PATH.
+    Permission Denied: Run with sudo if accessing restricted ports (e.g., < 1024).
+

dwani-0.1.1/pyproject.toml

@@ -0,0 +1,28 @@
+[build-system]
+requires = ["setuptools>=61.0", "wheel"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "dwani"
+version = "0.1.1"
+description = "Multimodal AI server for Indian languages (speech, vision, LLMs, TTS, ASR, etc.)"
+authors = [
+    { name="sachin", email="python@dwani.ai" }
+]
+readme = "README.md"
+license = { file = "LICENSE" }
+requires-python = ">=3.10"
+
+dependencies = [
+    "fastapi>=0.95.0",
+    "uvicorn[standard]>=0.22.0",
+    "pydantic>=2.0.0",
+    "requests>=2.25.0",
+    "python-multipart>=0.0.5",
+    "pydantic-settings>=2.0.0"
+]
+
+[project.urls]
+Homepage = "https://github.com/dwani-ai/dwani-server"
+Source = "https://github.com/dwani-ai/dwani-server"
+Issues = "https://github.com/dwani-ai/dwani-server/issues"

dwani-0.1.1/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+

dwani-0.1.1/src/__init__.py

@@ -0,0 +1,39 @@
+from .client import DhwaniClient
+from .chat import Chat
+from .audio import Audio
+from .vision import Vision
+from .asr import ASR
+from .exceptions import DhwaniAPIError
+
+__all__ = ["DhwaniClient", "Chat", "Audio", "Vision", "ASR", "DhwaniAPIError"]
+
+# Optionally, instantiate a default client for convenience
+api_key = None
+api_base = "http://localhost:7860"
+
+def _get_client():
+    global _client
+    if "_client" not in globals() or _client is None:
+        from .client import DhwaniClient
+        globals()["_client"] = DhwaniClient(api_key=api_key, api_base=api_base)
+    return globals()["_client"]
+
+class chat:
+    @staticmethod
+    def create(prompt, **kwargs):
+        return _get_client().chat(prompt, **kwargs)
+
+class audio:
+    @staticmethod
+    def speech(*args, **kwargs):
+        return _get_client().speech(*args, **kwargs)
+
+class vision:
+    @staticmethod
+    def caption(*args, **kwargs):
+        return _get_client().caption(*args, **kwargs)
+
+class asr:
+    @staticmethod
+    def transcribe(*args, **kwargs):
+        return _get_client().transcribe(*args, **kwargs)

dwani-0.1.1/src/asr.py

@@ -0,0 +1,34 @@
+from .exceptions import DhwaniAPIError
+
+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}",
+            headers=client._headers(),
+            files=files
+        )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+class ASR:
+    @staticmethod
+    def transcribe(*args, **kwargs):
+        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-0.1.1/src/audio.py

@@ -0,0 +1,29 @@
+from .exceptions import DhwaniAPIError
+
+def audio_speech(client, input, voice, model, response_format="mp3", output_file=None):
+    data = {
+        "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,
+        stream=True
+    )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    if output_file:
+        with open(output_file, "wb") as f:
+            for chunk in resp.iter_content(chunk_size=8192):
+                f.write(chunk)
+        return output_file
+    return resp.content
+
+class Audio:
+    @staticmethod
+    def speech(*args, **kwargs):
+        from . import _get_client
+        return _get_client().speech(*args, **kwargs)

dwani-0.1.1/src/chat.py

@@ -0,0 +1,17 @@
+from .exceptions import DhwaniAPIError
+
+def chat_create(client, prompt, **kwargs):
+    resp = requests.post(
+        f"{client.api_base}/chat",
+        headers={**client._headers(), "Content-Type": "application/json"},
+        json={"prompt": prompt, **kwargs}
+    )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+class Chat:
+    @staticmethod
+    def create(prompt, **kwargs):
+        from . import _get_client
+        return _get_client().chat(prompt, **kwargs)

