Chat with Sarvam-M

Sarvam-M is Sarvam AI's flagship language model optimized for Indian languages with hybrid thinking capabilities for complex reasoning.

Key Features

•24B parameters optimized for Indic languages
•Hybrid thinking mode for complex reasoning
•Free to use (no cost)
•OpenAI-compatible API
•Multilingual chat in English + 10 Indian languages

Installation

bash

pip install sarvamai
# or
pip install openai  # OpenAI-compatible

Quick Start

Sarvam SDK

python

from sarvamai import SarvamAI

client = SarvamAI()

response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "भारत की राजधानी क्या है?"
    }
]
)

print(response.choices[
    0
].message.content)

OpenAI-Compatible

python

from openai import OpenAI

client = OpenAI(
    api_key=os.environ[
    "SARVAM_API_KEY"
],
    base_url="https://api.sarvam.ai/v1"
)

response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "What is the capital of India?"
    }
]
)

print(response.choices[
    0
].message.content)

Hybrid Thinking Mode

Enable step-by-step reasoning for complex problems:

python

response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "Solve: If a train travels 120km in 2 hours, what is its average speed?"
    }
],
    thinking=True  # Enable reasoning
)

# Access thinking process
print("Thinking:", response.choices[
    0
].message.thinking)
print("Answer:", response.choices[
    0
].message.content)

System Prompts

Guide the model's behavior:

python

response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "system",
        "content": "You are a helpful Hindi tutor. Always respond in Hindi with English transliteration in parentheses."
    },
    {
        "role": "user",
        "content": "How do I say 'Good morning'?"
    }
]
)

Multi-turn Conversations

Maintain context across turns:

python

messages = [
    {
        "role": "system",
        "content": "You are a knowledgeable assistant."
    },
    {
        "role": "user",
        "content": "Tell me about the Taj Mahal"
    },
    {
        "role": "assistant",
        "content": "The Taj Mahal is a white marble mausoleum..."
    },
    {
        "role": "user",
        "content": "Who built it and when?"
    }
]

response = client.chat.completions.create(
    model="sarvam-m",
    messages=messages
)

Streaming

Stream responses token by token:

python

stream = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "Write a short poem about India"
    }
],
    stream=True
)

for chunk in stream:
    if chunk.choices[
    0
].delta.content:
        print(chunk.choices[
    0
].delta.content, end="", flush=True)

Temperature Control

Control randomness:

python

# Creative (higher temperature)
response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "Write a creative story"
    }
],
    temperature=0.9
)

# Factual (lower temperature)
response = client.chat.completions.create(
    model="sarvam-m",
    messages=[
    {
        "role": "user",
        "content": "What is 2+2?"
    }
],
    temperature=0.1
)

JavaScript

javascript

import { SarvamAI
} from "sarvamai";

const client = new SarvamAI();

const response = await client.chat.completions.create({
  model: "sarvam-m",
  messages: [
        { role: "user", content: "भारत की राजधानी क्या है?"
        }
    ]
});

console.log(response.choices[
    0
].message.content);

OpenAI SDK

javascript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.SARVAM_API_KEY,
  baseURL: "https://api.sarvam.ai/v1"
});

const response = await client.chat.completions.create({
  model: "sarvam-m",
  messages: [
        { role: "user", content: "Hello!"
        }
    ]
});

cURL

bash

curl -X POST "https://api.sarvam.ai/v1/chat/completions" \
  -H "api-subscription-key: $SARVAM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sarvam-m",
    "messages": [
        {
            "role": "user",
            "content": "What is the capital of India?"
        }
    ]
}'

Parameters

Parameter	Type	Required	Description
`model`	string	Yes	`sarvam-m`
`messages`	array	Yes	Conversation history
`temperature`	float	No	0.0-2.0 (default: 1.0)
`max_tokens`	int	No	Max response length
`stream`	bool	No	Enable streaming
`thinking`	bool	No	Enable hybrid thinking
`top_p`	float	No	Nucleus sampling (0.0-1.0)

Response

json

{
    "id": "chatcmpl-abc123",
    "object": "chat.completion",
    "model": "sarvam-m",
    "choices": [
        {
            "index": 0,
            "message": {
                "role": "assistant",
                "content": "भारत की राजधानी नई दिल्ली है।"
            },
            "finish_reason": "stop"
        }
    ],
    "usage": {
        "prompt_tokens": 15,
        "completion_tokens": 12,
        "total_tokens": 27
    }
}

See references/prompting.md for advanced prompting techniques.