@oh-my-pi/pi-ai 3.20.1 → 3.35.0

package/src/providers/google-gemini-cli.ts

@@ -18,8 +18,16 @@ import type {
 	ToolCall,
 } from "../types";
 import { AssistantMessageEventStream } from "../utils/event-stream";
+import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
 import { sanitizeSurrogates } from "../utils/sanitize-unicode";
-import { convertMessages, convertTools, mapStopReasonString, mapToolChoice } from "./google-shared";
+import {
+	convertMessages,
+	convertTools,
+	isThinkingPart,
+	mapStopReasonString,
+	mapToolChoice,
+	retainThoughtSignature,
+} from "./google-shared";
 
 /**
  * Thinking level for Gemini 3 models.
@@ -69,6 +77,87 @@ const ANTIGRAVITY_HEADERS = {
 	}),
 };
 
+// Antigravity system instruction (ported from CLIProxyAPI v6.6.89).
+const ANTIGRAVITY_SYSTEM_INSTRUCTION = `<identity>
+You are Antigravity, a powerful agentic AI coding assistant designed by the Google DeepMind team working on Advanced Agentic Coding.
+You are pair programming with a USER to solve their coding task. The task may require creating a new codebase, modifying or debugging an existing codebase, or simply answering a question.
+The USER will send you requests, which you must always prioritize addressing. Along with each USER request, we will attach additional metadata about their current state, such as what files they have open and where their cursor is.
+This information may or may not be relevant to the coding task, it is up for you to decide.
+</identity>
+
+<tool_calling>
+Call tools as you normally would. The following list provides additional guidance to help you avoid errors:
+  - **Absolute paths only**. When using tools that accept file path arguments, ALWAYS use the absolute file path.
+</tool_calling>
+
+<web_application_development>
+## Technology Stack
+Your web applications should be built using the following technologies:
+1. **Core**: Use HTML for structure and JavaScript for logic.
+2. **Styling (CSS)**: Use Vanilla CSS for maximum flexibility and control. Avoid using TailwindCSS unless the USER explicitly requests it; in this case, first confirm which TailwindCSS version to use.
+3. **Web App**: If the USER specifies that they want a more complex web app, use a framework like Next.js or Vite. Only do this if the USER explicitly requests a web app.
+4. **New Project Creation**: If you need to use a framework for a new app, use \`npx\` with the appropriate script, but there are some rules to follow:
+   - Use \`npx -y\` to automatically install the script and its dependencies
+   - You MUST run the command with \`--help\` flag to see all available options first
+   - Initialize the app in the current directory with \`./\` (example: \`npx -y create-vite-app@latest ./\`)
+   - You should run in non-interactive mode so that the user doesn't need to input anything
+5. **Running Locally**: When running locally, use \`npm run dev\` or equivalent dev server. Only build the production bundle if the USER explicitly requests it or you are validating the code for correctness.
+
+# Design Aesthetics
+1. **Use Rich Aesthetics**: The USER should be wowed at first glance by the design. Use best practices in modern web design (e.g. vibrant colors, dark modes, glassmorphism, and dynamic animations) to create a stunning first impression. Failure to do this is UNACCEPTABLE.
+2. **Prioritize Visual Excellence**: Implement designs that will WOW the user and feel extremely premium:
+   - Avoid generic colors (plain red, blue, green). Use curated, harmonious color palettes (e.g., HSL tailored colors, sleek dark modes).
+   - Using modern typography (e.g., from Google Fonts like Inter, Roboto, or Outfit) instead of browser defaults.
+   - Use smooth gradients
+   - Add subtle micro-animations for enhanced user experience
