feat (provider/perplexity): rewrite provider and support sources (#4864)

.changeset/forty-readers-clap.md

@@ -0,0 +1,5 @@
+---
+'@ai-sdk/perplexity': major
+---
+
+feat (provider/perplexity): rewrite provider and support sources

.changeset/rude-actors-fly.md

@@ -0,0 +1,6 @@
+---
+'@ai-sdk/provider-utils': patch
+'ai': patch
+---
+
+chore (ai): move mockId test helper into provider utils

content/providers/01-ai-sdk-providers/70-perplexity.mdx

@@ -77,12 +77,26 @@ const { text } = await generateText({
 });
 ```
 
-Perplexity models can be used in the `streamText` and `streamUI` functions (see
-[AI SDK Core](/docs/ai-sdk-core) and [AI SDK RSC](/docs/ai-sdk-rsc)).
+### Sources
 
-### Provider Metadata
+Websites that have been used to generate the response are included in the `sources` property of the result:
 
-The Perplexity provider includes additional experimental metadata in the response through `providerMetadata`:
+```ts
+import { perplexity } from '@ai-sdk/perplexity';
+import { generateText } from 'ai';
+
+const { text, sources } = await generateText({
+  model: perplexity('sonar-pro'),
+  prompt: 'What are the latest developments in quantum computing?',
+});
+
+console.log(sources);
+```
+
+### Provider Options & Metadata
+
+The Perplexity provider includes additional metadata in the response through `providerMetadata`.
+Additional configuration options are available through `providerOptions`.
 
 ```ts
 const result = await generateText({
@@ -99,10 +113,6 @@ console.log(result.providerMetadata);
 // Example output:
 // {
 //   perplexity: {
-//     citations: [
-//       'https://www.sfchronicle.com',
-//       'https://www.cbsnews.com/sanfrancisco/',
-//     ],
 //     usage: { citationTokens: 5286, numSearchQueries: 1 },
 //     images: [
 //       { imageUrl: "https://example.com/image1.jpg", originUrl: "https://elsewhere.com/page1", height: 1280, width: 720 },
@@ -114,7 +124,6 @@ console.log(result.providerMetadata);
 
 The metadata includes:
 
-- `citations`: Array of URLs used as sources for the response
 - `usage`: Object containing `citationTokens` and `numSearchQueries` metrics
 - `images`: Array of image URLs when `return_images` is enabled (Tier-2 users only)
 
@@ -132,14 +141,6 @@ You can enable image responses by setting `return_images: true` in the provider
 | `sonar-pro` | <Cross size={18} /> | <Cross size={18} /> | <Cross size={18} /> | <Cross size={18} /> |
 | `sonar`     | <Cross size={18} /> | <Cross size={18} /> | <Cross size={18} /> | <Cross size={18} /> |
 
-### Key Features
-
-- **Real-time Web Search**: Both models provide grounded responses using real-time web search
-- **Citations**: Sonar Pro provides 2x more citations than standard Sonar
-- **Data Privacy**: No training on customer data
-- **Self-serve API**: Immediate access with scalable pricing
-- **Advanced Queries**: Support for complex queries and follow-up questions
-
 <Note>
   Please see the [Perplexity docs](https://docs.perplexity.ai) for detailed API
   documentation and the latest updates.

examples/ai-core/src/e2e/perplexity.test.ts

@@ -1,28 +1,25 @@
 import 'dotenv/config';
 import { expect } from 'vitest';
-import {
-  perplexity as provider,
-  PerplexityErrorData,
-} from '@ai-sdk/perplexity';
+import { perplexity as provider } from '@ai-sdk/perplexity';
 import {
   createFeatureTestSuite,
   createLanguageModelWithCapabilities,
 } from './feature-test-suite';
 import { APICallError } from '@ai-sdk/provider';
 
 const createChatModel = (modelId: string) =>
-  createLanguageModelWithCapabilities(provider.chat(modelId));
+  createLanguageModelWithCapabilities(provider(modelId));
 
 createFeatureTestSuite({
   name: 'perplexity',
   models: {
-    invalidModel: provider.chat('no-such-model'),
+    invalidModel: provider('no-such-model'),
     languageModels: [createChatModel('sonar-pro'), createChatModel('sonar')],
   },
   timeout: 30000,
   customAssertions: {
     errorValidator: (error: APICallError) => {
-      expect((error.data as PerplexityErrorData).code).toBe(
+      expect((error.data as any).code).toBe(
         'Some requested entity was not found',
       );
     },

