fix: prevent context mixing in Roo/Sonic model (fixes #7292)

- Disabled prompt caching for roo/sonic model to prevent cross-session contamination
- Added unique session IDs to each RooHandler instance for request isolation
- Added request-specific headers (X-Session-Id, X-Request-Id, X-No-Cache) to prevent caching
- Added comprehensive tests to verify session isolation is working correctly
- Updated model configuration to set supportsPromptCache to false

This fix addresses the issue where the Roo/Sonic model was giving completely unrelated responses, suggesting context was being mixed between different users or sessions.

packages/types/src/providers/roo.ts

@@ -10,7 +10,7 @@ export const rooModels = {
 		maxTokens: 16_384,
 		contextWindow: 262_144,
 		supportsImages: false,
-		supportsPromptCache: true,
+		supportsPromptCache: false, // Disabled to prevent context mixing between sessions
 		inputPrice: 0,
 		outputPrice: 0,
 		description:

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

@@ -257,6 +257,7 @@ describe("RooHandler", () => {
 						expect.objectContaining({ role: "user", content: "Second message" }),
 					]),
 				}),
+				expect.any(Object), // Headers object
 			)
 		})
 	})
@@ -331,7 +332,7 @@ describe("RooHandler", () => {
 			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.supportsPromptCache).toBe(false) // Should be false now to prevent context mixing
 			expect(modelInfo.info.inputPrice).toBe(0)
 			expect(modelInfo.info.outputPrice).toBe(0)
 		})
@@ -361,6 +362,7 @@ describe("RooHandler", () => {
 				expect.not.objectContaining({
 					temperature: expect.anything(),
 				}),
+				expect.any(Object), // Headers object
 			)
 		})
 
@@ -378,6 +380,7 @@ describe("RooHandler", () => {
 				expect.objectContaining({
 					temperature: 0.9,
 				}),
+				expect.any(Object), // Headers object
 			)
 		})
 
@@ -433,4 +436,143 @@ describe("RooHandler", () => {
 			}).toThrow("Authentication required for Roo Code Cloud")
 		})
 	})
+
+	describe("session isolation", () => {
+		beforeEach(() => {
+			mockHasInstanceFn.mockReturnValue(true)
+			mockGetSessionTokenFn.mockReturnValue("test-session-token")
+			mockCreate.mockClear()
+		})
+
+		it("should include session isolation headers in requests", async () => {
+			handler = new RooHandler(mockOptions)
+			const stream = handler.createMessage(systemPrompt, messages)
+
+			// Consume the stream
+			for await (const _chunk of stream) {
+				// Just consume
+			}
+
+			// Verify that create was called with session isolation headers
+			expect(mockCreate).toHaveBeenCalledWith(
+				expect.any(Object),
+				expect.objectContaining({
+					headers: expect.objectContaining({
+						"X-Session-Id": expect.any(String),
+						"X-Request-Id": expect.any(String),
+						"X-No-Cache": "true",
+						"Cache-Control": "no-store, no-cache, must-revalidate",
+						Pragma: "no-cache",
+					}),
+				}),
+			)
+		})
+
+		it("should generate unique session IDs for different handler instances", async () => {
+			const handler1 = new RooHandler(mockOptions)
+			const handler2 = new RooHandler(mockOptions)
+
+			// Create messages with both handlers
+			const stream1 = handler1.createMessage(systemPrompt, messages)
+			for await (const _chunk of stream1) {
+				// Consume
+			}
+
+			const stream2 = handler2.createMessage(systemPrompt, messages)
+			for await (const _chunk of stream2) {
+				// Consume
+			}
+
+			// Get the session IDs from the calls
+			const call1Headers = mockCreate.mock.calls[0][1].headers
+			const call2Headers = mockCreate.mock.calls[1][1].headers
+
+			// Session IDs should be different for different handler instances
+			expect(call1Headers["X-Session-Id"]).toBeDefined()
+			expect(call2Headers["X-Session-Id"]).toBeDefined()
+			expect(call1Headers["X-Session-Id"]).not.toBe(call2Headers["X-Session-Id"])
+		})
+
+		it("should generate unique request IDs for each request", async () => {
+			handler = new RooHandler(mockOptions)
+
+			// Make two requests with the same handler
+			const stream1 = handler.createMessage(systemPrompt, messages)
+			for await (const _chunk of stream1) {
+				// Consume
+			}
+
+			const stream2 = handler.createMessage(systemPrompt, messages)
+			for await (const _chunk of stream2) {
+				// Consume
+			}
+
+			// Get the request IDs from the calls
+			const call1Headers = mockCreate.mock.calls[0][1].headers
+			const call2Headers = mockCreate.mock.calls[1][1].headers
+
+			// Request IDs should be different for each request
+			expect(call1Headers["X-Request-Id"]).toBeDefined()
+			expect(call2Headers["X-Request-Id"]).toBeDefined()
+			expect(call1Headers["X-Request-Id"]).not.toBe(call2Headers["X-Request-Id"])
+
+			// But session IDs should be the same for the same handler
+			expect(call1Headers["X-Session-Id"]).toBe(call2Headers["X-Session-Id"])
+		})
+
+		it("should include metadata in request params", async () => {
+			handler = new RooHandler(mockOptions)
+			const stream = handler.createMessage(systemPrompt, messages)
+
+			for await (const _chunk of stream) {
+				// Consume
+			}
+
+			// Verify metadata is included in the request
+			expect(mockCreate).toHaveBeenCalledWith(
+				expect.objectContaining({
+					metadata: expect.objectContaining({
+						session_id: expect.any(String),
+						request_id: expect.any(String),
+						timestamp: expect.any(String),
+					}),
+				}),
+				expect.any(Object),
+			)
+		})
+
+		it("should have prompt caching disabled for roo/sonic model", () => {
+			handler = new RooHandler(mockOptions)
+			const modelInfo = handler.getModel()
+
+			// Verify that prompt caching is disabled
+			expect(modelInfo.info.supportsPromptCache).toBe(false)
+		})
+
+		it("should maintain session ID consistency across multiple requests", async () => {
+			handler = new RooHandler(mockOptions)
+
+			// Make multiple requests
+			const requests = []
+			for (let i = 0; i < 3; i++) {
+				const stream = handler.createMessage(systemPrompt, messages)
+				for await (const _chunk of stream) {
+					// Consume
+				}
+				requests.push(i)
+			}
+
+			// All requests should have the same session ID
+			const sessionIds = mockCreate.mock.calls.map((call) => call[1].headers["X-Session-Id"])
+			const firstSessionId = sessionIds[0]
+
+			expect(sessionIds.every((id) => id === firstSessionId)).toBe(true)
+
+			// But all request IDs should be unique
+			const requestIds = mockCreate.mock.calls.map((call) => call[1].headers["X-Request-Id"])
+			const uniqueRequestIds = new Set(requestIds)
+
+			expect(uniqueRequestIds.size).toBe(requestIds.length)
+		})
+	})
 })