+3. **Use a Dynamic Design**: An interface that feels responsive and alive encourages interaction. Achieve this with hover effects and interactive elements. Micro-animations, in particular, are highly effective for improving user engagement.
+4. **Premium Designs**: Make a design that feels premium and state of the art. Avoid creating simple minimum viable products.
+5. **Don't use placeholders**: If you need an image, use your generate_image tool to create a working demonstration.
+
+## Implementation Workflow
+Follow this systematic approach when building web applications:
+1. **Plan and Understand**:
+   - Fully understand the user's requirements
+   - Draw inspiration from modern, beautiful, and dynamic web designs
+   - Outline the features needed for the initial version
+2. **Build the Foundation**:
+   - Start by creating/modifying \`index.css\`
+   - Implement the core design system with all tokens and utilities
+3. **Create Components**:
+   - Build necessary components using your design system
+   - Ensure all components use predefined styles, not ad-hoc utilities
+   - Keep components focused and reusable
+4. **Assemble Pages**:
+   - Update the main application to incorporate your design and components
+   - Ensure proper routing and navigation
+   - Implement responsive layouts
+5. **Polish and Optimize**:
+   - Review the overall user experience
+   - Ensure smooth interactions and transitions
+   - Optimize performance where needed
+
+## SEO Best Practices
+Automatically implement SEO best practices on every page:
+- **Title Tags**: Include proper, descriptive title tags for each page
+- **Meta Descriptions**: Add compelling meta descriptions that accurately summarize page content
+- **Heading Structure**: Use a single \`<h1>\` per page with proper heading hierarchy
+- **Semantic HTML**: Use appropriate HTML5 semantic elements
+- **Unique IDs**: Ensure all interactive elements have unique, descriptive IDs for browser testing
+- **Performance**: Ensure fast page load times through optimization
+CRITICAL REMINDER: AESTHETICS ARE VERY IMPORTANT. If your web app looks simple and basic then you have FAILED!
+</web_application_development>
+<ephemeral_message>
+There will be an <EPHEMERAL_MESSAGE> appearing in the conversation at times. This is not coming from the user, but instead injected by the system as important information to pay attention to. 
+Do not respond to nor acknowledge those messages, but do follow them strictly.
+</ephemeral_message>
+
+<communication_style>
+- **Formatting**. Format your responses in github-style markdown to make your responses easier for the USER to parse. For example, use headers to organize your responses and bolded or italicized text to highlight important keywords. Use backticks to format file, directory, function, and class names. If providing a URL to the user, format this in markdown as well, for example \`[label](example.com)\`.
+- **Proactiveness**. As an agent, you are allowed to be proactive, but only in the course of completing the user's task. For example, if the user asks you to add a new component, you can edit the code, verify build and test statuses, and take any other obvious follow-up actions, such as performing additional research. However, avoid surprising the user. For example, if the user asks HOW to approach something, you should answer their question and instead of jumping into editing a file.
+- **Helpfulness**. Respond like a helpful software engineer who is explaining your work to a friendly collaborator on the project. Acknowledge mistakes or any backtracking you do as a result of new information.
+- **Ask for clarification**. If you are unsure about the USER's intent, always ask for clarification rather than making assumptions.
+</communication_style>`;
+
 // Counter for generating unique tool call IDs
 let toolCallCounter = 0;
 
