@f5xc-salesdemos/xcsh 15.2.0 → 15.3.0

package/package.json

@@ -1,7 +1,7 @@
 {
 	"type": "module",
 	"name": "@f5xc-salesdemos/xcsh",
-	"version": "15.2.0",
+	"version": "15.3.0",
 	"description": "Coding agent CLI with read, bash, edit, write tools and session management",
 	"homepage": "https://github.com/f5xc-salesdemos/xcsh",
 	"author": "Can Boluk",
@@ -46,12 +46,12 @@
 	"dependencies": {
 		"@agentclientprotocol/sdk": "0.16.1",
 		"@mozilla/readability": "^0.6",
-		"@f5xc-salesdemos/xcsh-stats": "15.2.0",
-		"@f5xc-salesdemos/pi-agent-core": "15.2.0",
-		"@f5xc-salesdemos/pi-ai": "15.2.0",
-		"@f5xc-salesdemos/pi-natives": "15.2.0",
-		"@f5xc-salesdemos/pi-tui": "15.2.0",
-		"@f5xc-salesdemos/pi-utils": "15.2.0",
+		"@f5xc-salesdemos/xcsh-stats": "15.3.0",
+		"@f5xc-salesdemos/pi-agent-core": "15.3.0",
+		"@f5xc-salesdemos/pi-ai": "15.3.0",
+		"@f5xc-salesdemos/pi-natives": "15.3.0",
+		"@f5xc-salesdemos/pi-tui": "15.3.0",
+		"@f5xc-salesdemos/pi-utils": "15.3.0",
 		"@sinclair/typebox": "^0.34",
 		"@xterm/headless": "^6.0",
 		"ajv": "^8.18",

package/src/config/settings-schema.ts

@@ -1287,6 +1287,16 @@ export const SETTINGS_SCHEMA = {
 		},
 	},
 
+	"generate_image.enabled": {
+		type: "boolean",
+		default: true,
+		ui: {
+			tab: "tools",
+			label: "Generate Image",
+			description: "Enable the generate_image tool for AI-powered image and diagram generation",
+		},
+	},
+
 	"checkpoint.enabled": {
 		type: "boolean",
 		default: false,
@@ -1615,12 +1625,36 @@ export const SETTINGS_SCHEMA = {
 	},
 	"providers.image": {
 		type: "enum",
-		values: ["auto", "gemini", "openrouter"] as const,
+		values: ["auto", "gemini", "openrouter", "openai"] as const,
 		default: "auto",
 		ui: {
 			tab: "providers",
 			label: "Image Provider",
-			description: "Provider for image generation tool",
+			description: "Provider for image generation tool (auto detects from available API keys)",
+			submenu: true,
+		},
+	},
+
+	"providers.imageSize": {
+		type: "enum",
+		values: ["1024x1024", "1536x1024", "1024x1536"] as const,
+		default: "1536x1024",
+		ui: {
+			tab: "providers",
+			label: "Image Size",
+			description: "Default image dimensions for generation (landscape, square, or portrait)",
+			submenu: true,
+		},
+	},
+
+	"providers.imageQuality": {
+		type: "enum",
+		values: ["low", "medium", "high"] as const,
+		default: "high",
+		ui: {
+			tab: "providers",
+			label: "Image Quality",
+			description: "Rendering quality for generated images (higher = slower but more detailed)",
 			submenu: true,
 		},
 	},

package/src/modes/components/settings-defs.ts

@@ -337,10 +337,21 @@ const OPTION_PROVIDERS: Partial<Record<SettingPath, OptionProvider>> = {
 		{ value: "parallel", label: "Parallel", description: "Requires PARALLEL_API_KEY" },
 	],
 	"providers.image": [
-		{ value: "auto", label: "Auto", description: "Priority: OpenRouter > Gemini" },
+		{ value: "auto", label: "Auto", description: "Auto-detect from available API keys" },
+		{ value: "openai", label: "OpenAI", description: "gpt-image-1 via LITELLM_API_KEY or OPENAI_API_KEY" },
 		{ value: "gemini", label: "Gemini", description: "Requires GEMINI_API_KEY" },
 		{ value: "openrouter", label: "OpenRouter", description: "Requires OPENROUTER_API_KEY" },
 	],
+	"providers.imageSize": [
+		{ value: "1024x1024", label: "1024x1024", description: "Square" },
+		{ value: "1536x1024", label: "1536x1024", description: "Landscape (default)" },
+		{ value: "1024x1536", label: "1024x1536", description: "Portrait" },
+	],
+	"providers.imageQuality": [
+		{ value: "low", label: "Low", description: "Fastest generation, lower detail" },
+		{ value: "medium", label: "Medium", description: "Balanced speed and quality" },
+		{ value: "high", label: "High", description: "Best quality, slower generation (default)" },
+	],
 	"providers.kimiApiFormat": [
 		{ value: "openai", label: "OpenAI", description: "api.kimi.com" },
 		{ value: "anthropic", label: "Anthropic", description: "api.moonshot.ai" },

