feat: add dynamic model loading for Roo Code Cloud provider (RooCodeInc#8728)

Co-authored-by: Roo Code <[email protected]>
Co-authored-by: Matt Rubens <[email protected]>

Author: 3 people

packages/types/src/provider-settings.ts

@@ -19,7 +19,6 @@ import {
 	moonshotModels,
 	openAiNativeModels,
 	qwenCodeModels,
-	rooModels,
 	sambaNovaModels,
 	vertexModels,
 	vscodeLlmModels,
@@ -49,6 +48,7 @@ export const dynamicProviders = [
 	"requesty",
 	"unbound",
 	"glama",
+	"roo",
 ] as const
 
 export type DynamicProvider = (typeof dynamicProviders)[number]
@@ -677,7 +677,7 @@ export const MODELS_BY_PROVIDER: Record<
 		models: Object.keys(openAiNativeModels),
 	},
 	"qwen-code": { id: "qwen-code", label: "Qwen Code", models: Object.keys(qwenCodeModels) },
-	roo: { id: "roo", label: "Roo", models: Object.keys(rooModels) },
+	roo: { id: "roo", label: "Roo Code Cloud", models: [] },
 	sambanova: {
 		id: "sambanova",
 		label: "SambaNova",

packages/types/src/providers/roo.ts

@@ -1,53 +1,49 @@
+import { z } from "zod"
+
 import type { ModelInfo } from "../model.js"
 
-export type RooModelId =
-	| "xai/grok-code-fast-1"
-	| "roo/code-supernova-1-million"
-	| "xai/grok-4-fast"
-	| "deepseek/deepseek-chat-v3.1"
-
-export const rooDefaultModelId: RooModelId = "xai/grok-code-fast-1"
-
-export const rooModels = {
-	"xai/grok-code-fast-1": {
-		maxTokens: 16_384,
-		contextWindow: 262_144,
-		supportsImages: false,
-		supportsPromptCache: true,
-		inputPrice: 0,
-		outputPrice: 0,
-		description:
-			"A reasoning model that is blazing fast and excels at agentic coding, accessible for free through Roo Code Cloud for a limited time. (Note: the free prompts and completions are logged by xAI and used to improve the model.)",
-	},
-	"roo/code-supernova-1-million": {
-		maxTokens: 30_000,
-		contextWindow: 1_000_000,
-		supportsImages: true,
-		supportsPromptCache: true,
-		inputPrice: 0,
-		outputPrice: 0,
-		description:
-			"A versatile agentic coding stealth model with a 1M token context window that supports image inputs, accessible for free through Roo Code Cloud for a limited time. (Note: the free prompts and completions are logged by the model provider and used to improve the model.)",
-	},
-	"xai/grok-4-fast": {
-		maxTokens: 30_000,
-		contextWindow: 2_000_000,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0,
-		outputPrice: 0,
-		description:
-			"Grok 4 Fast is xAI's latest multimodal model with SOTA cost-efficiency and a 2M token context window. (Note: prompts and completions are logged by xAI and used to improve the model.)",
-		deprecated: true,
-	},
-	"deepseek/deepseek-chat-v3.1": {
-		maxTokens: 16_384,
-		contextWindow: 163_840,
-		supportsImages: false,
-		supportsPromptCache: false,
-		inputPrice: 0,
-		outputPrice: 0,
-		description:
-			"DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active). It extends the DeepSeek-V3 base with a two-phase long-context training process, reaching up to 128K tokens, and uses FP8 microscaling for efficient inference.",
-	},
-} as const satisfies Record<string, ModelInfo>
+/**
+ * Roo Code Cloud is a dynamic provider - models are loaded from the /v1/models API endpoint.
+ * Default model ID used as fallback when no model is specified.
+ */
+export const rooDefaultModelId = "xai/grok-code-fast-1"
+
+/**
+ * Empty models object maintained for type compatibility.
+ * All model data comes dynamically from the API.
+ */
+export const rooModels = {} as const satisfies Record<string, ModelInfo>
+
+/**
+ * Roo Code Cloud API response schemas
+ */
+
+export const RooPricingSchema = z.object({
+	input: z.string(),
+	output: z.string(),
+	input_cache_read: z.string().optional(),
+	input_cache_write: z.string().optional(),
+})
+
+export const RooModelSchema = z.object({
+	id: z.string(),
+	object: z.literal("model"),
+	created: z.number(),
+	owned_by: z.string(),
+	name: z.string(),
+	description: z.string(),
+	context_window: z.number(),
+	max_tokens: z.number(),
+	type: z.literal("language"),
+	tags: z.array(z.string()).optional(),
+	pricing: RooPricingSchema,
+	deprecated: z.boolean().optional(),
+})
+
+export const RooModelsResponseSchema = z.object({
+	object: z.literal("list"),
+	data: z.array(RooModelSchema),
+})
+
+export type RooModel = z.infer<typeof RooModelSchema>
+export type RooModelsResponse = z.infer<typeof RooModelsResponseSchema>

src/api/providers/__tests__/roo.spec.ts

@@ -1,7 +1,7 @@
 // npx vitest run api/providers/__tests__/roo.spec.ts
 
 import { Anthropic } from "@anthropic-ai/sdk"
-import { rooDefaultModelId, rooModels } from "@roo-code/types"
+import { rooDefaultModelId } from "@roo-code/types"
 
 import { ApiHandlerOptions } from "../../../shared/api"
 
