feat(ai): Add start event emission to all providers

- Emit start event with model and provider info after creating stream
- Add abort signal tests for all providers
- Update README abort signal section to reflect non-throwing API
- Fix model references in README examples

packages/ai/src/providers/anthropic.ts

@@ -129,6 +129,8 @@ export class AnthropicLLM implements LLM<AnthropicLLMOptions> {
 				},
 			);
 
+			options?.onEvent?.({ type: "start", model: this.modelInfo.id, provider: this.modelInfo.provider });
+
 			let blockType: "text" | "thinking" | "toolUse" | "other" = "other";
 			let blockContent = "";
 			let toolCall: (ToolCall & { partialJson: string }) | null = null;

packages/ai/src/providers/google.ts

@@ -89,6 +89,14 @@ export class GoogleLLM implements LLM<GoogleLLMOptions> {
 				};
 			}
 
+			// Abort signal
+			if (options?.signal) {
+				if (options.signal.aborted) {
+					throw new Error("Request aborted");
+				}
+				config.abortSignal = options.signal;
+			}
+
 			// Build the request parameters
 			const params: GenerateContentParameters = {
 				model: this.model.id,
@@ -98,6 +106,8 @@ export class GoogleLLM implements LLM<GoogleLLMOptions> {
 
 			const stream = await this.client.models.generateContentStream(params);
 
+			options?.onEvent?.({ type: "start", model: this.model.id, provider: this.model.provider });
+
 			const blocks: AssistantMessage["content"] = [];
 			let currentBlock: TextContent | ThinkingContent | null = null;
 			let usage: Usage = {

packages/ai/src/providers/openai-completions.ts

@@ -92,6 +92,8 @@ export class OpenAICompletionsLLM implements LLM<OpenAICompletionsLLMOptions> {
 				signal: options?.signal,
 			});
 
+			options?.onEvent?.({ type: "start", model: this.modelInfo.id, provider: this.modelInfo.provider });
+
 			const blocks: AssistantMessage["content"] = [];
 			let currentBlock: TextContent | ThinkingContent | (ToolCall & { partialArgs?: string }) | null = null;
 			let usage: Usage = {

packages/ai/src/providers/openai-responses.ts

@@ -85,6 +85,8 @@ export class OpenAIResponsesLLM implements LLM<OpenAIResponsesLLMOptions> {
 				signal: options?.signal,
 			});
 
+			options?.onEvent?.({ type: "start", model: this.modelInfo.id, provider: this.modelInfo.provider });
+
 			const outputItems: (ResponseReasoningItem | ResponseOutputMessage | ResponseFunctionToolCall)[] = []; // any for function_call items
 			let currentTextAccum = ""; // For delta accumulation
 			let currentThinkingAccum = ""; // For delta accumulation
@@ -184,9 +186,25 @@ export class OpenAIResponsesLLM implements LLM<OpenAIResponsesLLMOptions> {
 					} satisfies AssistantMessage;
 					options?.onEvent?.({ type: "error", error: errorOutput.error || "Unknown error" });
 					return errorOutput;
+				} else if (event.type === "response.failed") {
+					const errorOutput = {
+						role: "assistant",
+						content: [],
+						provider: this.modelInfo.provider,
+						model: this.modelInfo.id,
+						usage,
+						stopReason: "error",
+						error: "Unknown error",
+					} satisfies AssistantMessage;
+					options?.onEvent?.({ type: "error", error: errorOutput.error || "Unknown error" });
+					return errorOutput;
 				}
 			}
 
+			if (options?.signal?.aborted) {
+				throw new Error("Request was aborted");
+			}
+
 			// Convert output items to blocks
 			const blocks: AssistantMessage["content"] = [];