examples/ai-core/src/generate-object/perplexity.ts

@@ -0,0 +1,31 @@
+import 'dotenv/config';
+import { perplexity } from '@ai-sdk/perplexity';
+import { generateObject, generateText } from 'ai';
+import { z } from 'zod';
+
+async function main() {
+  const result = await generateObject({
+    model: perplexity('sonar-pro'),
+    prompt: 'What has happened in San Francisco recently?',
+    providerOptions: {
+      perplexity: {
+        search_recency_filter: 'week',
+      },
+    },
+    output: 'array',
+    schema: z.object({
+      title: z.string(),
+      summary: z.string(),
+    }),
+  });
+
+  console.log(result.object);
+  console.log();
+  console.log('Token usage:', result.usage);
+  console.log('Finish reason:', result.finishReason);
+  console.log('Metadata:', result.providerMetadata);
+}
+
+main().catch((error: Error) => {
+  console.error(JSON.stringify(error, null, 2));
+});

examples/ai-core/src/generate-text/perplexity.ts

@@ -5,11 +5,17 @@ import { generateText } from 'ai';
 async function main() {
   const result = await generateText({
     model: perplexity('sonar-pro'),
-    prompt: 'Invent a new holiday and describe its traditions.',
+    prompt: 'What has happened in San Francisco recently?',
+    providerOptions: {
+      perplexity: {
+        search_recency_filter: 'week',
+      },
+    },
   });
 
   console.log(result.text);
   console.log();
+  console.log('Sources:', result.sources);
   console.log('Token usage:', result.usage);
   console.log('Finish reason:', result.finishReason);
   console.log('Metadata:', result.providerMetadata);

examples/ai-core/src/stream-text/perplexity.ts

@@ -5,16 +5,20 @@ import 'dotenv/config';
 async function main() {
   const result = streamText({
     model: perplexity('sonar-pro'),
-    prompt:
-      'List the top 5 San Francisco news from the past week.' +
-      'You must include the date of each article.',
+    prompt: 'What has happened in San Francisco recently?',
+    providerOptions: {
+      perplexity: {
+        search_recency_filter: 'week',
+      },
+    },
   });
 
   for await (const textPart of result.textStream) {
     process.stdout.write(textPart);
   }
 
   console.log();
+  console.log('Sources:', await result.sources);
   console.log('Token usage:', await result.usage);
   console.log('Finish reason:', await result.finishReason);
   console.log(

examples/next-openai/app/api/use-chat-sources/route.ts

@@ -1,4 +1,5 @@
 import { vertex } from '@ai-sdk/google-vertex';
+import { perplexity } from '@ai-sdk/perplexity';
 import { streamText } from 'ai';
 
 // Allow streaming responses up to 30 seconds
@@ -8,7 +9,8 @@ export async function POST(req: Request) {
   const { messages } = await req.json();
 
   const result = streamText({
-    model: vertex('gemini-1.5-flash', { useSearchGrounding: true }),
+    // model: vertex('gemini-1.5-flash', { useSearchGrounding: true }),
+    model: perplexity('sonar-pro'),
     messages,
   });
 

examples/next-openai/app/use-chat-sources/page.tsx

@@ -21,27 +21,28 @@ export default function Chat() {
       {messages.map(message => (
         <div key={message.id} className="whitespace-pre-wrap">
           {message.role === 'user' ? 'User: ' : 'AI: '}
-          {message.parts.map((part, index) => {
-            if (part.type === 'text') {
-              return <div key={index}>{part.text}</div>;
-            }
-
-            if (part.type === 'source') {
-              return (
-                <span key={`source-${part.source.id}`}>
-                  [
-                  <a
-                    href={part.source.url}
-                    target="_blank"
-                    className="text-sm font-bold text-blue-500 hover:underline"
-                  >
-                    {part.source.title}
-                  </a>
-                  ]
-                </span>
-              );
-            }
-          })}
+          {message.parts
+            .filter(part => part.type !== 'source')
+            .map((part, index) => {
+              if (part.type === 'text') {
+                return <div key={index}>{part.text}</div>;
+              }
+            })}
+          {message.parts
+            .filter(part => part.type === 'source')
+            .map(part => (
+              <span key={`source-${part.source.id}`}>
+                [
+                <a
+                  href={part.source.url}
+                  target="_blank"
+                  className="text-sm font-bold text-blue-500 hover:underline"
+                >
+                  {part.source.title ?? new URL(part.source.url).hostname}
+                </a>
+                ]
+              </span>
+            ))}
         </div>
       ))}
 