@@ -154,7 +243,7 @@ interface CloudCodeAssistRequest {
 	model: string;
 	request: {
 		contents: Content[];
-		systemInstruction?: { parts: { text: string }[] };
+		systemInstruction?: { role?: string; parts: { text: string }[] };
 		generationConfig?: {
 			maxOutputTokens?: number;
 			temperature?: number;
@@ -167,6 +256,7 @@ interface CloudCodeAssistRequest {
 			};
 		};
 	};
+	requestType?: string;
 	userAgent?: string;
 	requestId?: string;
 }
@@ -250,12 +340,12 @@ export const streamGoogleGeminiCli: StreamFunction<"google-gemini-cli"> = (
 				throw new Error("Missing token or projectId in Google Cloud credentials. Use /login to re-authenticate.");
 			}
 
-			const requestBody = buildRequest(model, context, projectId, options);
 			const endpoint = model.baseUrl || DEFAULT_ENDPOINT;
 			const url = `${endpoint}/v1internal:streamGenerateContent?alt=sse`;
 
 			// Use Antigravity headers for sandbox endpoint, otherwise Gemini CLI headers
 			const isAntigravity = endpoint.includes("sandbox.googleapis.com");
+			const requestBody = buildRequest(model, context, projectId, options, isAntigravity);
 			const headers = isAntigravity ? ANTIGRAVITY_HEADERS : GEMINI_CLI_HEADERS;
 
 			// Fetch with retry logic for rate limits and transient errors
@@ -298,8 +388,11 @@ export const streamGoogleGeminiCli: StreamFunction<"google-gemini-cli"> = (
 					// Not retryable or max retries exceeded
 					throw new Error(`Cloud Code Assist API error (${response.status}): ${errorText}`);
 				} catch (error) {
-					if (error instanceof Error && error.message === "Request was aborted") {
-						throw error;
+					// Check for abort - fetch throws AbortError, our code throws "Request was aborted"
+					if (error instanceof Error) {
+						if (error.name === "AbortError" || error.message === "Request was aborted") {
+							throw new Error("Request was aborted");
+						}
 					}
 					lastError = error instanceof Error ? error : new Error(String(error));
 					// Network errors are retryable
@@ -331,46 +424,109 @@ export const streamGoogleGeminiCli: StreamFunction<"google-gemini-cli"> = (
 			const decoder = new TextDecoder();
 			let buffer = "";
 
-			while (true) {
-				const { done, value } = await reader.read();
-				if (done) break;
+			// Set up abort handler to cancel reader when signal fires
+			const abortHandler = () => {
+				void reader.cancel().catch(() => {});
+			};
+			options?.signal?.addEventListener("abort", abortHandler);
+
+			try {
+				while (true) {
+					// Check abort signal before each read
+					if (options?.signal?.aborted) {
+						throw new Error("Request was aborted");
+					}
 
-				buffer += decoder.decode(value, { stream: true });
-				const lines = buffer.split("\n");
-				buffer = lines.pop() || "";
+					const { done, value } = await reader.read();
+					if (done) break;
 
-				for (const line of lines) {
-					if (!line.startsWith("data:")) continue;
+					buffer += decoder.decode(value, { stream: true });
+					const lines = buffer.split("\n");
+					buffer = lines.pop() || "";
 
-					const jsonStr = line.slice(5).trim();
-					if (!jsonStr) continue;
+					for (const line of lines) {
+						if (!line.startsWith("data:")) continue;
 
-					let chunk: CloudCodeAssistResponseChunk;
-					try {
-						chunk = JSON.parse(jsonStr);
-					} catch {
-						continue;
-					}
+						const jsonStr = line.slice(5).trim();
+						if (!jsonStr) continue;
+
+						let chunk: CloudCodeAssistResponseChunk;
+						try {
+							chunk = JSON.parse(jsonStr);
+						} catch {
+							continue;
+						}
+
+						// Unwrap the response
+						const responseData = chunk.response;
+						if (!responseData) continue;
+
+						const candidate = responseData.candidates?.[0];
+						if (candidate?.content?.parts) {
+							for (const part of candidate.content.parts) {
+								if (part.text !== undefined) {
+									const isThinking = isThinkingPart(part);
+									if (
+										!currentBlock ||
+										(isThinking && currentBlock.type !== "thinking") ||
+										(!isThinking && currentBlock.type !== "text")
+									) {
+										if (currentBlock) {
+											if (currentBlock.type === "text") {
+												stream.push({
+													type: "text_end",
+													contentIndex: blocks.length - 1,
+													content: currentBlock.text,
+													partial: output,
+												});
+											} else {
+												stream.push({
+													type: "thinking_end",
+													contentIndex: blockIndex(),
+													content: currentBlock.thinking,
+													partial: output,
+												});
+											}
+										}
+										if (isThinking) {
+											currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
+											output.content.push(currentBlock);
+											stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
+										} else {
+											currentBlock = { type: "text", text: "" };
+											output.content.push(currentBlock);
+											stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
+										}
+									}
+									if (currentBlock.type === "thinking") {
+										currentBlock.thinking += part.text;
+										currentBlock.thinkingSignature = retainThoughtSignature(
+											currentBlock.thinkingSignature,
+											part.thoughtSignature,
+										);
+										stream.push({
+											type: "thinking_delta",
+											contentIndex: blockIndex(),
+											delta: part.text,
+											partial: output,
+										});
+									} else {
+										currentBlock.text += part.text;
+										stream.push({
+											type: "text_delta",
+											contentIndex: blockIndex(),
+											delta: part.text,
+											partial: output,
+										});
+									}
+								}
 
