feat: add query-context smart tool selection

- Add ToolSelectionAnalyzer class for intelligent tool filtering
- Analyze query complexity (simple/moderate/complex)
- Detect tool mentions and group needs
- Score tools based on relevance (0-1 scale)
- Select 6-12 most relevant tools per query
- Add configuration settings for the feature
- Include comprehensive unit tests

Fixes #9117

Author: roomote

packages/types/src/provider-settings.ts

@@ -182,6 +182,11 @@ const baseProviderSettingsSchema = z.object({
 
 	// Model verbosity.
 	verbosity: verbosityLevelsSchema.optional(),
+
+	// Smart tool selection.
+	smartToolSelectionEnabled: z.boolean().optional(),
+	smartToolSelectionMinTools: z.number().min(1).max(20).optional(),
+	smartToolSelectionMaxTools: z.number().min(1).max(20).optional(),
 })
 
 // Several of the providers share common model config properties.

src/core/prompts/system.ts

@@ -62,6 +62,7 @@ async function generatePrompt(
 	settings?: SystemPromptSettings,
 	todoList?: TodoItem[],
 	modelId?: string,
+	userQuery?: string,
 ): Promise<string> {
 	if (!context) {
 		throw new Error("Extension context is required for generating system prompt")
@@ -108,6 +109,7 @@ ${getToolDescriptionsForMode(
 	settings,
 	enableMcpServerCreation,
 	modelId,
+	userQuery,
 )}
 
 ${getToolUseGuidelinesSection(codeIndexManager)}
@@ -153,6 +155,7 @@ export const SYSTEM_PROMPT = async (
 	settings?: SystemPromptSettings,
 	todoList?: TodoItem[],
 	modelId?: string,
+	userQuery?: string,
 ): Promise<string> => {
 	if (!context) {
 		throw new Error("Extension context is required for generating system prompt")
@@ -225,5 +228,6 @@ ${customInstructions}`
 		settings,
 		todoList,
 		modelId,
+		userQuery,
 	)
 }

