Converts text between scripts while preserving the same language, with spoken-form conversion and international or native numeral formatting.
openapi: 3.0.1
info:
title: Sarvam AI API
description: >-
REST API for Sarvam AI, India's full-stack sovereign AI platform. Provides
chat completions over Sarvam's Indic LLMs, speech-to-text and
speech-to-text-translate, text-to-speech, text translation,
transliteration, and language identification across Indian languages.
All requests authenticate with an api-subscription-key header.
termsOfService: https://www.sarvam.ai/terms-of-service
contact:
name: Sarvam AI Support
url: https://www.sarvam.ai
version: '1.0'
servers:
- url: https://api.sarvam.ai
security:
- ApiSubscriptionKey: []
tags:
- name: Chat
- name: Speech to Text
- name: Text to Speech
- name: Translate
- name: Transliterate
- name: Language Identification
paths:
/v1/chat/completions:
post:
operationId: createChatCompletion
tags:
- Chat
summary: Create a chat completion
description: >-
Generates a model response for a conversation using Sarvam's Indic LLMs
(sarvam-m, sarvam-30b, sarvam-105b). Supports streaming via
Server-Sent Events when stream is true, tool use, and reasoning effort.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/ChatCompletionRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/ChatCompletionResponse'
text/event-stream:
schema:
type: string
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/speech-to-text:
post:
operationId: transcribeSpeech
tags:
- Speech to Text
summary: Transcribe audio to text
description: >-
Transcribes an audio file in an Indian language using the Saaras or
Saarika speech-recognition models. Supports auto language detection,
word-level timestamps, and speaker diarization.
requestBody:
required: true
content:
multipart/form-data:
schema:
$ref: '#/components/schemas/SpeechToTextRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/SpeechToTextResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/speech-to-text-translate:
post:
operationId: translateSpeech
tags:
- Speech to Text
summary: Transcribe and translate audio to English
description: >-
Transcribes Indic-language audio and translates the result into English
in a single request using the Saaras family of models.
requestBody:
required: true
content:
multipart/form-data:
schema:
$ref: '#/components/schemas/SpeechToTextTranslateRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/SpeechToTextResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/text-to-speech:
post:
operationId: convertTextToSpeech
tags:
- Text to Speech
summary: Convert text to speech
description: >-
Synthesizes natural speech for Indian languages using the Bulbul models
(v2, v3) with 30+ speakers and configurable pace, pitch, sample rate,
and output audio codec. Returns base64-encoded audio.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/TextToSpeechRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/TextToSpeechResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/translate:
post:
operationId: translateText
tags:
- Translate
summary: Translate text between Indian languages
description: >-
Translates text across Indian languages using Mayura (12 languages,
multiple tones) or Sarvam-Translate (all 22 scheduled languages), with
tone modes, output-script control, and numeral formatting.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/TranslateRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/TranslateResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/transliterate:
post:
operationId: transliterateText
tags:
- Transliterate
summary: Transliterate text between scripts
description: >-
Converts text between scripts while preserving the same language, with
spoken-form conversion and international or native numeral formatting.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/TransliterateRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/TransliterateResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
/text-lid:
post:
operationId: identifyLanguage
tags:
- Language Identification
summary: Identify the language of text
description: >-
Automatically detects the language and script of input text, returning
BCP-47 language and script codes.
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/LanguageIdRequest'
responses:
'200':
description: OK
content:
application/json:
schema:
$ref: '#/components/schemas/LanguageIdResponse'
'401':
$ref: '#/components/responses/Unauthorized'
'429':
$ref: '#/components/responses/TooManyRequests'
components:
securitySchemes:
ApiSubscriptionKey:
type: apiKey
in: header
name: api-subscription-key
description: API subscription key created in the Sarvam AI dashboard.
responses:
Unauthorized:
description: Missing or invalid api-subscription-key.
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
TooManyRequests:
description: Rate limit exceeded.
content:
application/json:
schema:
$ref: '#/components/schemas/Error'
schemas:
Error:
type: object
properties:
error:
type: object
properties:
message:
type: string
code:
type: string
ChatMessage:
type: object
required:
- role
- content
properties:
role:
type: string
enum:
- system
- user
- assistant
- tool
content:
type: string
ChatCompletionRequest:
type: object
required:
- model
- messages
properties:
model:
type: string
description: Sarvam LLM identifier.
enum:
- sarvam-m
- sarvam-30b
- sarvam-105b
example: sarvam-m
messages:
type: array
items:
$ref: '#/components/schemas/ChatMessage'
temperature:
type: number
default: 0.7
top_p:
type: number
default: 1
max_tokens:
type: integer
stream:
type: boolean
default: false
description: When true, partial deltas are streamed as Server-Sent Events.
reasoning_effort:
type: string
enum:
- low
- medium
- high
frequency_penalty:
type: number
presence_penalty:
type: number
n:
type: integer
default: 1
stop:
type: array
items:
type: string
seed:
type: integer
wiki_grounding:
type: boolean
description: Ground responses with Wikipedia knowledge.