-					// Unwrap the response
-					const responseData = chunk.response;
-					if (!responseData) continue;
-
-					const candidate = responseData.candidates?.[0];
-					if (candidate?.content?.parts) {
-						for (const part of candidate.content.parts) {
-							if (part.text !== undefined) {
-								const isThinking = part.thought === true;
-								if (
-									!currentBlock ||
-									(isThinking && currentBlock.type !== "thinking") ||
-									(!isThinking && currentBlock.type !== "text")
-								) {
+								if (part.functionCall) {
 									if (currentBlock) {
 										if (currentBlock.type === "text") {
 											stream.push({
 												type: "text_end",
-												contentIndex: blocks.length - 1,
+												contentIndex: blockIndex(),
 												content: currentBlock.text,
 												partial: output,
 											});
@@ -382,115 +538,70 @@ export const streamGoogleGeminiCli: StreamFunction<"google-gemini-cli"> = (
 												partial: output,
 											});
 										}
+										currentBlock = null;
 									}
-									if (isThinking) {
-										currentBlock = { type: "thinking", thinking: "", thinkingSignature: undefined };
-										output.content.push(currentBlock);
-										stream.push({ type: "thinking_start", contentIndex: blockIndex(), partial: output });
-									} else {
-										currentBlock = { type: "text", text: "" };
-										output.content.push(currentBlock);
-										stream.push({ type: "text_start", contentIndex: blockIndex(), partial: output });
-									}
-								}
-								if (currentBlock.type === "thinking") {
-									currentBlock.thinking += part.text;
-									currentBlock.thinkingSignature = part.thoughtSignature;
-									stream.push({
-										type: "thinking_delta",
-										contentIndex: blockIndex(),
-										delta: part.text,
-										partial: output,
-									});
-								} else {
-									currentBlock.text += part.text;
+
+									const providedId = part.functionCall.id;
+									const needsNewId =
+										!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
+									const toolCallId = needsNewId
+										? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
+										: providedId;
+
+									const toolCall: ToolCall = {
+										type: "toolCall",
+										id: toolCallId,
+										name: part.functionCall.name || "",
+										arguments: part.functionCall.args as Record<string, unknown>,
+										...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
+									};
+
+									output.content.push(toolCall);
+									stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
 									stream.push({
-										type: "text_delta",
+										type: "toolcall_delta",
 										contentIndex: blockIndex(),
-										delta: part.text,
+										delta: JSON.stringify(toolCall.arguments),
 										partial: output,
 									});
+									stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
 								}
 							}
-
-							if (part.functionCall) {
-								if (currentBlock) {
-									if (currentBlock.type === "text") {
-										stream.push({
-											type: "text_end",
-											contentIndex: blockIndex(),
-											content: currentBlock.text,
-											partial: output,
-										});
-									} else {
-										stream.push({
-											type: "thinking_end",
-											contentIndex: blockIndex(),
-											content: currentBlock.thinking,
-											partial: output,
-										});
-									}
-									currentBlock = null;
-								}
-
-								const providedId = part.functionCall.id;
-								const needsNewId =
-									!providedId || output.content.some((b) => b.type === "toolCall" && b.id === providedId);
-								const toolCallId = needsNewId
-									? `${part.functionCall.name}_${Date.now()}_${++toolCallCounter}`
-									: providedId;
-
-								const toolCall: ToolCall = {
-									type: "toolCall",
-									id: toolCallId,
-									name: part.functionCall.name || "",
-									arguments: part.functionCall.args as Record<string, unknown>,
-									...(part.thoughtSignature && { thoughtSignature: part.thoughtSignature }),
-								};
-
-								output.content.push(toolCall);
-								stream.push({ type: "toolcall_start", contentIndex: blockIndex(), partial: output });
-								stream.push({
-									type: "toolcall_delta",
-									contentIndex: blockIndex(),
-									delta: JSON.stringify(toolCall.arguments),
-									partial: output,
-								});
-								stream.push({ type: "toolcall_end", contentIndex: blockIndex(), toolCall, partial: output });
-							}
 						}
-					}
 
