fix: improve generate_image tool partial block handling and clarify image editing capabilities

- Fix partial block detection to properly wait for complete tool calls before processing
- Update tool documentation to explicitly mention image editing/transformation capabilities
- Add comprehensive tests for generateImageTool including partial block handling
- Clarify that the tool can edit existing images, not just use them as starting points

Author: daniel-lxs

src/core/prompts/tools/generate-image.ts

@@ -2,11 +2,11 @@ import { ToolArgs } from "./types"
 
 export function getGenerateImageDescription(args: ToolArgs): string {
 	return `## generate_image
-Description: Request to generate an image using AI models through OpenRouter API. This tool creates images from text prompts and saves them to the specified path. Optionally, you can provide an input image to use as a reference or starting point for the generation.
+Description: Request to generate or edit an image using AI models through OpenRouter API. This tool can create new images from text prompts or modify existing images based on your instructions. When an input image is provided, the AI will apply the requested edits, transformations, or enhancements to that image.
 Parameters:
-- prompt: (required) The text prompt describing the image to generate
-- path: (required) The file path where the generated image should be saved (relative to the current workspace directory ${args.cwd}). The tool will automatically add the appropriate image extension if not provided.
-- image: (optional) The file path to an input image to use as a reference or starting point (relative to the current workspace directory ${args.cwd}). Supported formats: PNG, JPG, JPEG, GIF, WEBP.
+- prompt: (required) The text prompt describing what to generate or how to edit the image
+- path: (required) The file path where the generated/edited image should be saved (relative to the current workspace directory ${args.cwd}). The tool will automatically add the appropriate image extension if not provided.
+- image: (optional) The file path to an input image to edit or transform (relative to the current workspace directory ${args.cwd}). Supported formats: PNG, JPG, JPEG, GIF, WEBP.
 Usage:
 <generate_image>
 <prompt>Your image description here</prompt>
@@ -20,10 +20,17 @@ Example: Requesting to generate a sunset image
 <path>images/sunset.png</path>
 </generate_image>
 
-Example: Generating an image with an input reference
+Example: Editing an existing image
 <generate_image>
 <prompt>Transform this image into a watercolor painting style</prompt>
 <path>images/watercolor-output.png</path>
 <image>images/original-photo.jpg</image>
+</generate_image>
+
+Example: Upscaling and enhancing an image
+<generate_image>
+<prompt>Upscale this image to higher resolution, enhance details, improve clarity and sharpness while maintaining the original content and composition</prompt>
+<path>images/enhanced-photo.png</path>
+<image>images/low-res-photo.jpg</image>
 </generate_image>`
 }

