@tanstack/openai-base 0.10.16 → 0.12.1

package/dist/esm/adapters/responses-text.js

@@ -1,9 +1,10 @@
-import { makeStructuredOutputCompatibleWithMap } from "../utils/schema-converter.js";
+import { makeStructuredOutputCompatibleWithMap, warnStrictFallback } from "../utils/schema-converter.js";
 import { buildResponsesUsage } from "../usage.js";
 import { extractRequestOptions } from "../utils/request-options.js";
 import { createToolInputNormalizer } from "../utils/tool-input-normalizer.js";
 import { convertToolsToResponsesFormat } from "./responses-tool-converter.js";
-import { EventType, normalizeSystemPrompts } from "@tanstack/ai";
+import { hostedShellCallIds, readUserExecutedCall, userToolRequestItem, userToolResultItem } from "./responses-user-tools.js";
+import { EventType, fileReferenceFor, isFileSource, normalizeSystemPrompts, unsupportedFileSourceError } from "@tanstack/ai";
 import { toRunErrorPayload, toRunErrorRawEvent } from "@tanstack/ai/adapter-internals";
 import { BaseTextAdapter } from "@tanstack/ai/adapters";
 import { generateId } from "@tanstack/ai-utils";
@@ -12,6 +13,40 @@ var PDF_BASE64_MAGIC = "JVBERi";
 function isRecord(value) {
 	return typeof value === "object" && value !== null;
 }
+function readURLCitation(value) {
+	if (!isRecord(value) || value.type !== "url_citation") return void 0;
+	if (typeof value.url !== "string" || typeof value.title !== "string" || typeof value.start_index !== "number" || typeof value.end_index !== "number") return;
+	return {
+		type: "url_citation",
+		url: value.url,
+		title: value.title,
+		start_index: value.start_index,
+		end_index: value.end_index
+	};
+}
+function readWebSearchCall(value) {
+	if (!isRecord(value) || value.type !== "web_search_call") return void 0;
+	if (typeof value.id !== "string" || typeof value.status !== "string" || !isRecord(value.action) || typeof value.action.type !== "string") return;
+	return value;
+}
+function collectWebSearchSources(item, citations) {
+	const sources = /* @__PURE__ */ new Map();
+	const add = (url, title) => {
+		if (typeof url !== "string" || url.length === 0) return;
+		const existing = sources.get(url);
+		if (existing) {
+			if (!existing.title && typeof title === "string" && title.length > 0) existing.title = title;
+			return;
+		}
+		sources.set(url, {
+			url,
+			...typeof title === "string" && title.length > 0 ? { title } : {}
+		});
+	};
+	if (item.action.type === "search") for (const source of item.action.sources ?? []) add(source.url);
+	for (const citation of citations) add(citation.url, citation.title);
+	return [...sources.values()];
+}
 function packResponsesReasoningSignature(id, encryptedContent) {
 	if (!id && !encryptedContent) return void 0;
 	return JSON.stringify({
@@ -54,10 +89,13 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 	kind = "text";
 	name;
 	client;
-	constructor(model, name, client) {
+	/** See {@link OpenAIBaseTextAdapterOptions.strictFallbackWarning}. */
+	strictFallbackWarning;
+	constructor(model, name, client, options = {}) {
 		super({}, model);
 		this.name = name;
 		this.client = client;
+		this.strictFallbackWarning = options.strictFallbackWarning ?? true;
 	}
 	async *chatStream(options) {
 		const toolCallMetadata = /* @__PURE__ */ new Map();
@@ -197,6 +235,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 		let hasClosedReasoning = false;
 		let model = chatOptions.model;
 		let usage;
+		let responseCompleted = false;
 		const closeReasoning = function* () {
 			if (reasoningMessageId && !hasClosedReasoning) {
 				hasClosedReasoning = true;
@@ -347,10 +386,11 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 					continue;
 				}
 				if (chunk.type === "response.completed") {
+					responseCompleted = true;
 					const response = chunk.response;
 					if (response.usage) usage = response.usage;
 					if (response.model) model = response.model;
-					continue;
+					break;
 				}
 				if (chunk.type === "response.failed") {
 					const response = chunk.response;
@@ -378,6 +418,22 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				model,
 				timestamp: Date.now()
 			};
+			if (!responseCompleted) {
+				const message = "Response stream ended before response.completed";
+				yield {
+					type: EventType.RUN_ERROR,
+					runId: aguiState.runId,
+					model,