package/src/prompts/system/system-prompt.md

@@ -281,7 +281,13 @@ Don't open a file hoping. Hope is not a strategy.
 ### Image inspection
 - For image understanding tasks: **MUST** use `inspect_image` over `read` to avoid overloading main session context.
 - Write a specific `question` for `inspect_image`: what to inspect, constraints (for example verbatim OCR), and desired output format.
+- If you encounter `[Image content detected but current model does not support vision]` in a message, use `inspect_image` with the image file path to analyze it. Do not ask the user to describe the image — analyze it yourself via the tool.
 {{/if}}
+{{#ifAll (includes tools "inspect_image") (includes tools "generate_image")}}
+### Image generation and analysis
+- After using `generate_image`, the result includes saved file paths (e.g. `/tmp/xcsh-image-*.png`). To analyze or describe the generated image, chain `inspect_image` using that file path.
+- Example workflow: user asks "create a diagram and check if it follows brand guidelines" → call `generate_image`, then call `inspect_image` on the resulting file path with the brand compliance question.
+{{/ifAll}}
 
 {{SECTION_SEPERATOR "Rules"}}
 

package/src/sdk.ts

@@ -672,7 +672,12 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 	}
 
 	const imageProvider = settings.get("providers.image");
-	if (imageProvider === "auto" || imageProvider === "gemini" || imageProvider === "openrouter") {
+	if (
+		imageProvider === "auto" ||
+		imageProvider === "gemini" ||
+		imageProvider === "openrouter" ||
+		imageProvider === "openai"
+	) {
 		setPreferredImageProvider(imageProvider);
 	}
 
@@ -1034,10 +1039,12 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		}
 	}
 
-	// Add Gemini image tools if GEMINI_API_KEY (or GOOGLE_API_KEY) is available
-	const geminiImageTools = await logger.time("getGeminiImageTools", getGeminiImageTools);
-	if (geminiImageTools.length > 0) {
-		customTools.push(...(geminiImageTools as unknown as CustomTool[]));
+	// Add image generation tools if an image API key is available and the tool is enabled
+	if (settings.get("generate_image.enabled")) {
+		const geminiImageTools = await logger.time("getGeminiImageTools", getGeminiImageTools);
+		if (geminiImageTools.length > 0) {
+			customTools.push(...(geminiImageTools as unknown as CustomTool[]));
+		}
 	}
 
 	// Add web search tools
@@ -1435,11 +1442,37 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 		});
 	};
 
-	// Final convertToLlm: chain block-images filter with secret obfuscation
+	// Replace unsupported image content with actionable warnings when model lacks vision
+	const convertToLlmWithImageRouting = (messages: Message[]): Message[] => {
+		const currentModel = agent?.state?.model;
+		if (!currentModel || currentModel.input.includes("image")) return messages;
+
+		return messages.map(msg => {
+			if (msg.role !== "user" && msg.role !== "toolResult") return msg;
+			const content = msg.content;
+			if (!Array.isArray(content)) return msg;
+
+			const hasImages = content.some(c => c.type === "image");
+			if (!hasImages) return msg;
+
+			const filtered = content.map(c =>
+				c.type === "image"
+					? {
+							type: "text" as const,
+							text: "[Image content detected but current model does not support vision. Use the inspect_image tool to analyze this image, or ask the user to switch to a vision-capable model.]",
+						}
+					: c,
+			);
+			return { ...msg, content: filtered };
+		});
+	};
+
+	// Final convertToLlm: chain block-images filter → image routing warnings → secret obfuscation
 	const convertToLlmFinal = (messages: AgentMessage[]): Message[] => {
 		const converted = convertToLlmWithBlockImages(messages);
-		if (!obfuscator?.hasSecrets()) return converted;
-		return obfuscateMessages(obfuscator, converted);
+		const routed = convertToLlmWithImageRouting(converted);
+		if (!obfuscator?.hasSecrets()) return routed;
+		return obfuscateMessages(obfuscator, routed);
 	};
 	const transformContext = extensionRunner
 		? async (messages: AgentMessage[], _signal?: AbortSignal) => {

