revert

2026-05-04 15:43:45 +02:00
parent 59a89e0c1c
commit cbf3c333ef
14 changed files with 31 additions and 395 deletions
--- a/packages/ai/src/providers/faux.ts
+++ b/packages/ai/src/providers/faux.ts
@@ -147,7 +147,7 @@ function contentToText(content: string | Array<TextContent | ImageContent>): str
 		.join("\n");
 }

-function assistantContentToText(content: Array<TextContent | ThinkingContent | ImageContent | ToolCall>): string {
+function assistantContentToText(content: Array<TextContent | ThinkingContent | ToolCall>): string {
 	return content
 		.map((block) => {
 			if (block.type === "text") {
@@ -156,9 +156,6 @@ function assistantContentToText(content: Array<TextContent | ThinkingContent | I
 			if (block.type === "thinking") {
 				return block.thinking;
 			}
-			if (block.type === "image") {
-				return `[image:${block.mimeType}]`;
-			}
 			return `${block.name}:${JSON.stringify(block.arguments)}`;
 		})
 		.join("\n");
@@ -365,13 +362,6 @@ async function streamWithDeltas(
 			continue;
 		}

-		if (block.type === "image") {
-			partial.content = [...partial.content, block];
-			stream.push({ type: "image_start", contentIndex: index, partial: { ...partial } });
-			stream.push({ type: "image_end", contentIndex: index, image: block, partial: { ...partial } });
-			continue;
-		}
-
 		partial.content = [...partial.content, { type: "toolCall", id: block.id, name: block.name, arguments: {} }];
 		stream.push({ type: "toolcall_start", contentIndex: index, partial: { ...partial } });
 		for (const chunk of splitStringByTokenSize(JSON.stringify(block.arguments), minTokenSize, maxTokenSize)) {
--- a/packages/ai/src/providers/google.ts
+++ b/packages/ai/src/providers/google.ts
@@ -10,7 +10,6 @@ import type {
 	Api,
 	AssistantMessage,
 	Context,
-	ImageContent,
 	Model,
 	SimpleStreamOptions,
 	StreamFunction,
@@ -154,43 +153,6 @@ export const streamGoogle: StreamFunction<"google-generative-ai", GoogleOptions>
 							}
 						}

-						if (part.inlineData) {
-							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;
-							}
-
-							if (part.inlineData.data && part.inlineData.mimeType) {
-								const imageBlock: ImageContent = {
-									type: "image",
-									data: part.inlineData.data,
-									mimeType: part.inlineData.mimeType,
-								};
-								output.content.push(imageBlock);
-								stream.push({ type: "image_start", contentIndex: blockIndex(), partial: output });
-								stream.push({
-									type: "image_end",
-									contentIndex: blockIndex(),
-									image: imageBlock,
-									partial: output,
-								});
-							}
-						}
-
 						if (part.functionCall) {
 							if (currentBlock) {
 								if (currentBlock.type === "text") {
--- a/packages/ai/src/providers/mistral.ts
+++ b/packages/ai/src/providers/mistral.ts
@@ -524,9 +524,6 @@ function toChatMessages(messages: Message[], supportsImages: boolean): ChatCompl
 					}
 					continue;
 				}
-				if (block.type !== "toolCall") {
-					continue;
-				}
 				toolCalls.push({
 					id: block.id,
 					type: "function",
--- a/packages/ai/src/providers/openai-completions.ts
+++ b/packages/ai/src/providers/openai-completions.ts
@@ -1,6 +1,5 @@
 import OpenAI from "openai";
 import type {
-	ChatCompletion,
 	ChatCompletionAssistantMessageParam,
 	ChatCompletionChunk,
 	ChatCompletionContentPart,
@@ -99,26 +98,6 @@ type ChatCompletionToolWithCacheControl = OpenAI.Chat.Completions.ChatCompletion
 	cache_control?: OpenAICompatCacheControl;
 };

