zgsm-ai
diff --git a/‎packages/types/src/providers/mistral.ts‎
Lines changed: 3 additions & 3 deletions b/‎packages/types/src/providers/mistral.ts‎
Lines changed: 3 additions & 3 deletions
diff --git a/‎releases/3.29.0-release.png‎
1.03 MB b/‎releases/3.29.0-release.png‎
1.03 MB
diff --git a/‎src/api/providers/fetchers/roo.ts‎
Lines changed: 5 additions & 4 deletions b/‎src/api/providers/fetchers/roo.ts‎
Lines changed: 5 additions & 4 deletions
diff --git a/‎src/api/providers/roo.ts‎
Lines changed: 12 additions & 2 deletions b/‎src/api/providers/roo.ts‎
Lines changed: 12 additions & 2 deletions
diff --git a/‎src/core/checkpoints/__tests__/checkpoint.test.ts‎
Lines changed: 7 additions & 7 deletions b/‎src/core/checkpoints/__tests__/checkpoint.test.ts‎
Lines changed: 7 additions & 7 deletions
diff --git a/‎src/core/checkpoints/index.ts‎
Lines changed: 48 additions & 15 deletions b/‎src/core/checkpoints/index.ts‎
Lines changed: 48 additions & 15 deletions
diff --git a/‎src/core/config/ProviderSettingsManager.ts‎
Lines changed: 26 additions & 0 deletions b/‎src/core/config/ProviderSettingsManager.ts‎
Lines changed: 26 additions & 0 deletions
@@ -7,9 +7,9 @@ export const mistralDefaultModelId: MistralModelId = "codestral-latest"
 
 export const mistralModels = {
 	"magistral-medium-latest": {
-		maxTokens: 41_000,
-		contextWindow: 41_000,
-		supportsImages: false,
+		maxTokens: 8192,
+		contextWindow: 128_000,
+		supportsImages: true,
 		supportsPromptCache: false,
 		inputPrice: 2.0,
 		outputPrice: 5.0,
 
@@ -1,6 +1,7 @@
 import { RooModelsResponseSchema } from "@roo-code/types"
 
 import type { ModelRecord } from "../../../shared/api"
+import { parseApiPrice } from "../../../shared/cost"
 
 import { DEFAULT_HEADERS } from "../constants"
 
@@ -71,10 +72,10 @@ export async function getRooModels(baseUrl: string, apiKey?: string): Promise<Mo
 				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
+				const inputPrice = parseApiPrice(pricing.input)
+				const outputPrice = parseApiPrice(pricing.output)
+				const cacheReadPrice = pricing.input_cache_read ? parseApiPrice(pricing.input_cache_read) : undefined
+				const cacheWritePrice = pricing.input_cache_write ? parseApiPrice(pricing.input_cache_write) : undefined
 
 				models[modelId] = {
 					maxTokens,
 
@@ -12,6 +12,12 @@ import { DEFAULT_HEADERS } from "./constants"
 import { BaseOpenAiCompatibleProvider } from "./base-openai-compatible-provider"
 import { getModels, flushModels, getModelsFromCache } from "../providers/fetchers/modelCache"
 
+// Extend OpenAI's CompletionUsage to include Roo specific fields
+interface RooUsage extends OpenAI.CompletionUsage {
+	cache_creation_input_tokens?: number
+	cost?: number
+}
+
 export class RooHandler extends BaseOpenAiCompatibleProvider<string> {
 	private authStateListener?: (state: { state: AuthState }) => void
 	private fetcherBaseURL: string
@@ -124,10 +130,14 @@ export class RooHandler extends BaseOpenAiCompatibleProvider<string> {
 			}
 
 			if (chunk.usage) {
+				const usage = chunk.usage as RooUsage
 				yield {
 					type: "usage",
-					inputTokens: chunk.usage.prompt_tokens || 0,
-					outputTokens: chunk.usage.completion_tokens || 0,
+					inputTokens: usage.prompt_tokens || 0,
+					outputTokens: usage.completion_tokens || 0,
+					cacheWriteTokens: usage.cache_creation_input_tokens,
+					cacheReadTokens: usage.prompt_tokens_details?.cached_tokens,
+					totalCost: usage.cost ?? 0,
 				}
 			}
 		}
 
@@ -329,7 +329,7 @@ describe("Checkpoint functionality", () => {
 			]
 		})
 
-		it("should show diff for full mode", async () => {
+		it("should show diff for to-current mode", async () => {
 			const mockChanges = [
 				{
 					paths: { absolute: "/test/file.ts", relative: "file.ts" },
@@ -341,7 +341,7 @@ describe("Checkpoint functionality", () => {
 			await checkpointDiff(mockTask, {
 				ts: 4,
 				commitHash: "commit2",
-				mode: "full",
+				mode: "to-current",
 			})
 
 			expect(mockCheckpointService.getDiff).toHaveBeenCalledWith({
@@ -350,7 +350,7 @@ describe("Checkpoint functionality", () => {
 			})
 			expect(vscode.commands.executeCommand).toHaveBeenCalledWith(
 				"vscode.changes",
-				"Changes since task started",
+				"common:errors.checkpoint_diff_to_current",
 				expect.any(Array),
 			)
 		})
@@ -375,7 +375,7 @@ describe("Checkpoint functionality", () => {
 			})
 			expect(vscode.commands.executeCommand).toHaveBeenCalledWith(
 				"vscode.changes",
-				"Changes compare with next checkpoint",
+				"common:errors.checkpoint_diff_with_next",
 				expect.any(Array),
 			)
 		})
@@ -407,10 +407,10 @@ describe("Checkpoint functionality", () => {
 			await checkpointDiff(mockTask, {
 				ts: 4,
 				commitHash: "commit2",
-				mode: "full",
+				mode: "to-current",
 			})
 
