Merge remote-tracking branch 'upstream/main' into feat/vertex-prompt-caching

Author: aitoroses

.changeset/sour-parents-hug.md

@@ -0,0 +1,5 @@
+---
+"roo-cline": patch
+---
+
+Stop removing commas from terminal output

CHANGELOG.md

@@ -1,5 +1,25 @@
 # Roo Code Changelog
 
+## [3.7.6]
+
+- Handle really long text better in the in the ChatRow similar to TaskHeader (thanks @joemanley201!)
+- Support multiple files in drag-and-drop
+- Truncate search_file output to avoid crashing the extension
+- Better OpenRouter error handling (no more "Provider Error")
+- Add slider to control max output tokens for thinking models
+
+## [3.7.5]
+
+- Fix context window truncation math (see [#1173](https://github.com/RooVetGit/Roo-Code/issues/1173))
+- Fix various issues with the model picker (thanks @System233!)
+- Fix model input / output cost parsing (thanks @System233!)
+- Add drag-and-drop for files
+- Enable the "Thinking Budget" slider for Claude 3.7 Sonnet on OpenRouter
+
+## [3.7.4]
+
+- Fix a bug that prevented the "Thinking" setting from properly updating when switching profiles.
+
 ## [3.7.3]
 
 - Support for ["Thinking"](https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking) Sonnet 3.7 when using the Anthropic provider.

package-lock.json

@@ -1,9 +1,9 @@
 {
 	"name": "roo-cline",
 	"displayName": "Roo Code (prev. Roo Cline)",
-	"description": "An AI-powered autonomous coding agent that lives in your editor.",
+	"description": "A whole dev team of AI agents in your editor.",
 	"publisher": "RooVeterinaryInc",
-	"version": "3.7.3",
+	"version": "3.7.6",
 	"icon": "assets/icons/rocket.png",
 	"galleryBanner": {
 		"color": "#617A91",

package.json

@@ -14,8 +14,6 @@ import { ApiStream } from "../transform/stream"
 
 const ANTHROPIC_DEFAULT_TEMPERATURE = 0
 
-const THINKING_MODELS = ["claude-3-7-sonnet-20250219"]
-
 export class AnthropicHandler implements ApiHandler, SingleCompletionHandler {
 	private options: ApiHandlerOptions
 	private client: Anthropic
@@ -32,16 +30,27 @@ export class AnthropicHandler implements ApiHandler, SingleCompletionHandler {
 	async *createMessage(systemPrompt: string, messages: Anthropic.Messages.MessageParam[]): ApiStream {
 		let stream: AnthropicStream<Anthropic.Messages.RawMessageStreamEvent>
 		const cacheControl: CacheControlEphemeral = { type: "ephemeral" }
-		const modelId = this.getModel().id
-		const maxTokens = this.getModel().info.maxTokens || 8192
+		let { id: modelId, info: modelInfo } = this.getModel()
+		const maxTokens = this.options.modelMaxTokens || modelInfo.maxTokens || 8192
 		let temperature = this.options.modelTemperature ?? ANTHROPIC_DEFAULT_TEMPERATURE
 		let thinking: BetaThinkingConfigParam | undefined = undefined
 
-		if (THINKING_MODELS.includes(modelId)) {
-			thinking = this.options.anthropicThinking
-				? { type: "enabled", budget_tokens: this.options.anthropicThinking }
-				: { type: "disabled" }
-
+		// Anthropic "Thinking" models require a temperature of 1.0.
+		if (modelId === "claude-3-7-sonnet-20250219:thinking") {
+			// The `:thinking` variant is a virtual identifier for the
+			// `claude-3-7-sonnet-20250219` model with a thinking budget.
+			// We can handle this more elegantly in the future.
+			modelId = "claude-3-7-sonnet-20250219"
+
+			// Clamp the thinking budget to be at most 80% of max tokens and at
+			// least 1024 tokens.
+			const maxBudgetTokens = Math.floor(maxTokens * 0.8)
+			const budgetTokens = Math.max(
+				Math.min(this.options.anthropicThinking ?? maxBudgetTokens, maxBudgetTokens),
+				1024,
+			)
+
+			thinking = { type: "enabled", budget_tokens: budgetTokens }
 			temperature = 1.0
 		}
 
@@ -114,8 +123,8 @@ export class AnthropicHandler implements ApiHandler, SingleCompletionHandler {
 			default: {
 				stream = (await this.client.messages.create({
 					model: modelId,
-					max_tokens: this.getModel().info.maxTokens || 8192,
-					temperature: this.options.modelTemperature ?? ANTHROPIC_DEFAULT_TEMPERATURE,
+					max_tokens: maxTokens,
+					temperature,
 					system: [{ text: systemPrompt, type: "text" }],
 					messages,
 					// tools,

src/api/providers/anthropic.ts

@@ -1,10 +1,12 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import axios from "axios"
 import OpenAI from "openai"
-import { ApiHandler, SingleCompletionHandler } from "../"
+
 import { ApiHandlerOptions, ModelInfo, glamaDefaultModelId, glamaDefaultModelInfo } from "../../shared/api"
+import { parseApiPrice } from "../../utils/cost"
 import { convertToOpenAiMessages } from "../transform/openai-format"
 import { ApiStream } from "../transform/stream"
+import { ApiHandler, SingleCompletionHandler } from "../"
 
 const GLAMA_DEFAULT_TEMPERATURE = 0
 
@@ -69,7 +71,7 @@ export class GlamaHandler implements ApiHandler, SingleCompletionHandler {
 		let maxTokens: number | undefined
 