src/api/providers/roo.ts

@@ -1,15 +1,20 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import { rooDefaultModelId, rooModels, type RooModelId } from "@roo-code/types"
 import { CloudService } from "@roo-code/cloud"
+import { randomUUID } from "crypto"
+import OpenAI from "openai"
 
 import type { ApiHandlerOptions } from "../../shared/api"
 import { ApiStream } from "../transform/stream"
 import { t } from "../../i18n"
+import { convertToOpenAiMessages } from "../transform/openai-format"
 
 import type { ApiHandlerCreateMessageMetadata } from "../index"
 import { BaseOpenAiCompatibleProvider } from "./base-openai-compatible-provider"
 
 export class RooHandler extends BaseOpenAiCompatibleProvider<RooModelId> {
+	private sessionId: string
+
 	constructor(options: ApiHandlerOptions) {
 		// Check if CloudService is available and get the session token.
 		if (!CloudService.hasInstance()) {
@@ -22,6 +27,9 @@ export class RooHandler extends BaseOpenAiCompatibleProvider<RooModelId> {
 			throw new Error(t("common:errors.roo.authenticationRequired"))
 		}
 
+		// Generate a unique session ID for this handler instance to ensure request isolation
+		const sessionId = randomUUID()
+
 		super({
 			...options,
 			providerName: "Roo Code Cloud",
@@ -31,6 +39,53 @@ export class RooHandler extends BaseOpenAiCompatibleProvider<RooModelId> {
 			providerModels: rooModels,
 			defaultTemperature: 0.7,
 		})
+
+		this.sessionId = sessionId
+	}
+
+	protected override createStream(
+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		metadata?: ApiHandlerCreateMessageMetadata,
+	) {
+		const {
+			id: model,
+			info: { maxTokens: max_tokens },
+		} = this.getModel()
+
+		// Generate unique request ID for this specific request
+		const requestId = randomUUID()
+
+		// Create the request with session isolation metadata
+		const params: OpenAI.Chat.Completions.ChatCompletionCreateParamsStreaming = {
+			model,
+			max_tokens,
+			messages: [{ role: "system", content: systemPrompt }, ...convertToOpenAiMessages(messages)],
+			stream: true,
+			stream_options: { include_usage: true },
+			// Add session isolation metadata to prevent context mixing
+			metadata: {
+				session_id: this.sessionId,
+				request_id: requestId,
+				timestamp: new Date().toISOString(),
+			} as any,
+		}
+
+		// Only include temperature if explicitly set
+		if (this.options.modelTemperature !== undefined) {
+			params.temperature = this.options.modelTemperature
+		}
+
+		// Create the stream with additional headers for session isolation
+		return this.client.chat.completions.create(params, {
+			headers: {
+				"X-Session-Id": this.sessionId,
+				"X-Request-Id": requestId,
+				"X-No-Cache": "true", // Prevent any server-side caching
+				"Cache-Control": "no-store, no-cache, must-revalidate",
+				Pragma: "no-cache",
+			},
+		})
 	}
 
 	override async *createMessage(
@@ -78,13 +133,14 @@ export class RooHandler extends BaseOpenAiCompatibleProvider<RooModelId> {
 		}
 
 		// Return the requested model ID even if not found, with fallback info.
+		// Note: supportsPromptCache is now false to prevent context mixing
 		return {
 			id: modelId as RooModelId,
 			info: {
 				maxTokens: 16_384,
 				contextWindow: 262_144,
 				supportsImages: false,
-				supportsPromptCache: true,
+				supportsPromptCache: false, // Disabled to prevent context mixing
 				inputPrice: 0,
 				outputPrice: 0,
 			},