Sarvam

Sarvam AI Skill

This skill provides access to Sarvam AI's suite of Indian language models.

Dependencies

This skill requires the following Python packages:

Security Note

The requests>=2.20.0 requirement addresses PYSEC-2018-28, a vulnerability where older versions of requests did not properly check SSL certificates. Version 2.20.0+ includes the fix for this security issue.

pip install -r requirements.txt

Usage

Text to Speech (TTS)

Generate speech from text in various Indian languages.

python skills/sarvam/scripts/sarvam_cli.py tts "Namaste, kaise hain aap?" --lang hi-IN --speaker meera --output hello.wav

Parameters:

• text: The text to speak.
• --lang: Language code (e.g., hi-IN for Hindi, bn-IN for Bengali, etc.).
• --speaker: Voice ID (e.g., meera, pavithra, arvind).
• --output: Output file path (default: output.wav).

Speech to Text (STT)

Transcribe audio files.

python skills/sarvam/scripts/sarvam_cli.py stt path/to/audio.wav --model saaras:v3

Parameters:

• file: Path to the audio file (wav, mp3).
• --model: Model to use (default: saaras:v3).
• --mode: STT Mode: transcribe (default), translate (to English), verbatim, translit, codemix.

Translation

Translate text between Indian languages and English.

python skills/sarvam/scripts/sarvam_cli.py translate "Hello, how are you?" --source en-IN --target hi-IN

Parameters:

• text: Text to translate.
• --source: Source language code.
• --target: Target language code.

Chat

Interact with Sarvam's LLM (sarvam-2g).

python skills/sarvam/scripts/sarvam_cli.py chat "What is the capital of India?"

Parameters:

• message: User message.
• --model: Model to use (default: sarvam-2g).
• --system: Optional system prompt.

Setup

1. Environment Variable: Ensure your API key is set in .env (already done for this workspace):

   SARVAM_API_KEY="sk_..."
   

2. Virtual Environment: The skill uses a local virtual environment at skills/sarvam/.venv. Dependencies (requests) are pre-installed here.

Usage

Use the virtual environment's Python to run commands:

Text to Speech (TTS)

skills/sarvam/.venv/bin/python skills/sarvam/scripts/sarvam_cli.py tts "Namaste, kaise hain aap?" --lang hi-IN --speaker meera --output hello.wav

Speech to Text (STT)

skills/sarvam/.venv/bin/python skills/sarvam/scripts/sarvam_cli.py stt path/to/audio.wav --model saaras:v3

Translation

skills/sarvam/.venv/bin/python skills/sarvam/scripts/sarvam_cli.py translate "Hello, how are you?" --source en-IN --target hi-IN

Chat

skills/sarvam/.venv/bin/python skills/sarvam/scripts/sarvam_cli.py chat "What is the capital of India?"