dwani-0.1.1/src/client.py

@@ -0,0 +1,41 @@
+import os
+import requests
+from .exceptions import DhwaniAPIError
+
+class DhwaniClient:
+    def __init__(self, api_key=None, api_base=None):
+        self.api_key = api_key or os.getenv("DHWANI_API_KEY")
+        self.api_base = api_base or os.getenv("DHWANI_API_BASE", "http://localhost:7860")
+        if not self.api_key:
+            raise ValueError("DHWANI_API_KEY not set")
+
+    def _headers(self):
+        return {"X-API-Key": self.api_key}
+
+    def chat(self, prompt, **kwargs):
+        from .chat import chat_create
+        return chat_create(self, prompt, **kwargs)
+
+    def speech(self, *args, **kwargs):
+        from .audio import audio_speech
+        return audio_speech(self, *args, **kwargs)
+
+    def caption(self, *args, **kwargs):
+        from .vision import vision_caption
+        return vision_caption(self, *args, **kwargs)
+
+    def transcribe(self, *args, **kwargs):
+        from .asr import asr_transcribe
+        return asr_transcribe(self, *args, **kwargs)
+    def document_ocr(self, file_path, language=None):
+        from .docs import document_ocr
+        return document_ocr(self, file_path, language)
+
+    def document_translate(self, file_path, src_lang, tgt_lang):
+        from .docs import document_translate
+        return document_translate(self, file_path, src_lang, tgt_lang)
+
+    def document_summarize(self, file_path, language=None):
+        from .docs import document_summarize
+        return document_summarize(self, file_path, language)
+

dwani-0.1.1/src/docs.py

@@ -0,0 +1,70 @@
+import requests
+from .exceptions import DhwaniAPIError
+
+def document_ocr(client, file_path, language=None):
+    """OCR a document (image/PDF) and return extracted text."""
+    with open(file_path, "rb") as f:
+        files = {"file": f}
+        data = {}
+        if language:
+            data["language"] = language
+        resp = requests.post(
+            f"{client.api_base}/v1/document/ocr",
+            headers=client._headers(),
+            files=files,
+            data=data
+        )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+def document_translate(client, file_path, src_lang, tgt_lang):
+    """Translate a document (image/PDF with text) from src_lang to tgt_lang."""
+    with open(file_path, "rb") as f:
+        files = {"file": f}
+        data = {
+            "src_lang": src_lang,
+            "tgt_lang": tgt_lang
+        }
+        resp = requests.post(
+            f"{client.api_base}/v1/document/translate",
+            headers=client._headers(),
+            files=files,
+            data=data
+        )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+def document_summarize(client, file_path, language=None):
+    """Summarize a document (image/PDF/text)."""
+    with open(file_path, "rb") as f:
+        files = {"file": f}
+        data = {}
+        if language:
+            data["language"] = language
+        resp = requests.post(
+            f"{client.api_base}/v1/document/summarize",
+            headers=client._headers(),
+            files=files,
+            data=data
+        )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+class Documents:
+    @staticmethod
+    def ocr(file_path, language=None):
+        from . import _get_client
+        return _get_client().document_ocr(file_path, language)
+
+    @staticmethod
+    def translate(file_path, src_lang, tgt_lang):
+        from . import _get_client
+        return _get_client().document_translate(file_path, src_lang, tgt_lang)
+
+    @staticmethod
+    def summarize(file_path, language=None):
+        from . import _get_client
+        return _get_client().document_summarize(file_path, language)