-interface OpenRouterGeneratedImage {
-	image_url?: string | { url?: string };
-}
-
-type OpenRouterImageGenerationMessage = ChatCompletion["choices"][number]["message"] & {
-	images?: OpenRouterGeneratedImage[];
-};
-
-type OpenRouterImageGenerationChoice = ChatCompletion["choices"][number] & {
-	message: OpenRouterImageGenerationMessage;
-};
-
-type OpenRouterImageGenerationResponse = ChatCompletion & {
-	choices: OpenRouterImageGenerationChoice[];
-};
-
-function isOpenRouterImageGenerationModel(model: Model<"openai-completions">): boolean {
-	return model.provider === "openrouter" && model.compat?.openRouterImageGeneration === true;
-}
-
 function resolveCacheRetention(cacheRetention?: CacheRetention): CacheRetention {
 	if (cacheRetention) {
 		return cacheRetention;
@@ -171,75 +150,6 @@ export const streamOpenAICompletions: StreamFunction<"openai-completions", OpenA
 				...(options?.timeoutMs !== undefined ? { timeout: options.timeoutMs } : {}),
 				...(options?.maxRetries !== undefined ? { maxRetries: options.maxRetries } : {}),
 			};
-			if (isOpenRouterImageGenerationModel(model)) {
-				const { stream_options: _streamOptions, ...nonStreamingBaseParams } = params;
-				const nonStreamingParams = {
-					...nonStreamingBaseParams,
-					stream: false,
-				} as OpenAI.Chat.Completions.ChatCompletionCreateParamsNonStreaming;
-				const { data: response, response: rawResponse } = await client.chat.completions
-					.create(nonStreamingParams, requestOptions)
-					.withResponse();
-				await options?.onResponse?.(
-					{ status: rawResponse.status, headers: headersToRecord(rawResponse.headers) },
-					model,
-				);
-				stream.push({ type: "start", partial: output });
-
-				const imageResponse = response as OpenRouterImageGenerationResponse;
-				output.responseId ||= imageResponse.id;
-				output.usage = parseChunkUsage(imageResponse.usage ?? {}, model);
-
-				const choice = imageResponse.choices[0];
-				if (choice) {
-					const text = choice.message.content;
-					if (text) {
-						const textBlock: TextContent = { type: "text", text };
-						output.content.push(textBlock);
-						const contentIndex = output.content.length - 1;
-						stream.push({ type: "text_start", contentIndex, partial: output });
-						stream.push({ type: "text_delta", contentIndex, delta: text, partial: output });
-						stream.push({ type: "text_end", contentIndex, content: text, partial: output });
-					}
-
-					for (const image of choice.message.images ?? []) {
-						const imageUrl = typeof image.image_url === "string" ? image.image_url : image.image_url?.url;
-						if (!imageUrl?.startsWith("data:")) {
-							continue;
-						}
-						const matches = imageUrl.match(/^data:([^;]+);base64,(.+)$/);
-						if (!matches) {
-							continue;
-						}
-						const imageBlock: ImageContent = {
-							type: "image",
-							mimeType: matches[1],
-							data: matches[2],
-						};
-						output.content.push(imageBlock);
-						const contentIndex = output.content.length - 1;
-						stream.push({ type: "image_start", contentIndex, partial: output });
-						stream.push({ type: "image_end", contentIndex, image: imageBlock, partial: output });
-					}
-
-					if (choice.finish_reason) {
-						const finishReasonResult = mapStopReason(choice.finish_reason);
-						output.stopReason = finishReasonResult.stopReason;
-						if (finishReasonResult.errorMessage) {
-							output.errorMessage = finishReasonResult.errorMessage;
-						}
-					}
-				}
-
-				if (output.stopReason === "stop" || output.stopReason === "length" || output.stopReason === "toolUse") {
-					stream.push({ type: "done", reason: output.stopReason, message: output });
-				} else {
-					stream.push({ type: "error", reason: output.stopReason, error: output });
-				}
-				stream.end();
-				return;
-			}
-
 			const { data: openaiStream, response } = await client.chat.completions
 				.create(params, requestOptions)
 				.withResponse();
@@ -575,10 +485,6 @@ function buildParams(
 		prompt_cache_retention: cacheRetention === "long" && compat.supportsLongCacheRetention ? "24h" : undefined,
 	};