src/core/prompts/tools/ToolSelectionAnalyzer.test.ts

@@ -0,0 +1,308 @@
+import { describe, it, expect, beforeEach } from "vitest"
+import { ToolSelectionAnalyzer, SmartToolSelectionConfig } from "./ToolSelectionAnalyzer"
+import type { ModeConfig, ToolName, ToolGroup } from "@roo-code/types"
+
+describe("ToolSelectionAnalyzer", () => {
+	let analyzer: ToolSelectionAnalyzer
+
+	beforeEach(() => {
+		analyzer = new ToolSelectionAnalyzer()
+	})
+
+	const createModeConfig = (groups: ToolGroup[]): ModeConfig => ({
+		slug: "test",
+		name: "Test Mode",
+		roleDefinition: "Test role",
+		groups: groups as any, // Groups can be string or tuple format
+	})
+
+	const createAvailableTools = (tools: ToolName[]): Set<ToolName> => new Set(tools)
+
+	describe("query analysis", () => {
+		it("should detect simple queries", () => {
+			const simpleQueries = [
+				"what does this function do?",
+				"explain this code",
+				"what is the purpose of this file?",
+				"can you summarize this?",
+			]
+
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"search_files",
+				"list_code_definition_names",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			simpleQueries.forEach((query) => {
+				const selected = analyzer.selectTools(query, modeConfig, availableTools)
+				expect(selected.length).toBeGreaterThanOrEqual(6)
+				expect(selected.length).toBeLessThanOrEqual(8)
+				// Should include essential read tools
+				expect(selected).toContain("read_file")
+				expect(selected).toContain("list_files")
+			})
+		})
+
+		it("should detect complex queries", () => {
+			const complexQueries = [
+				"refactor the entire authentication system to use JWT tokens",
+				"implement a comprehensive logging system across all modules",
+				"redesign the database schema and migrate all existing data",
+				"build a complete REST API with authentication and rate limiting",
+			]
+
+			const modeConfig = createModeConfig(["read", "edit"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"apply_diff",
+				"list_files",
+				"search_files",
+				"execute_command",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			complexQueries.forEach((query) => {
+				const selected = analyzer.selectTools(query, modeConfig, availableTools)
+				expect(selected.length).toBeGreaterThanOrEqual(8)
+				expect(selected.length).toBeLessThanOrEqual(12)
+				// Should include both read and edit tools
+				expect(selected).toContain("read_file")
+				expect(selected).toContain("write_to_file")
+				expect(selected).toContain("apply_diff")
+			})
+		})
+	})
+
+	describe("tool-specific queries", () => {
+		it("should prioritize mentioned tools", () => {
+			const query = "run npm test and show me the results"
+			const modeConfig = createModeConfig(["read", "command"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"execute_command",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+			// execute_command should be included and highly ranked (top 5) due to explicit mention
+			expect(selected).toContain("execute_command")
+			const executeIndex = selected.indexOf("execute_command")
+			expect(executeIndex).toBeGreaterThanOrEqual(0)
+			expect(executeIndex).toBeLessThan(5) // Within top 5 is still highly ranked
+		})
+
+		it("should detect file operations", () => {
+			const query = "create a new config file with the settings I provided"
+			const modeConfig = createModeConfig(["read", "edit"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"apply_diff",
+				"list_files",
+				"search_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+			// write_to_file should be highly ranked
+			expect(selected.slice(0, 3)).toContain("write_to_file")
+		})
+
+		it("should detect search operations", () => {
+			const query = "find all places where the API key is used"
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"search_files",
+				"list_code_definition_names",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+			// search_files should be included and highly ranked (top 5) due to explicit search intent
+			expect(selected).toContain("search_files")
+			const searchIndex = selected.indexOf("search_files")
+			expect(searchIndex).toBeGreaterThanOrEqual(0)
+			expect(searchIndex).toBeLessThan(5) // Within top 5 is still highly ranked
+		})
+	})
+
+	describe("settings configuration", () => {
+		it("should respect minimum tools setting", () => {
+			const query = "simple task"
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const customAnalyzer = new ToolSelectionAnalyzer({
+				enabled: true,
+				minTools: 3,
+				maxTools: 12,
+			})
+
+			const selected = customAnalyzer.selectTools(query, modeConfig, availableTools)
+			expect(selected.length).toBeGreaterThanOrEqual(3)
+		})
+
+		it("should respect maximum tools setting", () => {
+			const query = "complex refactoring task involving multiple systems"
+			const modeConfig = createModeConfig(["read", "edit", "command"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"apply_diff",
+				"insert_content",
+				"list_files",
+				"search_files",
+				"list_code_definition_names",
+				"execute_command",
+				"ask_followup_question",
+				"attempt_completion",
+				"switch_mode",
+				"new_task",
+				"update_todo_list",
+			])
+
+			const customAnalyzer = new ToolSelectionAnalyzer({
+				enabled: true,
+				minTools: 6,
+				maxTools: 8,
+			})
+
+			const selected = customAnalyzer.selectTools(query, modeConfig, availableTools)
+			expect(selected.length).toBeLessThanOrEqual(8)
+		})
+
+		it("should return all tools when smart selection is disabled", () => {
+			const query = "some task"
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"search_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const disabledAnalyzer = new ToolSelectionAnalyzer({
+				enabled: false,
+			})
+
+			const selected = disabledAnalyzer.selectTools(query, modeConfig, availableTools)
+			expect(selected.length).toBe(availableTools.size)
+			expect(selected.sort()).toEqual(Array.from(availableTools).sort())
+		})
+	})
+
+	describe("essential tools", () => {
+		it("should always include ask_followup_question and attempt_completion", () => {
+			const queries = ["simple query", "complex refactoring", "create a file", "run tests"]
+
+			const modeConfig = createModeConfig(["read", "edit"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"list_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			queries.forEach((query) => {
+				const selected = analyzer.selectTools(query, modeConfig, availableTools)
+				expect(selected).toContain("ask_followup_question")
+				expect(selected).toContain("attempt_completion")
+			})
+		})
+	})
+
+	describe("empty or invalid inputs", () => {
+		it("should handle empty query gracefully", () => {
+			const query = ""
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"list_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+			expect(selected.length).toBeGreaterThan(0)
+			expect(selected).toContain("ask_followup_question")
+			expect(selected).toContain("attempt_completion")
+		})
+
+		it("should handle empty available tools", () => {
+			const query = "do something"
+			const modeConfig = createModeConfig(["read"])
+			const availableTools = createAvailableTools([])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+			expect(selected).toEqual([])
+		})
+	})
+
+	describe("tool scoring", () => {
+		it("should score tools based on query relevance", () => {
+			const query = "write a test file for the authentication module"
+			const modeConfig = createModeConfig(["read", "edit"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"apply_diff",
+				"list_files",
+				"search_files",
+				"execute_command",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+
+			// write_to_file should be highly ranked for "write a test file"
+			const writeIndex = selected.indexOf("write_to_file")
+			expect(writeIndex).toBeGreaterThanOrEqual(0)
+			expect(writeIndex).toBeLessThan(4) // Should be in top 4 tools
+
+			// read_file should also be included for context
+			expect(selected).toContain("read_file")
+		})
+
+		it("should handle multiple tool mentions in query", () => {
+			const query = "read the config file, update it, and run the build command"
+			const modeConfig = createModeConfig(["read", "edit", "command"])
+			const availableTools = createAvailableTools([
+				"read_file",
+				"write_to_file",
+				"apply_diff",
+				"execute_command",
+				"list_files",
+				"search_files",
+				"ask_followup_question",
+				"attempt_completion",
+			])
+
+			const selected = analyzer.selectTools(query, modeConfig, availableTools)
+
+			// Should include all mentioned operations
+			expect(selected).toContain("read_file")
+			expect(selected).toContain("apply_diff") // for "update"
+			expect(selected).toContain("execute_command") // for "run"
+		})
+	})
+})