dwani-0.1.1/src/dwani.egg-info/PKG-INFO

@@ -0,0 +1,193 @@
+Metadata-Version: 2.4
+Name: dwani
+Version: 0.1.1
+Summary: Multimodal AI server for Indian languages (speech, vision, LLMs, TTS, ASR, etc.)
+Author-email: sachin <python@dwani.ai>
+License: MIT License
+        
+        Copyright (c) 2025 Sachin Shetty
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+        
+Project-URL: Homepage, https://github.com/dwani-ai/dwani-server
+Project-URL: Source, https://github.com/dwani-ai/dwani-server
+Project-URL: Issues, https://github.com/dwani-ai/dwani-server/issues
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: fastapi>=0.95.0
+Requires-Dist: uvicorn[standard]>=0.22.0
+Requires-Dist: pydantic>=2.0.0
+Requires-Dist: requests>=2.25.0
+Requires-Dist: python-multipart>=0.0.5
+Requires-Dist: pydantic-settings>=2.0.0
+Dynamic: license-file
+
+# Dhwani Server
+
+Dhwani API is a FastAPI-based application providing AI-powered services for Indian languages, including text-to-speech (TTS), language model (LLM) chat, vision-language model (VLM) capabilities, and automatic speech recognition (ASR). It supports lazy loading of models for fast startup and includes endpoints for various tasks.
+
+## Features
+- **Text-to-Speech (TTS)**: Generate audio from text in Indian languages using Parler TTS.
+- **Chat**: Process Kannada prompts and respond in Kannada via translation and LLM.
+- **Vision-Language Model (VLM)**: Caption images, answer visual queries, detect, and point objects.
+- **Automatic Speech Recognition (ASR)**: Transcribe audio files in multiple Indian languages.
+- **Lazy Loading**: Models load on-demand or via an explicit endpoint for fast startup.
+
+## Prerequisites
+- **System Requirements - User **:
+    - **Python**: 3.10
+    - Ubuntu 22.04
+    - git 
+    - vscode
+- **System Requirements - Server **:
+  - Ubuntu with sufficient RAM (16GB+ recommended for models).
+  - Optional: NVIDIA GPU with CUDA support for faster inference.
+- **FFmpeg**: Required for audio processing (ASR).
+
+- Server Setup
+```bash
+export HF_HOME=/home/ubuntu/data-dhwani-models
+export HF_TOKEN='YOur-HF-token'
+python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+```
+## Installation
+
+1. **Clone the Repository**:
+   ```bash
+   git clone https://github.com/slabstech/dhwani-server
+   cd dhwani-server
+   ```
+
+2. Install Libraries:
+    - On Ubuntu: ```sudo apt-get install ffmpeg build-essential```
+
+3. Set Up Virtual Environment:
+    ```bash
+    python -m venv venv
+    source venv/bin/activate 
+    ```
+4. Install Dependencies:
+    ```bash
+    sudo apt-get install -y ffmpeg build-essential
+    curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --no-modify-path --profile minimal
+    . "$HOME/.cargo/env"
+    export CC=/usr/bin/gcc
+    export ENV CXX=/usr/bin/g++
+    ```
+    ```bash
+    pip install --no-cache-dir --upgrade pip setuptools psutil setuptools-rust torch==2.6.0
+    pip install --no-cache-dir flash-attn --no-build-isolation
+    ```
+
+    ```bash
+    pip install -r requirements.txt
+    ```
+    
+4. Set Environment Variable:
+    Create a .env file in the root directory and add your API key:
+    plaintext
+    ```bash
+    API_KEY=your_secret_key
+    ```
+
+5. Running the Server
+    - Start the Server:
+    ```bash
+        python src/server/main.py --host 0.0.0.0 --port 7860 --config config_two
+    ```
+
+    - The server starts with models loaded on start
+    - Access the interactive API docs at http://localhost:7860/docs.
+    
+    - (Optional) Load All Models:
+    Preload all models (LLM, Translation, TTS, VLM, ASR) with:
+    - 
+    ```bash
+    curl -X POST "http://localhost:7860/load_all_models" -H "X-API-Key: your_secret_key"
+    ```
+
+- Usage
+    - Endpoints
+        - All endpoints require the X-API-Key header with the value from your .env file.
+
+    -    Health Check: GET /health
+    ```bash
+        curl "http://localhost:7860/health"
+    ```
+    - Response: 
+        ```bash 
+        {"status": "healthy", "model": "Qwen/Qwen2.5-3B-Instruct"}
+        ```
+    
+    - Text-to-Speech: POST /v1/audio/speech
+    ```  bash
+    curl -X POST "http://localhost:7860/v1/audio/speech" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"input": "ನಮಸ್ಕಾರ", "voice": "Female voice", "model": "ai4bharat/indic-parler-tts", "response_format": "mp3"}' --output speech.mp3
+    ```
+    - Chat: POST /chat
+    ``` bash
+        curl -X POST "http://localhost:7860/chat" -H "X-API-Key: your_secret_key" -H "Content-Type: application/json" -d '{"prompt": "ನೀವು ಹೇಗಿದ್ದೀರಿ?"}'
+    ```
+
+    - Response: 
+            ```{"response": "<Kannada response>"}```
+    - Image Captioning: POST /caption/
+    ```bash
+        curl -X POST "http://localhost:7860/caption/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "length=short"
+    ```
+    - Response:``` {"caption": "<short caption>"}```
+    - Visual Query: POST /visual_query/
+    ```bash
+        curl -X POST "http://localhost:7860/visual_query/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "query=What is this?"
+    ```
+    - Response: ```{"answer": "<answer>"}```
+    - Object Detection: POST /detect/
+    ```bash
+        curl -X POST "http://localhost:7860/detect/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=face"
+    ```
+    - Response: ```{"objects": [<list of detected objects>]}```
+    - Object Pointing: POST /point/
+    ```bash
+
+        curl -X POST "http://localhost:7860/point/" -H "X-API-Key: your_secret_key" -F "file=@image.jpg" -F "object_type=person"
+    ```
+    - Response: ```{"points": [<list of points>]}```
+    - Transcription: POST /transcribe/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe/?language=kannada" -H "X-API-Key: your_secret_key" -F "file=@audio.wav"
+    ```
+    - Response: ```{"text": "<transcribed text>"}```
+    - Batch Transcription: POST /transcribe_batch/
+    ```bash
+    curl -X POST "http://localhost:7860/transcribe_batch/?language=kannada" -H "X-API-Key: your_secret_key" -F "files=@audio1.wav" -F "files=@audio2.mp3"
+    ```
+    - Response: ```{"transcriptions": ["<text1>", "<text2>"]}```
+
+- Notes
+    - Lazy Loading: Models load on first use or via /load_all_models. Expect a delay on the first request for each model type.
+    Supported Languages: ASR supports multiple Indian languages (e.g., kannada, hindi, tamil); see models/asr.py for the full list.
+    Logs: Check dhwani_api.log for detailed logs (rotated at 10MB, 5 backups).
+    Performance: Use a GPU with flash-attn installed for faster TTS and ASR inference.
+
+- Troubleshooting
+
+    - Module Errors: Ensure all dependencies are installed. Re-run pip install if needed.
+    FFmpeg Not Found: Install FFmpeg and ensure it’s in your PATH.
+    Permission Denied: Run with sudo if accessing restricted ports (e.g., < 1024).
+