@@ -301,16 +301,19 @@ describe("RooHandler", () => {
 			const modelInfo = handler.getModel()
 			expect(modelInfo.id).toBe(mockOptions.apiModelId)
 			expect(modelInfo.info).toBeDefined()
-			// xai/grok-code-fast-1 is a valid model in rooModels
-			expect(modelInfo.info).toBe(rooModels["xai/grok-code-fast-1"])
+			// Models are loaded dynamically, so we just verify the structure
+			expect(modelInfo.info.maxTokens).toBeDefined()
+			expect(modelInfo.info.contextWindow).toBeDefined()
 		})
 
 		it("should return default model when no model specified", () => {
 			const handlerWithoutModel = new RooHandler({})
 			const modelInfo = handlerWithoutModel.getModel()
 			expect(modelInfo.id).toBe(rooDefaultModelId)
 			expect(modelInfo.info).toBeDefined()
-			expect(modelInfo.info).toBe(rooModels[rooDefaultModelId])
+			// Models are loaded dynamically
+			expect(modelInfo.info.maxTokens).toBeDefined()
+			expect(modelInfo.info.contextWindow).toBeDefined()
 		})
 
 		it("should handle unknown model ID with fallback info", () => {
@@ -320,24 +323,27 @@ describe("RooHandler", () => {
 			const modelInfo = handlerWithUnknownModel.getModel()
 			expect(modelInfo.id).toBe("unknown-model-id")
 			expect(modelInfo.info).toBeDefined()
-			// Should return fallback info for unknown models
-			expect(modelInfo.info.maxTokens).toBe(16_384)
-			expect(modelInfo.info.contextWindow).toBe(262_144)
-			expect(modelInfo.info.supportsImages).toBe(false)
-			expect(modelInfo.info.supportsPromptCache).toBe(true)
-			expect(modelInfo.info.inputPrice).toBe(0)
-			expect(modelInfo.info.outputPrice).toBe(0)
+			// Should return fallback info for unknown models (dynamic models will be merged in real usage)
+			expect(modelInfo.info.maxTokens).toBeDefined()
+			expect(modelInfo.info.contextWindow).toBeDefined()
+			expect(modelInfo.info.supportsImages).toBeDefined()
+			expect(modelInfo.info.supportsPromptCache).toBeDefined()
+			expect(modelInfo.info.inputPrice).toBeDefined()
+			expect(modelInfo.info.outputPrice).toBeDefined()
 		})
 
-		it("should return correct model info for all Roo models", () => {
-			// Test each model in rooModels
-			const modelIds = Object.keys(rooModels) as Array<keyof typeof rooModels>
+		it("should handle any model ID since models are loaded dynamically", () => {
+			// Test with various model IDs - they should all work since models are loaded dynamically
+			const testModelIds = ["xai/grok-code-fast-1", "roo/sonic", "deepseek/deepseek-chat-v3.1"]
 
-			for (const modelId of modelIds) {
+			for (const modelId of testModelIds) {
 				const handlerWithModel = new RooHandler({ apiModelId: modelId })
 				const modelInfo = handlerWithModel.getModel()
 				expect(modelInfo.id).toBe(modelId)
-				expect(modelInfo.info).toBe(rooModels[modelId])
+				expect(modelInfo.info).toBeDefined()
+				// Verify the structure has required fields
+				expect(modelInfo.info.maxTokens).toBeDefined()
+				expect(modelInfo.info.contextWindow).toBeDefined()
 			}
 		})
 	})

src/api/providers/fetchers/modelCache.ts

@@ -24,6 +24,7 @@ import { getLMStudioModels } from "./lmstudio"
 import { getIOIntelligenceModels } from "./io-intelligence"
 import { getDeepInfraModels } from "./deepinfra"
 import { getHuggingFaceModels } from "./huggingface"
+import { getRooModels } from "./roo"
 
 const memoryCache = new NodeCache({ stdTTL: 5 * 60, checkperiod: 5 * 60 })
 