-	if (isOpenRouterImageGenerationModel(model)) {
-		Reflect.set(params, "modalities", ["image"]);
-	}
-
 	if (compat.supportsUsageInStreaming !== false) {
 		(params as any).stream_options = { include_usage: true };
 	}
@@ -599,12 +505,12 @@ function buildParams(
 		params.temperature = options.temperature;
 	}

-	if (!isOpenRouterImageGenerationModel(model) && context.tools && context.tools.length > 0) {
+	if (context.tools && context.tools.length > 0) {
 		params.tools = convertTools(context.tools, compat);
 		if (compat.zaiToolStream) {
 			(params as any).tool_stream = true;
 		}
-	} else if (!isOpenRouterImageGenerationModel(model) && hasToolHistory(context.messages)) {
+	} else if (hasToolHistory(context.messages)) {
 		// Anthropic (via LiteLLM/proxy) requires tools param when conversation has tool_calls/tool_results
 		params.tools = [];
 	}
@@ -1174,7 +1080,6 @@ function detectCompat(model: Model<"openai-completions">): ResolvedOpenAIComplet
 					? "openrouter"
 					: "openai",
 		openRouterRouting: {},
-		openRouterImageGeneration: false,
 		vercelGatewayRouting: {},
 		zaiToolStream: false,
 		supportsStrictMode: true,
@@ -1208,7 +1113,6 @@ function getCompat(model: Model<"openai-completions">): ResolvedOpenAICompletion
 			detected.requiresReasoningContentOnAssistantMessages,
 		thinkingFormat: model.compat.thinkingFormat ?? detected.thinkingFormat,
 		openRouterRouting: model.compat.openRouterRouting ?? {},
-		openRouterImageGeneration: model.compat.openRouterImageGeneration ?? detected.openRouterImageGeneration,
 		vercelGatewayRouting: model.compat.vercelGatewayRouting ?? detected.vercelGatewayRouting,
 		zaiToolStream: model.compat.zaiToolStream ?? detected.zaiToolStream,
 		supportsStrictMode: model.compat.supportsStrictMode ?? detected.supportsStrictMode,
--- a/packages/ai/src/types.ts
+++ b/packages/ai/src/types.ts
@@ -212,7 +212,7 @@ export interface UserMessage {

 export interface AssistantMessage {
 	role: "assistant";
-	content: (TextContent | ThinkingContent | ImageContent | ToolCall)[];
+	content: (TextContent | ThinkingContent | ToolCall)[];
 	api: Api;
 	provider: Provider;
 	model: string;
@@ -265,8 +265,6 @@ export type AssistantMessageEvent =
 	| { type: "thinking_start"; contentIndex: number; partial: AssistantMessage }
 	| { type: "thinking_delta"; contentIndex: number; delta: string; partial: AssistantMessage }
 	| { type: "thinking_end"; contentIndex: number; content: string; partial: AssistantMessage }
-	| { type: "image_start"; contentIndex: number; partial: AssistantMessage }
-	| { type: "image_end"; contentIndex: number; image: ImageContent; partial: AssistantMessage }
 	| { type: "toolcall_start"; contentIndex: number; partial: AssistantMessage }
 	| { type: "toolcall_delta"; contentIndex: number; delta: string; partial: AssistantMessage }
 	| { type: "toolcall_end"; contentIndex: number; toolCall: ToolCall; partial: AssistantMessage }
@@ -302,8 +300,6 @@ export interface OpenAICompletionsCompat {
 	thinkingFormat?: "openai" | "openrouter" | "deepseek" | "zai" | "qwen" | "qwen-chat-template";
 	/** OpenRouter-specific routing preferences. Only used when baseUrl points to OpenRouter. */
 	openRouterRouting?: OpenRouterRouting;
-	/** Whether the model uses OpenRouter's image-generation response shape with assistant `message.images`. */
-	openRouterImageGeneration?: boolean;
 	/** Vercel AI Gateway routing preferences. Only used when baseUrl points to Vercel AI Gateway. */
 	vercelGatewayRouting?: VercelGatewayRouting;
 	/** Whether z.ai supports top-level `tool_stream: true` for streaming tool call deltas. Default: false. */