-					if (candidate?.finishReason) {
-						output.stopReason = mapStopReasonString(candidate.finishReason);
-						if (output.content.some((b) => b.type === "toolCall")) {
-							output.stopReason = "toolUse";
+						if (candidate?.finishReason) {
+							output.stopReason = mapStopReasonString(candidate.finishReason);
+							if (output.content.some((b) => b.type === "toolCall")) {
+								output.stopReason = "toolUse";
+							}
 						}
-					}
 
-					if (responseData.usageMetadata) {
-						// promptTokenCount includes cachedContentTokenCount, so subtract to get fresh input
-						const promptTokens = responseData.usageMetadata.promptTokenCount || 0;
-						const cacheReadTokens = responseData.usageMetadata.cachedContentTokenCount || 0;
-						output.usage = {
-							input: promptTokens - cacheReadTokens,
-							output:
-								(responseData.usageMetadata.candidatesTokenCount || 0) +
-								(responseData.usageMetadata.thoughtsTokenCount || 0),
-							cacheRead: cacheReadTokens,
-							cacheWrite: 0,
-							totalTokens: responseData.usageMetadata.totalTokenCount || 0,
-							cost: {
-								input: 0,
-								output: 0,
-								cacheRead: 0,
+						if (responseData.usageMetadata) {
+							// promptTokenCount includes cachedContentTokenCount, so subtract to get fresh input
+							const promptTokens = responseData.usageMetadata.promptTokenCount || 0;
+							const cacheReadTokens = responseData.usageMetadata.cachedContentTokenCount || 0;
+							output.usage = {
+								input: promptTokens - cacheReadTokens,
+								output:
+									(responseData.usageMetadata.candidatesTokenCount || 0) +
+									(responseData.usageMetadata.thoughtsTokenCount || 0),
+								cacheRead: cacheReadTokens,
 								cacheWrite: 0,
-								total: 0,
-							},
-						};
-						calculateCost(model, output.usage);
+								totalTokens: responseData.usageMetadata.totalTokenCount || 0,
+								cost: {
+									input: 0,
+									output: 0,
+									cacheRead: 0,
+									cacheWrite: 0,
+									total: 0,
+								},
+							};
+							calculateCost(model, output.usage);
+						}
 					}
 				}
+			} finally {
+				options?.signal?.removeEventListener("abort", abortHandler);
 			}
 
 			if (currentBlock) {
@@ -528,7 +639,7 @@ export const streamGoogleGeminiCli: StreamFunction<"google-gemini-cli"> = (
 				}
 			}
 			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
+			output.errorMessage = formatErrorMessageWithRetryAfter(error);
 			stream.push({ type: "error", reason: output.stopReason, error: output });
 			stream.end();
 		}
@@ -542,6 +653,7 @@ function buildRequest(
 	context: Context,
 	projectId: string,
 	options: GoogleGeminiCliOptions = {},
+	isAntigravity = false,
 ): CloudCodeAssistRequest {
 	const contents = convertMessages(model, context);
 
