feat: add Perplexity as a first-class API provider

Introduce a new Perplexity provider that targets Perplexity's
OpenAI-compatible chat-completions endpoint at https://api.perplexity.ai.
Models exposed: sonar, sonar-pro, sonar-reasoning, sonar-reasoning-pro
(all 128k context).

API key resolution order:
  1. perplexityApiKey from settings
  2. PERPLEXITY_API_KEY env var
  3. PPLX_API_KEY env var (fallback)

Wired into the provider registry (packages/types), api factory,
profile validation, webview settings UI, model picker, validation,
and English locale strings.

Tests cover construction, base URL/auth, env-var fallbacks, model
selection, streaming, and error propagation.

Author: james-pplx

packages/types/src/global-settings.ts

@@ -275,6 +275,7 @@ export const SECRET_STATE_KEYS = [
 	"sambaNovaApiKey",
 	"zaiApiKey",
 	"fireworksApiKey",
+	"perplexityApiKey",
 	"vercelAiGatewayApiKey",
 	"basetenApiKey",
 ] as const

packages/types/src/provider-settings.ts

@@ -13,6 +13,7 @@ import {
 	moonshotModels,
 	openAiCodexModels,
 	openAiNativeModels,
+	perplexityModels,
 	qwenCodeModels,
 	sambaNovaModels,
 	vertexModels,
@@ -114,6 +115,7 @@ export const providerNames = [
 	"minimax",
 	"openai-codex",
 	"openai-native",
+	"perplexity",
 	"qwen-code",
 	"sambanova",
 	"vertex",
@@ -368,6 +370,10 @@ const fireworksSchema = apiModelIdProviderModelSchema.extend({
 	fireworksApiKey: z.string().optional(),
 })
 
+const perplexitySchema = apiModelIdProviderModelSchema.extend({
+	perplexityApiKey: z.string().optional(),
+})
+
 const qwenCodeSchema = apiModelIdProviderModelSchema.extend({
 	qwenCodeOauthPath: z.string().optional(),
 })
@@ -412,6 +418,7 @@ export const providerSettingsSchemaDiscriminated = z.discriminatedUnion("apiProv
 	sambaNovaSchema.merge(z.object({ apiProvider: z.literal("sambanova") })),
 	zaiSchema.merge(z.object({ apiProvider: z.literal("zai") })),
 	fireworksSchema.merge(z.object({ apiProvider: z.literal("fireworks") })),
+	perplexitySchema.merge(z.object({ apiProvider: z.literal("perplexity") })),
 	qwenCodeSchema.merge(z.object({ apiProvider: z.literal("qwen-code") })),
 	vercelAiGatewaySchema.merge(z.object({ apiProvider: z.literal("vercel-ai-gateway") })),
 	defaultSchema,
@@ -445,6 +452,7 @@ export const providerSettingsSchema = z.object({
 	...sambaNovaSchema.shape,
 	...zaiSchema.shape,
 	...fireworksSchema.shape,
+	...perplexitySchema.shape,
 	...qwenCodeSchema.shape,
 	...vercelAiGatewaySchema.shape,
 	...codebaseIndexProviderSchema.shape,
@@ -520,6 +528,7 @@ export const modelIdKeysByProvider: Record<TypicalProvider, ModelIdKey> = {
 	sambanova: "apiModelId",
 	zai: "apiModelId",
 	fireworks: "apiModelId",
+	perplexity: "apiModelId",
 	"vercel-ai-gateway": "vercelAiGatewayModelId",
 }
 
@@ -575,6 +584,11 @@ export const MODELS_BY_PROVIDER: Record<
 		label: "Fireworks",
 		models: Object.keys(fireworksModels),
 	},
+	perplexity: {
+		id: "perplexity",
+		label: "Perplexity",
+		models: Object.keys(perplexityModels),
+	},
 	gemini: {
 		id: "gemini",
 		label: "Google Gemini",

packages/types/src/providers/index.ts

@@ -13,6 +13,7 @@ export * from "./openai.js"
 export * from "./openai-codex.js"
 export * from "./openai-codex-rate-limits.js"
 export * from "./openrouter.js"
+export * from "./perplexity.js"
 export * from "./poe.js"
 export * from "./qwen-code.js"
 export * from "./requesty.js"
@@ -36,6 +37,7 @@ import { mistralDefaultModelId } from "./mistral.js"
 import { moonshotDefaultModelId } from "./moonshot.js"
 import { openAiCodexDefaultModelId } from "./openai-codex.js"
 import { openRouterDefaultModelId } from "./openrouter.js"
+import { perplexityDefaultModelId } from "./perplexity.js"
 import { poeDefaultModelId } from "./poe.js"
 import { qwenCodeDefaultModelId } from "./qwen-code.js"
 import { requestyDefaultModelId } from "./requesty.js"
@@ -103,6 +105,8 @@ export function getProviderDefaultModelId(
 			return sambaNovaDefaultModelId
 		case "fireworks":
 			return fireworksDefaultModelId
+		case "perplexity":
+			return perplexityDefaultModelId
 		case "qwen-code":
 			return qwenCodeDefaultModelId
 		case "poe":