-			expect(vscode.window.showInformationMessage).toHaveBeenCalledWith("No changes found.")
+			expect(vscode.window.showInformationMessage).toHaveBeenCalledWith("common:errors.checkpoint_no_changes")
 			expect(vscode.commands.executeCommand).not.toHaveBeenCalled()
 		})
 
@@ -420,7 +420,7 @@ describe("Checkpoint functionality", () => {
 			await checkpointDiff(mockTask, {
 				ts: 4,
 				commitHash: "commit2",
-				mode: "full",
+				mode: "to-current",
 			})
 
 			expect(mockTask.enableCheckpoints).toBe(false)
 
@@ -338,10 +338,16 @@ export async function checkpointRestore(
 }
 
 export type CheckpointDiffOptions = {
-	ts: number
+	ts?: number
 	previousCommitHash?: string
 	commitHash: string
-	mode: "full" | "checkpoint"
+	/**
+	 * from-init: Compare from the first checkpoint to the selected checkpoint.
+	 * checkpoint: Compare the selected checkpoint to the next checkpoint.
+	 * to-current: Compare the selected checkpoint to the current workspace.
+	 * full: Compare from the first checkpoint to the current workspace.
+	 */
+	mode: "from-init" | "checkpoint" | "to-current" | "full"
 }
 
 export async function checkpointDiff(task: Task, { ts, previousCommitHash, commitHash, mode }: CheckpointDiffOptions) {
@@ -353,30 +359,57 @@ export async function checkpointDiff(task: Task, { ts, previousCommitHash, commi
 
 	TelemetryService.instance.captureCheckpointDiffed(task.taskId)
 
-	let prevHash = commitHash
-	let nextHash: string | undefined = undefined
+	let fromHash: string | undefined
+	let toHash: string | undefined
+	let title: string
 
-	if (mode !== "full") {
-		const checkpoints = task.clineMessages.filter(({ say }) => say === "checkpoint_saved").map(({ text }) => text!)
-		const idx = checkpoints.indexOf(commitHash)
-		if (idx !== -1 && idx < checkpoints.length - 1) {
-			nextHash = checkpoints[idx + 1]
-		} else {
-			nextHash = undefined
-		}
+	const checkpoints = task.clineMessages.filter(({ say }) => say === "checkpoint_saved").map(({ text }) => text!)
+
+	if (["from-init", "full"].includes(mode) && checkpoints.length < 1) {
+		vscode.window.showInformationMessage(t("common:errors.checkpoint_no_first"))
+		return
+	}
+
+	const idx = checkpoints.indexOf(commitHash)
+	switch (mode) {
+		case "checkpoint":
+			fromHash = commitHash
+			toHash = idx !== -1 && idx < checkpoints.length - 1 ? checkpoints[idx + 1] : undefined
+			title = t("common:errors.checkpoint_diff_with_next")
+			break
+		case "from-init":
+			fromHash = checkpoints[0]
+			toHash = commitHash
+			title = t("common:errors.checkpoint_diff_since_first")
+			break
+		case "to-current":
+			fromHash = commitHash
+			toHash = undefined
+			title = t("common:errors.checkpoint_diff_to_current")
+			break
+		case "full":
+			fromHash = checkpoints[0]
+			toHash = undefined
+			title = t("common:errors.checkpoint_diff_since_first")
+			break
+	}
+
+	if (!fromHash) {
+		vscode.window.showInformationMessage(t("common:errors.checkpoint_no_previous"))
+		return
 	}
 
 	try {
-		const changes = await service.getDiff({ from: prevHash, to: nextHash })
+		const changes = await service.getDiff({ from: fromHash, to: toHash })
 
 		if (!changes?.length) {
-			vscode.window.showInformationMessage("No changes found.")
+			vscode.window.showInformationMessage(t("common:errors.checkpoint_no_changes"))
 			return
 		}
 
 		await vscode.commands.executeCommand(
 			"vscode.changes",
-			mode === "full" ? "Changes since task started" : "Changes compare with next checkpoint",
+			title,
 			changes.map((change) => [
 				vscode.Uri.file(change.paths.absolute),
 				vscode.Uri.parse(`${DIFF_VIEW_URI_SCHEME}:${change.paths.relative}`).with({
 
@@ -16,6 +16,7 @@ import {
 import { TelemetryService } from "@roo-code/telemetry"
 
 import { Mode, modes } from "../../shared/modes"
+import { buildApiHandler } from "../../api"
 
 // Type-safe model migrations mapping
 type ModelMigrations = {
@@ -549,6 +550,31 @@ export class ProviderSettingsManager {
 				for (const name in configs) {
 					// Avoid leaking properties from other providers.
 					configs[name] = discriminatedProviderSettingsWithIdSchema.parse(configs[name])
+
+					// If it has no apiProvider, skip filtering
+					if (!configs[name].apiProvider) {
+						continue
+					}
+
+					// Try to build an API handler to get model information
+					try {
+						const apiHandler = buildApiHandler(configs[name])
+						const modelInfo = apiHandler.getModel().info
+
+						// Check if the model supports reasoning budgets
+						const supportsReasoningBudget =
+							modelInfo.supportsReasoningBudget || modelInfo.requiredReasoningBudget
+
+						// If the model doesn't support reasoning budgets, remove the token fields
+						if (!supportsReasoningBudget) {
+							delete configs[name].modelMaxTokens
+							delete configs[name].modelMaxThinkingTokens
+						}
+					} catch (error) {
+						// If we can't build the API handler or get model info, skip filtering
+						// to avoid accidental data loss from incomplete configurations
+						console.warn(`Skipping token field filtering for config '${name}': ${error}`)
+					}
 				}
 				return profiles
 			})