examples/next-openai/package.json

@@ -13,6 +13,7 @@
     "@ai-sdk/deepseek": "0.1.9",
     "@ai-sdk/openai": "1.1.10",
     "@ai-sdk/google-vertex": "2.1.13",
+    "@ai-sdk/perplexity": "0.0.8",
     "@ai-sdk/ui-utils": "1.1.13",
     "@ai-sdk/react": "1.1.13",
     "@vercel/blob": "^0.26.0",

packages/ai/core/generate-text/generate-text.test.ts

@@ -1,10 +1,10 @@
 import { LanguageModelV1CallOptions } from '@ai-sdk/provider';
+import { mockId } from '@ai-sdk/provider-utils/test';
 import { jsonSchema } from '@ai-sdk/ui-utils';
 import assert from 'node:assert';
 import { z } from 'zod';
 import { Output } from '.';
 import { ToolExecutionError } from '../../errors';
-import { mockId } from '../test/mock-id';
 import { MockLanguageModelV1 } from '../test/mock-language-model-v1';
 import { MockTracer } from '../test/mock-tracer';
 import { tool } from '../tool/tool';

packages/ai/core/generate-text/stream-text.test.ts

@@ -10,14 +10,14 @@ import {
   convertAsyncIterableToArray,
   convertReadableStreamToArray,
   convertResponseStreamToArray,
+  mockId,
 } from '@ai-sdk/provider-utils/test';
 import { jsonSchema } from '@ai-sdk/ui-utils';
 import assert from 'node:assert';
 import { z } from 'zod';
 import { ToolExecutionError } from '../../errors/tool-execution-error';
 import { StreamData } from '../../streams/stream-data';
 import { createDataStream } from '../data-stream/create-data-stream';
-import { mockId } from '../test/mock-id';
 import { MockLanguageModelV1 } from '../test/mock-language-model-v1';
 import { createMockServerResponse } from '../test/mock-server-response';
 import { MockTracer } from '../test/mock-tracer';

packages/ai/core/middleware/extract-reasoning-middleware.test.ts

@@ -1,10 +1,10 @@
 import {
   convertArrayToReadableStream,
   convertAsyncIterableToArray,
+  mockId,
 } from '@ai-sdk/provider-utils/test';
 import { generateText, streamText } from '../generate-text';
 import { wrapLanguageModel } from '../middleware/wrap-language-model';
-import { mockId } from '../test/mock-id';
 import { MockLanguageModelV1 } from '../test/mock-language-model-v1';
 import { extractReasoningMiddleware } from './extract-reasoning-middleware';
 

packages/ai/test/index.ts

@@ -1,6 +1,8 @@
-export { convertArrayToReadableStream } from '@ai-sdk/provider-utils/test';
+export {
+  convertArrayToReadableStream,
+  mockId,
+} from '@ai-sdk/provider-utils/test';
 export { MockEmbeddingModelV1 } from '../core/test/mock-embedding-model-v1';
-export { mockId } from '../core/test/mock-id';
 export { MockLanguageModelV1 } from '../core/test/mock-language-model-v1';
 export { mockValues } from '../core/test/mock-values';
 

packages/perplexity/package.json

@@ -30,7 +30,6 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/openai-compatible": "0.1.9",
     "@ai-sdk/provider": "1.0.7",
     "@ai-sdk/provider-utils": "2.1.7"
   },

packages/perplexity/src/__snapshots__/convert-to-perplexity-messages.test.ts.snap

@@ -0,0 +1,28 @@
+// Vitest Snapshot v1, https://vitest.dev/guide/snapshot.html
+
+exports[`convertToPerplexityMessages > assistant messages > should convert an assistant message with text content 1`] = `
+[
+  {
+    "content": "Assistant reply",
+    "role": "assistant",
+  },
+]
+`;
+
+exports[`convertToPerplexityMessages > system messages > should convert a system message with text content 1`] = `
+[
+  {
+    "content": "System initialization",
+    "role": "system",
+  },
+]
+`;
+
+exports[`convertToPerplexityMessages > user messages > should convert a user message with text parts 1`] = `
+[
+  {
+    "content": "Hello World",
+    "role": "user",
+  },
+]
+`;