packages/types/src/providers/perplexity.ts

@@ -0,0 +1,50 @@
+import type { ModelInfo } from "../model.js"
+
+// Perplexity
+// https://docs.perplexity.ai/docs/getting-started
+// https://docs.perplexity.ai/guides/pricing
+export type PerplexityModelId = keyof typeof perplexityModels
+
+export const perplexityDefaultModelId: PerplexityModelId = "sonar-pro"
+
+export const perplexityModels = {
+	sonar: {
+		maxTokens: 8192,
+		contextWindow: 128_000,
+		supportsImages: false,
+		supportsPromptCache: false,
+		inputPrice: 1.0,
+		outputPrice: 1.0,
+		description:
+			"Lightweight, cost-effective model with built-in web search grounding. Best for quick lookups and short answers.",
+	},
+	"sonar-pro": {
+		maxTokens: 8192,
+		contextWindow: 128_000,
+		supportsImages: false,
+		supportsPromptCache: false,
+		inputPrice: 3.0,
+		outputPrice: 15.0,
+		description:
+			"Perplexity's flagship model with built-in web search grounding. Best for complex queries that benefit from up-to-date information.",
+	},
+	"sonar-reasoning": {
+		maxTokens: 8192,
+		contextWindow: 128_000,
+		supportsImages: false,
+		supportsPromptCache: false,
+		inputPrice: 1.0,
+		outputPrice: 5.0,
+		description: "Reasoning model with chain-of-thought and built-in web search grounding.",
+	},
+	"sonar-reasoning-pro": {
+		maxTokens: 8192,
+		contextWindow: 128_000,
+		supportsImages: false,
+		supportsPromptCache: false,
+		inputPrice: 2.0,
+		outputPrice: 8.0,
+		description:
+			"Reasoning model with extended chain-of-thought reasoning and built-in web search grounding for complex multi-step problems.",
+	},
+} as const satisfies Record<string, ModelInfo>

src/api/index.ts

@@ -30,6 +30,7 @@ import {
 	SambaNovaHandler,
 	ZAiHandler,
 	FireworksHandler,
+	PerplexityHandler,
 	VercelAiGatewayHandler,
 	MiniMaxHandler,
 	BasetenHandler,
@@ -169,6 +170,8 @@ export function buildApiHandler(configuration: ProviderSettings): ApiHandler {
 			return new ZAiHandler(options)
 		case "fireworks":
 			return new FireworksHandler(options)
+		case "perplexity":
+			return new PerplexityHandler(options)
 		case "vercel-ai-gateway":
 			return new VercelAiGatewayHandler(options)
 		case "minimax":

src/api/providers/__tests__/base-openai-compatible-provider.spec.ts

@@ -326,6 +326,42 @@ describe("BaseOpenAiCompatibleProvider", () => {
 			// Should yield reasoning with spaces (only pure whitespace is filtered)
 			expect(chunks).toEqual([{ type: "reasoning", text: "  content with spaces  " }])
 		})
+
+		it("should yield reasoning_content before content when both are present in a delta", async () => {
+			mockCreate.mockImplementationOnce(() => {
+				return {
+					[Symbol.asyncIterator]: () => ({
+						next: vi
+							.fn()
+							.mockResolvedValueOnce({
+								done: false,
+								value: {
+									choices: [
+										{
+											delta: {
+												reasoning_content: "Thinking first",
+												content: "Final answer",
+											},
+										},
+									],
+								},
+							})
+							.mockResolvedValueOnce({ done: true }),
+					}),
+				}
+			})
+
+			const stream = handler.createMessage("system prompt", [])
+			const chunks = []
+			for await (const chunk of stream) {
+				chunks.push(chunk)
+			}
+
+			expect(chunks).toEqual([
+				{ type: "reasoning", text: "Thinking first" },
+				{ type: "text", text: "Final answer" },
+			])
+		})
 	})
 
 	describe("Basic functionality", () => {