feat (docs): add Sarvam community provider (#6079) (#6096)

Co-authored-by: Mrinank Bhowmick <[email protected]>

Author: samdenty

content/docs/02-foundations/02-providers-and-models.mdx

@@ -72,6 +72,7 @@ The open-source community has created the following providers:
 - [Spark Provider](/providers/community-providers/spark) (`spark-ai-provider`)
 - [AnthropicVertex Provider](/providers/community-providers/anthropic-vertex-ai) (`anthropic-vertex-ai`)
 - [LangDB Provider](/providers/community-providers/langdb) (`@langdb/vercel-provider`)
+- [Sarvam Provider](/providers/community-providers/sarvam) (`sarvam-ai-provider`)
 
 ## Self-Hosted Models
 

content/providers/03-community-providers/97-sarvam.mdx

@@ -0,0 +1,139 @@
+---
+title: 'Sarvam'
+description: 'Learn how to use the Sarvam AI provider for the AI SDK.'
+---
+
+# Sarvam Provider
+
+The Sarvam AI Provider is a library developed to integrate with the AI SDK. This library brings Speech to Text (STT) capabilities to your applications, allowing for seamless interaction with audio and text data.
+
+## Setup
+
+The Sarvam provider is available in the `sarvam-ai-provider` module. You can install it with:
+
+<Tabs items={['pnpm', 'npm', 'yarn']}>
+  <Tab>
+    <Snippet text="pnpm add sarvam-ai-provider" dark />
+  </Tab>
+  
+  <Tab>
+    <Snippet text="npm install sarvam-ai-provider" dark />
+  </Tab>
+
+  <Tab>
+    <Snippet text="yarn add sarvam-ai-provider" dark />
+  </Tab>
+</Tabs>
+
+## Provider Instance
+
+First, get your **Sarvam API Key** from the [Sarvam Dashboard](https://dashboard.sarvam.ai/auth/signin).
+
+Then initialize `Sarvam` in your application:
+
+```ts
+import { createSarvam } from 'sarvam-ai-provider';
+
+const sarvam = createSarvam({
+  headers: {
+    'api-subscription-key': 'YOUR_API_KEY',
+  },
+});
+```
+
+<Note>
+  The `api-subscription-key` needs to be passed in headers. Consider using
+  `YOUR_API_KEY` as environment variables for security.
+</Note>
+
+- Transcribe speech to text
+
+```ts
+import { experimental_transcribe as transcribe } from 'ai';
+import { readFile } from 'fs/promises';
+
+await transcribe({
+  model: sarvam.transcription('saarika:v2'),
+  audio: await readFile('./src/transcript-test.mp3'),
+  providerOptions: {
+    sarvam: {
+      language_code: 'en-IN',
+    },
+  },
+});
+```
+
+## Features
+
+### Changing parameters
+
+- Change language_code
+
+```ts
+providerOptions: {
+    sarvam: {
+      language_code: 'en-IN',
+    },
+  },
+```
+
+<Note>
+  `language_code` specifies the language of the input audio and is required for
+  accurate transcription. • It is mandatory for the `saarika:v1` model (this
+  model does not support `unknown`). • It is optional for the `saarika:v2`
+  model. • Use `unknown` when the language is not known; in that case, the API
+  will auto‑detect it. Available options: `unknown`, `hi-IN`, `bn-IN`, `kn-IN`,
+  `ml-IN`, `mr-IN`, `od-IN`, `pa-IN`, `ta-IN`, `te-IN`, `en-IN`, `gu-IN`.
+</Note>
+
+- with_timestamps?
+
+```ts
+providerOptions: {
+  sarvam: {
+    with_timestamps: true,
+  },
+},
+```
+
+<Note>
+  `with_timestamps` specifies whether to include start/end timestamps for each
+  word/token. • Type: boolean • When true, each word/token will include
+  start/end timestamps. • Default: false
+</Note>
+
+- with_diarization?
+
+```ts
+providerOptions: {
+  sarvam: {
+    with_diarization: true,
+  },
+},
+```
+
+<Note>
+  `with_diarization` enables speaker diarization (Beta). • Type: boolean • When
+  true, enables speaker diarization. • Default: false
+</Note>
+
+- num_speakers?
+
+```ts
+providerOptions: {
+  sarvam: {
+    with_diarization: true,
+    num_speakers: 2,
+  },
+},
+```
+
+<Note>
+  `num_speakers` sets the number of distinct speakers to detect (only when
+  `with_diarization` is true). • Type: number | null • Number of distinct
+  speakers to detect. • Default: null
+</Note>
+
+## References
+
+- [Sarvam API Docs](https://docs.sarvam.ai/api-reference-docs/endpoints/speech-to-text)