tools:
type: array
items:
type: object
tool_choice:
oneOf:
- type: string
- type: object
response_format:
type: object
ChatCompletionResponse:
type: object
properties:
id:
type: string
object:
type: string
example: chat.completion
created:
type: integer
model:
type: string
choices:
type: array
items:
type: object
properties:
index:
type: integer
message:
$ref: '#/components/schemas/ChatMessage'
finish_reason:
type: string
usage:
type: object
properties:
prompt_tokens:
type: integer
completion_tokens:
type: integer
total_tokens:
type: integer
SpeechToTextRequest:
type: object
required:
- file
properties:
file:
type: string
format: binary
description: >-
Audio file (WAV, MP3, AAC, AIFF, OGG, OPUS, FLAC, MP4/M4A, AMR, WMA,
WebM, PCM). Optimized for 16kHz.
model:
type: string
enum:
- saarika:v2.5
- saaras:v3
default: saarika:v2.5
language_code:
type: string
description: BCP-47 code (e.g. hi-IN) or "unknown" for auto-detection.
mode:
type: string
description: saaras:v3 only.
enum:
- transcribe
- translate
- verbatim
- translit
- codemix
input_audio_codec:
type: string
description: Required only for PCM files.
enum:
- pcm_s16le
- pcm_l16
- pcm_raw
SpeechToTextTranslateRequest:
type: object
required:
- file
properties:
file:
type: string
format: binary
description: Audio file to transcribe and translate to English.
model:
type: string
enum:
- saaras:v2.5
- saaras:v3
default: saaras:v2.5
prompt:
type: string
description: Optional context to improve translation quality.
SpeechToTextResponse:
type: object
properties:
request_id:
type: string
transcript:
type: string
language_code:
type: string
language_probability:
type: number
timestamps:
type: object
description: Word-level timing data when requested.
diarized_transcript:
type: object
description: Speaker-attributed segments when diarization is enabled.
TextToSpeechRequest:
type: object
required:
- text
- target_language_code
properties:
text:
type: string
description: Max 2500 chars (bulbul:v3) or 1500 chars (bulbul:v2).
target_language_code:
type: string
enum:
- bn-IN
- en-IN
- gu-IN
- hi-IN
- kn-IN
- ml-IN
- mr-IN
- od-IN
- pa-IN
- ta-IN
- te-IN
speaker:
type: string
description: Voice name; default shubh (v3) or anushka (v2).
model:
type: string
enum:
- bulbul:v2
- bulbul:v3
default: bulbul:v3
pitch:
type: number
description: -0.75 to 0.75. bulbul:v2 only.
pace:
type: number
default: 1.0
description: v3 0.5-2.0; v2 0.3-3.0.
loudness:
type: number
description: 0.3 to 3.0. bulbul:v2 only.
speech_sample_rate:
type: integer
default: 24000
enum:
- 8000
- 16000
- 22050
- 24000
- 32000
- 44100
- 48000
enable_preprocessing:
type: boolean
default: false
temperature:
type: number
default: 0.6
description: 0.01 to 2.0. v3 only.
output_audio_codec:
type: string
default: wav
enum:
- mp3
- linear16
- mulaw
- alaw
- opus
- flac
- aac
- wav
TextToSpeechResponse:
type: object
properties:
request_id:
type: string
audios:
type: array
items:
type: string
description: Base64-encoded audio.
TranslateRequest:
type: object
required:
- input
- source_language_code
- target_language_code
properties:
input:
type: string
description: Text to translate (up to ~1000-2000 chars).
source_language_code:
type: string
description: BCP-47 code or "auto" for detection.
target_language_code:
type: string
speaker_gender:
type: string
enum:
- Male
- Female
mode:
type: string
enum:
- formal
- modern-colloquial
- classic-colloquial
- code-mixed
model:
type: string
enum:
- mayura:v1
- sarvam-translate:v1
output_script:
type: string
enum:
- roman
- fully-native
- spoken-form-in-native
numerals_format:
type: string
enum:
- international
- native
default: international
TranslateResponse:
type: object
properties:
request_id:
type: string
translated_text:
type: string
source_language_code:
type: string
TransliterateRequest:
type: object
required:
- input
- source_language_code
- target_language_code
properties:
input:
type: string
source_language_code:
type: string
target_language_code:
type: string
numerals_format:
type: string
enum:
- international
- native
default: international
spoken_form:
type: boolean
default: false
spoken_form_numerals_language:
type: string
enum:
- english
- native
default: native
TransliterateResponse:
type: object
properties:
request_id:
type: string
transliterated_text:
type: string
source_language_code:
type: string
LanguageIdRequest:
type: object
required:
- input
properties:
input:
type: string
LanguageIdResponse:
type: object
properties:
request_id:
type: string
language_code:
type: string
script_code:
type: string