 		if (this.getModel().id.startsWith("anthropic/")) {
-			maxTokens = 8_192
+			maxTokens = this.getModel().info.maxTokens
 		}
 
 		const requestOptions: OpenAI.Chat.ChatCompletionCreateParams = {
@@ -177,7 +179,7 @@ export class GlamaHandler implements ApiHandler, SingleCompletionHandler {
 			}
 
 			if (this.getModel().id.startsWith("anthropic/")) {
-				requestOptions.max_tokens = 8192
+				requestOptions.max_tokens = this.getModel().info.maxTokens
 			}
 
 			const response = await this.client.chat.completions.create(requestOptions)
@@ -190,3 +192,44 @@ export class GlamaHandler implements ApiHandler, SingleCompletionHandler {
 		}
 	}
 }
+
+export async function getGlamaModels() {
+	const models: Record<string, ModelInfo> = {}
+
+	try {
+		const response = await axios.get("https://glama.ai/api/gateway/v1/models")
+		const rawModels = response.data
+
+		for (const rawModel of rawModels) {
+			const modelInfo: ModelInfo = {
+				maxTokens: rawModel.maxTokensOutput,
+				contextWindow: rawModel.maxTokensInput,
+				supportsImages: rawModel.capabilities?.includes("input:image"),
+				supportsComputerUse: rawModel.capabilities?.includes("computer_use"),
+				supportsPromptCache: rawModel.capabilities?.includes("caching"),
+				inputPrice: parseApiPrice(rawModel.pricePerToken?.input),
+				outputPrice: parseApiPrice(rawModel.pricePerToken?.output),
+				description: undefined,
+				cacheWritesPrice: parseApiPrice(rawModel.pricePerToken?.cacheWrite),
+				cacheReadsPrice: parseApiPrice(rawModel.pricePerToken?.cacheRead),
+			}
+
+			switch (rawModel.id) {
+				case rawModel.id.startsWith("anthropic/claude-3-7-sonnet"):
+					modelInfo.maxTokens = 16384
+					break
+				case rawModel.id.startsWith("anthropic/"):
+					modelInfo.maxTokens = 8192
+					break
+				default:
+					break
+			}
+
+			models[rawModel.id] = modelInfo
+		}
+	} catch (error) {
+		console.error(`Error fetching Glama models: ${JSON.stringify(error, Object.getOwnPropertyNames(error), 2)}`)
+	}
+
+	return models
+}

src/api/providers/glama.ts

@@ -1,5 +1,7 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import OpenAI from "openai"
+import axios from "axios"
+
 import { ApiHandler, SingleCompletionHandler } from "../"
 import { ApiHandlerOptions, ModelInfo, openAiModelInfoSaneDefaults } from "../../shared/api"
 import { convertToOpenAiMessages } from "../transform/openai-format"
@@ -72,3 +74,17 @@ export class LmStudioHandler implements ApiHandler, SingleCompletionHandler {
 		}
 	}
 }
+
+export async function getLmStudioModels(baseUrl = "http://localhost:1234") {
+	try {
+		if (!URL.canParse(baseUrl)) {
+			return []
+		}
+
+		const response = await axios.get(`${baseUrl}/v1/models`)
+		const modelsArray = response.data?.data?.map((model: any) => model.id) || []
+		return [...new Set<string>(modelsArray)]
+	} catch (error) {
+		return []
+	}
+}

src/api/providers/lmstudio.ts

@@ -1,5 +1,7 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import OpenAI from "openai"
+import axios from "axios"
+
 import { ApiHandler, SingleCompletionHandler } from "../"
 import { ApiHandlerOptions, ModelInfo, openAiModelInfoSaneDefaults } from "../../shared/api"
 import { convertToOpenAiMessages } from "../transform/openai-format"
@@ -88,3 +90,17 @@ export class OllamaHandler implements ApiHandler, SingleCompletionHandler {
 		}
 	}
 }
+
+export async function getOllamaModels(baseUrl = "http://localhost:11434") {
+	try {
+		if (!URL.canParse(baseUrl)) {
+			return []
+		}
+
+		const response = await axios.get(`${baseUrl}/api/tags`)
+		const modelsArray = response.data?.models?.map((model: any) => model.name) || []
+		return [...new Set<string>(modelsArray)]
+	} catch (error) {
+		return []
+	}
+}

src/api/providers/ollama.ts

@@ -1,5 +1,6 @@
 import { Anthropic } from "@anthropic-ai/sdk"
 import OpenAI, { AzureOpenAI } from "openai"
+import axios from "axios"
 
 import {
 	ApiHandlerOptions,
@@ -166,3 +167,27 @@ export class OpenAiHandler implements ApiHandler, SingleCompletionHandler {
 		}
 	}
 }
+
+export async function getOpenAiModels(baseUrl?: string, apiKey?: string) {
+	try {
+		if (!baseUrl) {
+			return []
+		}
+
+		if (!URL.canParse(baseUrl)) {
+			return []
+		}
+
+		const config: Record<string, any> = {}
+
+		if (apiKey) {
+			config["headers"] = { Authorization: `Bearer ${apiKey}` }
+		}
+
+		const response = await axios.get(`${baseUrl}/models`, config)
+		const modelsArray = response.data?.data?.map((model: any) => model.id) || []
+		return [...new Set<string>(modelsArray)]
+	} catch (error) {
+		return []
+	}
+}