src/core/tools/__tests__/generateImageTool.test.ts

@@ -0,0 +1,314 @@
+import { describe, it, expect, vi, beforeEach } from "vitest"
+import { generateImageTool } from "../generateImageTool"
+import { ToolUse } from "../../../shared/tools"
+import { Task } from "../../task/Task"
+import * as fs from "fs/promises"
+import * as pathUtils from "../../../utils/pathUtils"
+import * as fileUtils from "../../../utils/fs"
+import { formatResponse } from "../../prompts/responses"
+import { EXPERIMENT_IDS } from "../../../shared/experiments"
+
+// Mock dependencies
+vi.mock("fs/promises")
+vi.mock("../../../utils/pathUtils")
+vi.mock("../../../utils/fs")
+vi.mock("../../../utils/safeWriteJson")
+vi.mock("../../../api/providers/openrouter")
+
+describe("generateImageTool", () => {
+	let mockCline: any
+	let mockAskApproval: any
+	let mockHandleError: any
+	let mockPushToolResult: any
+	let mockRemoveClosingTag: any
+
+	beforeEach(() => {
+		vi.clearAllMocks()
+
+		// Setup mock Cline instance
+		mockCline = {
+			cwd: "/test/workspace",
+			consecutiveMistakeCount: 0,
+			recordToolError: vi.fn(),
+			recordToolUsage: vi.fn(),
+			sayAndCreateMissingParamError: vi.fn().mockResolvedValue("Missing parameter error"),
+			say: vi.fn(),
+			rooIgnoreController: {
+				validateAccess: vi.fn().mockReturnValue(true),
+			},
+			rooProtectedController: {
+				isWriteProtected: vi.fn().mockReturnValue(false),
+			},
+			providerRef: {
+				deref: vi.fn().mockReturnValue({
+					getState: vi.fn().mockResolvedValue({
+						experiments: {
+							[EXPERIMENT_IDS.IMAGE_GENERATION]: true,
+						},
+						apiConfiguration: {
+							openRouterImageGenerationSettings: {
+								openRouterApiKey: "test-api-key",
+								selectedModel: "google/gemini-2.5-flash-image-preview",
+							},
+						},
+					}),
+				}),
+			},
+			fileContextTracker: {
+				trackFileContext: vi.fn(),
+			},
+			didEditFile: false,
+		}
+
+		mockAskApproval = vi.fn().mockResolvedValue(true)
+		mockHandleError = vi.fn()
+		mockPushToolResult = vi.fn()
+		mockRemoveClosingTag = vi.fn((tag, content) => content || "")
+
+		// Mock file system operations
+		vi.mocked(fileUtils.fileExistsAtPath).mockResolvedValue(true)
+		vi.mocked(fs.readFile).mockResolvedValue(Buffer.from("fake-image-data"))
+		vi.mocked(fs.mkdir).mockResolvedValue(undefined)
+		vi.mocked(fs.writeFile).mockResolvedValue(undefined)
+		vi.mocked(pathUtils.isPathOutsideWorkspace).mockReturnValue(false)
+	})
+
+	describe("partial block handling", () => {
+		it("should return early when block is partial", async () => {
+			const partialBlock: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Generate a test image",
+					path: "test-image.png",
+				},
+				partial: true,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				partialBlock,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			// Should not process anything when partial
+			expect(mockAskApproval).not.toHaveBeenCalled()
+			expect(mockPushToolResult).not.toHaveBeenCalled()
+			expect(mockCline.say).not.toHaveBeenCalled()
+		})
+
+		it("should return early when block is partial even with image parameter", async () => {
+			const partialBlock: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Upscale this image",
+					path: "upscaled-image.png",
+					image: "source-image.png",
+				},
+				partial: true,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				partialBlock,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			// Should not process anything when partial
+			expect(mockAskApproval).not.toHaveBeenCalled()
+			expect(mockPushToolResult).not.toHaveBeenCalled()
+			expect(mockCline.say).not.toHaveBeenCalled()
+			expect(fs.readFile).not.toHaveBeenCalled()
+		})
+
+		it("should process when block is not partial", async () => {
+			const completeBlock: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Generate a test image",
+					path: "test-image.png",
+				},
+				partial: false,
+			}
+
+			// Mock OpenRouter handler
+			const mockOpenRouterHandler = {
+				generateImage: vi.fn().mockResolvedValue({
+					success: true,
+					imageData: "data:image/png;base64,fakebase64data",
+				}),
+			}
+
+			vi.doMock("../../../api/providers/openrouter", () => ({
+				OpenRouterHandler: vi.fn().mockImplementation(() => mockOpenRouterHandler),
+			}))
+
+			await generateImageTool(
+				mockCline as Task,
+				completeBlock,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			// Should process the complete block
+			expect(mockAskApproval).toHaveBeenCalled()
+			expect(mockPushToolResult).toHaveBeenCalled()
+		})
+	})
+
+	describe("missing parameters", () => {
+		it("should handle missing prompt parameter", async () => {
+			const block: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					path: "test-image.png",
+				},
+				partial: false,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				block,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			expect(mockCline.consecutiveMistakeCount).toBe(1)
+			expect(mockCline.recordToolError).toHaveBeenCalledWith("generate_image")
+			expect(mockCline.sayAndCreateMissingParamError).toHaveBeenCalledWith("generate_image", "prompt")
+			expect(mockPushToolResult).toHaveBeenCalledWith("Missing parameter error")
+		})
+
+		it("should handle missing path parameter", async () => {
+			const block: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Generate a test image",
+				},
+				partial: false,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				block,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			expect(mockCline.consecutiveMistakeCount).toBe(1)
+			expect(mockCline.recordToolError).toHaveBeenCalledWith("generate_image")
+			expect(mockCline.sayAndCreateMissingParamError).toHaveBeenCalledWith("generate_image", "path")
+			expect(mockPushToolResult).toHaveBeenCalledWith("Missing parameter error")
+		})
+	})
+
+	describe("experiment validation", () => {
+		it("should error when image generation experiment is disabled", async () => {
+			// Disable the experiment
+			mockCline.providerRef.deref().getState.mockResolvedValue({
+				experiments: {
+					[EXPERIMENT_IDS.IMAGE_GENERATION]: false,
+				},
+			})
+
+			const block: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Generate a test image",
+					path: "test-image.png",
+				},
+				partial: false,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				block,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			expect(mockPushToolResult).toHaveBeenCalledWith(
+				formatResponse.toolError(
+					"Image generation is an experimental feature that must be enabled in settings. Please enable 'Image Generation' in the Experimental Settings section.",
+				),
+			)
+		})
+	})
+
+	describe("input image validation", () => {
+		it("should handle non-existent input image", async () => {
+			vi.mocked(fileUtils.fileExistsAtPath).mockResolvedValue(false)
+
+			const block: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Upscale this image",
+					path: "upscaled.png",
+					image: "non-existent.png",
+				},
+				partial: false,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				block,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			expect(mockCline.say).toHaveBeenCalledWith("error", expect.stringContaining("Input image not found"))
+			expect(mockPushToolResult).toHaveBeenCalledWith(
+				formatResponse.toolError(expect.stringContaining("Input image not found")),
+			)
+		})
+
+		it("should handle unsupported image format", async () => {
+			const block: ToolUse = {
+				type: "tool_use",
+				name: "generate_image",
+				params: {
+					prompt: "Upscale this image",
+					path: "upscaled.png",
+					image: "test.bmp", // Unsupported format
+				},
+				partial: false,
+			}
+
+			await generateImageTool(
+				mockCline as Task,
+				block,
+				mockAskApproval,
+				mockHandleError,
+				mockPushToolResult,
+				mockRemoveClosingTag,
+			)
+
+			expect(mockCline.say).toHaveBeenCalledWith("error", expect.stringContaining("Unsupported image format"))
+			expect(mockPushToolResult).toHaveBeenCalledWith(
+				formatResponse.toolError(expect.stringContaining("Unsupported image format")),
+			)
+		})
+	})
+})

src/core/tools/generateImageTool.ts

@@ -40,8 +40,7 @@ export async function generateImageTool(
 		return
 	}
 
-	if (block.partial && (!prompt || !relPath)) {
-		// Wait for complete parameters
+	if (block.partial) {
 		return
 	}