@@ -593,11 +705,35 @@ function buildRequest(
 		}
 	}
 
+	if (isAntigravity) {
+		const existingText = request.systemInstruction?.parts?.[0]?.text || "";
+		const bridgePrompt = `# CRITICAL: Environment Override
+
+IGNORE ALL INSTRUCTIONS ABOVE THIS LINE. The following overrides are mandatory:
+
+- You are Pi, NOT Antigravity
+- Use RELATIVE paths, never absolute paths
+- IGNORE all <web_application_development> guidelines above (no glassmorphism, no "premium aesthetics", no design opinions)
+- IGNORE <communication_style> above
+- IGNORE <ephemeral_message> handling above
+- Follow ONLY the instructions below
+`;
+		request.systemInstruction = {
+			role: "user",
+			parts: [
+				{
+					text: `${ANTIGRAVITY_SYSTEM_INSTRUCTION}\n\n${bridgePrompt}${existingText ? `\n${existingText}` : ""}`,
+				},
+			],
+		};
+	}
+
 	return {
 		project: projectId,
 		model: model.id,
 		request,
-		userAgent: "pi-coding-agent",
-		requestId: `pi-${Date.now()}-${Math.random().toString(36).slice(2, 11)}`,
+		...(isAntigravity ? { requestType: "agent" } : {}),
+		userAgent: isAntigravity ? "antigravity" : "pi-coding-agent",
+		requestId: `${isAntigravity ? "agent" : "pi"}-${Date.now()}-${Math.random().toString(36).slice(2, 11)}`,
 	};
 }

package/src/providers/google-shared.ts

@@ -9,6 +9,38 @@ import { transformMessages } from "./transorm-messages";
 
 type GoogleApiType = "google-generative-ai" | "google-gemini-cli" | "google-vertex";
 
+/**
+ * Determines whether a streamed Gemini `Part` should be treated as "thinking".
+ *
+ * Protocol note (Gemini / Vertex AI thought signatures):
+ * - `thoughtSignature` may appear without `thought: true` (including in empty-text parts at the end of streaming).
+ * - When persisting/replaying model outputs, signature-bearing parts must be preserved as-is;
+ *   do not merge/move signatures across parts.
+ * - Our streaming representation uses content blocks, so we classify any non-empty `thoughtSignature`
+ *   as thinking to avoid leaking thought content into normal assistant text.
+ *
+ * Some Google backends send thought content with `thoughtSignature` but omit `thought: true`
+ * on subsequent deltas. We treat any non-empty `thoughtSignature` as thinking to avoid
+ * leaking thought text into the normal assistant text stream.
+ */
+export function isThinkingPart(part: Pick<Part, "thought" | "thoughtSignature">): boolean {
+	return part.thought === true || (typeof part.thoughtSignature === "string" && part.thoughtSignature.length > 0);
+}
+
+/**
+ * Retain thought signatures during streaming.
+ *
+ * Some backends only send `thoughtSignature` on the first delta for a given part/block; later deltas may omit it.
+ * This helper preserves the last non-empty signature for the current block.
+ *
+ * Note: this does NOT merge or move signatures across distinct response parts. It only prevents
+ * a signature from being overwritten with `undefined` within the same streamed block.
+ */
+export function retainThoughtSignature(existing: string | undefined, incoming: string | undefined): string | undefined {
+	if (typeof incoming === "string" && incoming.length > 0) return incoming;
+	return existing;
+}
+
 /**
  * Convert internal messages to Gemini Content[] format.
  */
@@ -45,6 +77,8 @@ export function convertMessages<T extends GoogleApiType>(model: Model<T>, contex
 			}
 		} else if (msg.role === "assistant") {
 			const parts: Part[] = [];
+			// Check if message is from same provider and model - only then keep thinking blocks
+			const isSameProviderAndModel = msg.provider === model.provider && msg.model === model.id;
 
