Exploring Sarvam AI

This repository contains small, practical examples for building with Sarvam AI:

• direct chat completion calls using the Sarvam SDK
• an OpenAI-compatible local bridge server backed by Sarvam
• a client script that talks to that bridge through the OpenAI SDK

Prerequisites

• Python (project metadata currently declares >=3.14 in pyproject.toml)
• A valid Sarvam API key

Setup

1. Clone the repository:

git clone https://github.com/harshdadiya-wappnet/explore_sarvam
cd explore_sarvam

2. Create your environment file:

cp .env.example .env

Then set your key in .env:

SARVAM_API_KEY=your_actual_key

3. Install dependencies (choose one):

Using uv:

uv venv
source .venv/bin/activate
uv sync

Using pip:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
pip install fastapi uvicorn openai

Project Structure

.
├── README.md
├── pyproject.toml
├── requirements.txt
└── learning/
    ├── simple_usage.py
    ├── openai_chat_completion_server.py
    └── simple_usage_using_openai.py

Usage

1. Direct Sarvam SDK chat loop

Run:

python -m learning.simple_usage

What it does:

• reads SARVAM_API_KEY from .env
• sends chat completion requests through the sarvamai SDK
• runs an interactive CLI loop until you type exit, quit, or q

2. Start OpenAI-compatible bridge server

Run:

python -m learning.openai_chat_completion_server

The server starts at http://localhost:8000 and exposes:

• GET /v1/models
• POST /v1/chat/completions

It accepts OpenAI-style chat completion requests and forwards them to Sarvam.

3. Use OpenAI SDK client against the local bridge

In a second terminal (with the same virtual env active), run:

python learning/simple_usage_using_openai.py

This script uses:

• base_url="http://localhost:8000"
• api_key=$SARVAM_API_KEY
• model="sarvam-m"

Quick API Check (optional)

With the bridge server running:

curl -s http://localhost:8000/v1/models

Example chat request:

curl -s http://localhost:8000/v1/chat/completions \
  -H "Authorization: Bearer $SARVAM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sarvam-m",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Say hello in one line."}
    ]
  }'

Notes

• If you are using only requirements.txt, install extra runtime packages for the bridge/client (fastapi, uvicorn, openai) as shown above.
• Keep .env out of version control.