package/src/tools/gemini-image.ts

@@ -20,6 +20,9 @@ import { resolveReadPath } from "./path-utils";
 const DEFAULT_MODEL = "gemini-3-pro-image-preview";
 const DEFAULT_OPENROUTER_MODEL = "google/gemini-3-pro-image-preview";
 const DEFAULT_ANTIGRAVITY_MODEL = "gemini-3-pro-image";
+const DEFAULT_OPENAI_IMAGE_MODEL = "gpt-image-1";
+const DEFAULT_OPENAI_IMAGE_SIZE = "1536x1024";
+const DEFAULT_OPENAI_IMAGE_QUALITY = "high";
 const IMAGE_TIMEOUT = 3 * 60 * 1000; // 3 minutes
 const MAX_IMAGE_SIZE = 35 * 1024 * 1024;
 
@@ -27,7 +30,7 @@ const ANTIGRAVITY_ENDPOINT = "https://daily-cloudcode-pa.sandbox.googleapis.com"
 const IMAGE_SYSTEM_INSTRUCTION =
 	"You are an AI image generator. Generate images based on user descriptions. Focus on creating high-quality, visually appealing images that match the user's request.";
 
-type ImageProvider = "antigravity" | "gemini" | "openrouter";
+type ImageProvider = "antigravity" | "gemini" | "openrouter" | "openai";
 interface ImageApiKey {
 	provider: ImageProvider;
 	apiKey: string;
@@ -207,6 +210,21 @@ interface OpenRouterResponse {
 	choices?: OpenRouterChoice[];
 }
 
+interface OpenAIImageResponseData {
+	b64_json: string;
+	revised_prompt?: string | null;
+}
+
+interface OpenAIImageResponse {
+	created: number;
+	data: OpenAIImageResponseData[];
+	usage?: {
+		total_tokens: number;
+		input_tokens: number;
+		output_tokens: number;
+	};
+}
+
 interface AntigravityRequest {
 	project: string;
 	model: string;
@@ -396,9 +414,13 @@ async function findImageApiKey(modelRegistry?: ModelRegistry): Promise<ImageApiK
 		const openRouterKey = getEnvApiKey("openrouter");
 		if (openRouterKey) return { provider: "openrouter", apiKey: openRouterKey };
 		// Fall through to auto-detect if preferred provider key not found
+	} else if (preferredImageProvider === "openai") {
+		const openaiKey = getEnvApiKey("litellm") ?? getEnvApiKey("openai");
+		if (openaiKey) return { provider: "openai", apiKey: openaiKey };
+		// Fall through to auto-detect if preferred provider key not found
 	}
 
-	// Auto-detect: Antigravity takes priority, then OpenRouter, then Gemini
+	// Auto-detect: Antigravity takes priority, then OpenRouter, then OpenAI, then Gemini
 	if (modelRegistry) {
 		const antigravity = await findAntigravityCredentials(modelRegistry);
 		if (antigravity) return antigravity;
@@ -407,6 +429,9 @@ async function findImageApiKey(modelRegistry?: ModelRegistry): Promise<ImageApiK
 	const openRouterKey = getEnvApiKey("openrouter");
 	if (openRouterKey) return { provider: "openrouter", apiKey: openRouterKey };
 
+	const openaiKey = getEnvApiKey("litellm") ?? getEnvApiKey("openai");
+	if (openaiKey) return { provider: "openai", apiKey: openaiKey };
+
 	const geminiKey = getEnvApiKey("google");
 	if (geminiKey) return { provider: "gemini", apiKey: geminiKey };
 