dwani-0.1.1/src/dwani.egg-info/SOURCES.txt

@@ -0,0 +1,16 @@
+LICENSE
+README.md
+pyproject.toml
+src/__init__.py
+src/asr.py
+src/audio.py
+src/chat.py
+src/client.py
+src/docs.py
+src/exceptions.py
+src/vision.py
+src/dwani.egg-info/PKG-INFO
+src/dwani.egg-info/SOURCES.txt
+src/dwani.egg-info/dependency_links.txt
+src/dwani.egg-info/requires.txt
+src/dwani.egg-info/top_level.txt

dwani-0.1.1/src/dwani.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

dwani-0.1.1/src/dwani.egg-info/requires.txt

@@ -0,0 +1,6 @@
+fastapi>=0.95.0
+uvicorn[standard]>=0.22.0
+pydantic>=2.0.0
+requests>=2.25.0
+python-multipart>=0.0.5
+pydantic-settings>=2.0.0

dwani-0.1.1/src/dwani.egg-info/top_level.txt

@@ -0,0 +1,8 @@
+__init__
+asr
+audio
+chat
+client
+docs
+exceptions
+vision

dwani-0.1.1/src/exceptions.py

@@ -0,0 +1,5 @@
+class DhwaniAPIError(Exception):
+    def __init__(self, response):
+        super().__init__(f"API Error {response.status_code}: {response.text}")
+        self.status_code = response.status_code
+        self.response = response

dwani-0.1.1/src/vision.py

@@ -0,0 +1,21 @@
+from .exceptions import DhwaniAPIError
+
+def vision_caption(client, file_path, length="short"):
+    with open(file_path, "rb") as f:
+        files = {"file": f}
+        data = {"length": length}
+        resp = requests.post(
+            f"{client.api_base}/caption/",
+            headers=client._headers(),
+            files=files,
+            data=data
+        )
+    if resp.status_code != 200:
+        raise DhwaniAPIError(resp)
+    return resp.json()
+
+class Vision:
+    @staticmethod
+    def caption(*args, **kwargs):
+        from . import _get_client
+        return _get_client().caption(*args, **kwargs)