@@ -99,6 +100,13 @@ export const getModels = async (options: GetModelsOptions): Promise<ModelRecord>
 			case "huggingface":
 				models = await getHuggingFaceModels()
 				break
+			case "roo": {
+				// Roo Code Cloud provider requires baseUrl and optional apiKey
+				const rooBaseUrl =
+					options.baseUrl ?? process.env.ROO_CODE_PROVIDER_URL ?? "https://api.roocode.com/proxy"
+				models = await getRooModels(rooBaseUrl, options.apiKey)
+				break
+			}
 			default: {
 				// Ensures router is exhaustively checked if RouterName is a strict union.
 				const exhaustiveCheck: never = provider

src/api/providers/fetchers/roo.ts

@@ -0,0 +1,119 @@
+import { RooModelsResponseSchema } from "@roo-code/types"
+
+import type { ModelRecord } from "../../../shared/api"
+
+import { DEFAULT_HEADERS } from "../constants"
+
+/**
+ * Fetches available models from the Roo Code Cloud provider
+ *
+ * @param baseUrl The base URL of the Roo Code Cloud provider
+ * @param apiKey The API key (session token) for the Roo Code Cloud provider
+ * @returns A promise that resolves to a record of model IDs to model info
+ * @throws Will throw an error if the request fails or the response is not as expected.
+ */
+export async function getRooModels(baseUrl: string, apiKey?: string): Promise<ModelRecord> {
+	try {
+		const headers: Record<string, string> = {
+			"Content-Type": "application/json",
+			...DEFAULT_HEADERS,
+		}
+
+		if (apiKey) {
+			headers["Authorization"] = `Bearer ${apiKey}`
+		}
+
+		// Construct the models endpoint URL
+		// Strip trailing /v1 or /v1/ to avoid /v1/v1/models
+		const normalizedBase = baseUrl.replace(/\/?v1\/?$/, "")
+		const url = `${normalizedBase}/v1/models`
+
+		// Use fetch with AbortController for better timeout handling
+		const controller = new AbortController()
+		const timeoutId = setTimeout(() => controller.abort(), 10000)
+
+		try {
+			const response = await fetch(url, {
+				headers,
+				signal: controller.signal,
+			})
+
+			if (!response.ok) {
+				throw new Error(`HTTP ${response.status}: ${response.statusText}`)
+			}
+
+			const data = await response.json()
+			const models: ModelRecord = {}
+
+			// Validate response against schema
+			const parsed = RooModelsResponseSchema.safeParse(data)
+
+			if (!parsed.success) {
+				console.error("Error fetching Roo Code Cloud models: Unexpected response format", data)
+				console.error("Validation errors:", parsed.error.format())
+				throw new Error("Failed to fetch Roo Code Cloud models: Unexpected response format.")
+			}
+
+			// Process the validated model data
+			for (const model of parsed.data.data) {
+				const modelId = model.id
+
+				if (!modelId) continue
+
+				// Extract model data from the validated API response
+				// All required fields are guaranteed by the schema
+				const contextWindow = model.context_window
+				const maxTokens = model.max_tokens
+				const tags = model.tags || []
+				const pricing = model.pricing
+
+				// Determine if the model supports images based on tags
+				const supportsImages = tags.includes("vision")
+
+				// Parse pricing (API returns strings, convert to numbers)
+				const inputPrice = parseFloat(pricing.input)
+				const outputPrice = parseFloat(pricing.output)
+				const cacheReadPrice = pricing.input_cache_read ? parseFloat(pricing.input_cache_read) : undefined
+				const cacheWritePrice = pricing.input_cache_write ? parseFloat(pricing.input_cache_write) : undefined
+
+				models[modelId] = {
+					maxTokens,
+					contextWindow,
+					supportsImages,
+					supportsPromptCache: Boolean(cacheReadPrice !== undefined),
+					inputPrice,
+					outputPrice,
+					cacheWritesPrice: cacheWritePrice,
+					cacheReadsPrice: cacheReadPrice,
+					description: model.description || model.name,
+					deprecated: model.deprecated || false,
+				}
+			}
+
+			return models
+		} finally {
+			clearTimeout(timeoutId)
+		}
+	} catch (error: any) {
+		console.error("Error fetching Roo Code Cloud models:", error.message ? error.message : error)
+
+		// Handle abort/timeout
+		if (error.name === "AbortError") {
+			throw new Error("Failed to fetch Roo Code Cloud models: Request timed out after 10 seconds.")
+		}
+
+		// Handle fetch errors
+		if (error.message?.includes("HTTP")) {
+			throw new Error(`Failed to fetch Roo Code Cloud models: ${error.message}. Check base URL and API key.`)
+		}
+
+		// Handle network errors
+		if (error instanceof TypeError) {
+			throw new Error(
+				"Failed to fetch Roo Code Cloud models: No response from server. Check Roo Code Cloud server status and base URL.",
+			)
+		}
+
+		throw new Error(`Failed to fetch Roo Code Cloud models: ${error.message || "An unknown error occurred."}`)
+	}
+}