@@ -614,7 +639,7 @@ export const geminiImageTool: CustomTool<typeof geminiImageSchema, GeminiImageTo
 			const apiKey = await findImageApiKey(ctx.modelRegistry);
 			if (!apiKey) {
 				throw new Error(
-					"No image API credentials found. Login with google-antigravity, or set OPENROUTER_API_KEY, GEMINI_API_KEY, or GOOGLE_API_KEY.",
+					"No image API credentials found. Set LITELLM_API_KEY, OPENAI_API_KEY, OPENROUTER_API_KEY, GEMINI_API_KEY, or GOOGLE_API_KEY.",
 				);
 			}
 
@@ -624,7 +649,9 @@ export const geminiImageTool: CustomTool<typeof geminiImageSchema, GeminiImageTo
 					? DEFAULT_ANTIGRAVITY_MODEL
 					: provider === "openrouter"
 						? DEFAULT_OPENROUTER_MODEL
-						: DEFAULT_MODEL;
+						: provider === "openai"
+							? DEFAULT_OPENAI_IMAGE_MODEL
+							: DEFAULT_MODEL;
 			const resolvedModel = provider === "openrouter" ? resolveOpenRouterModel(model) : model;
 			const cwd = ctx.sessionManager.getCwd();
 
@@ -786,6 +813,86 @@ export const geminiImageTool: CustomTool<typeof geminiImageSchema, GeminiImageTo
 				};
 			}
 
+			if (provider === "openai") {
+				const openaiPrompt = assemblePrompt(params);
+				const size = params.image_size ?? ctx.settings?.get("providers.imageSize") ?? DEFAULT_OPENAI_IMAGE_SIZE;
+				const quality = ctx.settings?.get("providers.imageQuality") ?? DEFAULT_OPENAI_IMAGE_QUALITY;
+				const baseUrl = $env.LITELLM_BASE_URL ?? $env.OPENAI_BASE_URL ?? "https://api.openai.com";
+
+				const requestBody = {
+					model: DEFAULT_OPENAI_IMAGE_MODEL,
+					prompt: openaiPrompt,
+					n: 1,
+					size,
+					quality,
+				};
+
+				const response = await fetch(`${baseUrl}/openai/v1/images/generations`, {
+					method: "POST",
+					headers: {
+						Authorization: `Bearer ${apiKey.apiKey}`,
+						"Content-Type": "application/json",
+					},
+					body: JSON.stringify(requestBody),
+					signal: requestSignal,
+				});
+
+				const rawText = await response.text();
+				if (!response.ok) {
+					let message = rawText;
+					try {
+						const parsed = JSON.parse(rawText) as { error?: { message?: string } };
+						message = parsed.error?.message ?? message;
+					} catch {
+						// Keep raw text.
+					}
+					throw new Error(`OpenAI image request failed (${response.status}): ${message}`);
+				}
+
+				const data = JSON.parse(rawText) as OpenAIImageResponse;
+				const b64 = data.data?.[0]?.b64_json;
+				if (!b64) {
+					return {
+						content: [{ type: "text", text: "No image data returned from OpenAI." }],
+						details: {
+							provider,
+							model: DEFAULT_OPENAI_IMAGE_MODEL,
+							imageCount: 0,
+							imagePaths: [],
+							images: [],
+						},
+					};
+				}
+
+				const image: InlineImageData = { data: b64, mimeType: "image/png" };
+				const imagePaths = await saveImagesToTemp([image]);
+				const revisedPrompt = data.data[0]?.revised_prompt ?? undefined;
+
+				return {
+					content: [
+						{
+							type: "text",
+							text: buildResponseSummary(provider, DEFAULT_OPENAI_IMAGE_MODEL, imagePaths, revisedPrompt),
+						},
+					],
+					details: {
+						provider,
+						model: DEFAULT_OPENAI_IMAGE_MODEL,
+						imageCount: 1,
+						imagePaths,
+						images: [image],
+						responseText: revisedPrompt,
+						usage: data.usage
+							? {
+									promptTokenCount: data.usage.input_tokens,
+									candidatesTokenCount: data.usage.output_tokens,
+									totalTokenCount: data.usage.total_tokens,
+								}
+							: undefined,
+					},
+				};
+			}
+
 			const parts = [] as Array<{ text?: string; inlineData?: InlineImageData }>;
 			for (const image of resolvedImages) {
 				parts.push({ inlineData: image });