 			for (const block of msg.content) {
 				if (block.type === "text") {
@@ -52,17 +86,19 @@ export function convertMessages<T extends GoogleApiType>(model: Model<T>, contex
 					if (!block.text || block.text.trim() === "") continue;
 					parts.push({ text: sanitizeSurrogates(block.text) });
 				} else if (block.type === "thinking") {
-					// Thinking blocks require signatures for Claude via Antigravity.
-					// If signature is missing (e.g. from GPT-OSS), convert to regular text with delimiters.
-					if (block.thinkingSignature) {
+					// Skip empty thinking blocks
+					if (!block.thinking || block.thinking.trim() === "") continue;
+					// Only keep as thinking block if same provider AND same model
+					// Otherwise convert to plain text (no tags to avoid model mimicking them)
+					if (isSameProviderAndModel) {
 						parts.push({
 							thought: true,
 							text: sanitizeSurrogates(block.thinking),
-							thoughtSignature: block.thinkingSignature,
+							...(block.thinkingSignature && { thoughtSignature: block.thinkingSignature }),
 						});
 					} else {
 						parts.push({
-							text: `<thinking>\n${sanitizeSurrogates(block.thinking)}\n</thinking>`,
+							text: sanitizeSurrogates(block.thinking),
 						});
 					}
 				} else if (block.type === "toolCall") {
@@ -73,6 +109,9 @@ export function convertMessages<T extends GoogleApiType>(model: Model<T>, contex
 							args: block.arguments,
 						},
 					};
+					if (model.provider === "google-vertex" && part?.functionCall?.id) {
+						delete part.functionCall.id; // Vertex AI does not support 'id' in functionCall
+					}
 					if (block.thoughtSignature) {
 						part.thoughtSignature = block.thoughtSignature;
 					}
@@ -121,6 +160,10 @@ export function convertMessages<T extends GoogleApiType>(model: Model<T>, contex
 				},
 			};
 
+			if (model.provider === "google-vertex" && functionResponsePart.functionResponse?.id) {
+				delete functionResponsePart.functionResponse.id; // Vertex AI does not support 'id' in functionResponse
+			}
+
 			// Cloud Code Assist API requires all function responses to be in a single user turn.
 			// Check if the last content is already a user turn with function responses and merge.
 			const lastContent = contents[contents.length - 1];

package/src/providers/google-vertex.ts

@@ -18,9 +18,17 @@ import type {
 	ToolCall,
 } from "../types";
 import { AssistantMessageEventStream } from "../utils/event-stream";
+import { formatErrorMessageWithRetryAfter } from "../utils/retry-after";
 import { sanitizeSurrogates } from "../utils/sanitize-unicode";
 import type { GoogleThinkingLevel } from "./google-gemini-cli";
-import { convertMessages, convertTools, mapStopReason, mapToolChoice } from "./google-shared";
+import {
+	convertMessages,
+	convertTools,
+	isThinkingPart,
+	mapStopReason,
+	mapToolChoice,
+	retainThoughtSignature,
+} from "./google-shared";
 
 export interface GoogleVertexOptions extends StreamOptions {
 	toolChoice?: "auto" | "none" | "any";
@@ -88,7 +96,7 @@ export const streamGoogleVertex: StreamFunction<"google-vertex"> = (
 				if (candidate?.content?.parts) {
 					for (const part of candidate.content.parts) {
 						if (part.text !== undefined) {
-							const isThinking = part.thought === true;
+							const isThinking = isThinkingPart(part);
 							if (
 								!currentBlock ||
 								(isThinking && currentBlock.type !== "thinking") ||
@@ -123,7 +131,10 @@ export const streamGoogleVertex: StreamFunction<"google-vertex"> = (
 							}
 							if (currentBlock.type === "thinking") {
 								currentBlock.thinking += part.text;
-								currentBlock.thinkingSignature = part.thoughtSignature;
+								currentBlock.thinkingSignature = retainThoughtSignature(
+									currentBlock.thinkingSignature,
+									part.thoughtSignature,
+								);
 								stream.push({
 									type: "thinking_delta",
 									contentIndex: blockIndex(),
@@ -252,7 +263,7 @@ export const streamGoogleVertex: StreamFunction<"google-vertex"> = (
 				}
 			}
 			output.stopReason = options?.signal?.aborted ? "aborted" : "error";
-			output.errorMessage = error instanceof Error ? error.message : JSON.stringify(error);
+			output.errorMessage = formatErrorMessageWithRetryAfter(error);
 			stream.push({ type: "error", reason: output.stopReason, error: output });
 			stream.end();
 		}