+					timestamp: Date.now(),
+					message,
+					code: "incomplete-stream",
+					error: {
+						message,
+						code: "incomplete-stream"
+					}
+				};
+				return;
+			}
 			if (accumulatedContent.length === 0) {
 				yield {
 					type: EventType.RUN_ERROR,
@@ -566,8 +622,61 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 		let closedReasoningStepId;
 		let hasClosedReasoning = false;
 		let runFinishedEmitted = false;
+		const providerWebSearchCalls = /* @__PURE__ */ new Map();
+		const webSearchCitations = [];
 		const adapterName = this.name;
 		const emitModel = () => model || options.model;
+		const recordProviderWebSearchCall = (value, index) => {
+			const item = readWebSearchCall(value);
+			if (!item) return;
+			const existing = providerWebSearchCalls.get(item.id);
+			if (existing) {
+				existing.item = item;
+				existing.index = index;
+			} else providerWebSearchCalls.set(item.id, {
+				item,
+				index,
+				started: false
+			});
+		};
+		const emitProviderWebSearchCalls = function* (assistantMessage, completedOnly = false) {
+			for (const entry of providerWebSearchCalls.values()) {
+				if (entry.started || completedOnly && entry.item.status !== "completed") continue;
+				const callUrls = new Set(entry.item.action.type === "search" ? (entry.item.action.sources ?? []).map((source) => source.url) : []);
+				const citations = webSearchCitations.filter((citation) => callUrls.has(citation.url));
+				const metadata = {
+					itemId: entry.item.id,
+					providerExecuted: true,
+					sources: collectWebSearchSources(entry.item, citations),
+					openai: {
+						webSearchCall: entry.item,
+						urlCitations: citations,
+						...assistantMessage ? { assistantMessage } : {}
+					}
+				};
+				entry.started = true;
+				yield {
+					type: EventType.TOOL_CALL_START,
+					toolCallId: entry.item.id,
+					toolCallName: "web_search",
+					toolName: "web_search",
+					parentMessageId: aguiState.messageId,
+					model: emitModel(),
+					timestamp: Date.now(),
+					index: entry.index,
+					metadata
+				};
+				yield {
+					type: EventType.TOOL_CALL_END,
+					toolCallId: entry.item.id,
+					toolCallName: "web_search",
+					toolName: "web_search",
+					model: emitModel(),
+					timestamp: Date.now(),
+					input: entry.item.action
+				};
+			}
+		};
 		const openReasoning = function* () {
 			if (reasoningMessageId) return;
 			reasoningMessageId = generateId(adapterName);
@@ -639,6 +748,46 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 			hasClosedReasoning = false;
 			accumulatedReasoning = "";
 		};
+		const userToolChunks = (item, outputIndex, bareShell) => {
+			const call = readUserExecutedCall(item, { bareShell });
+			if (!call || toolCallMetadata.get(call.callId)?.ended) return [];
+			const tracked = toolCallMetadata.get(call.callId) ?? {
+				callId: call.callId,
+				index: outputIndex,
+				name: call.name,
+				started: false
+			};
+			toolCallMetadata.set(call.callId, tracked);
+			tracked.started = true;
+			tracked.ended = true;
+			tracked.name = call.name;
+			const timestamp = Date.now();
+			const modelName = model || options.model;
+			const metadata = {
+				openaiUserTool: call.name,
+				...call.itemId ? { itemId: call.itemId } : {},
+				...call.maxOutputLength !== void 0 ? { maxOutputLength: call.maxOutputLength } : {}
+			};
+			return [{
+				type: EventType.TOOL_CALL_START,
+				toolCallId: call.callId,
+				toolCallName: call.name,
+				toolName: call.name,
+				parentMessageId: aguiState.messageId,
+				model: modelName,
+				timestamp,
+				index: outputIndex,
+				metadata
+			}, {
+				type: EventType.TOOL_CALL_END,
+				toolCallId: call.callId,
+				toolCallName: call.name,
+				toolName: call.name,
+				model: modelName,
+				timestamp,
+				input: call.input
+			}];
+		};
 		const emitReasoningDelta = function* (text) {
 			if (!text) return;
 			yield* openReasoning();
@@ -709,6 +858,10 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 					accumulatedReasoning = "";
 				}
 				if (chunk.type === "response.failed" || chunk.type === "response.incomplete") {
+					if (chunk.type === "response.incomplete") {
+						for (const [index, item] of (chunk.response.output ?? []).entries()) recordProviderWebSearchCall(item, index);
+						yield* emitProviderWebSearchCalls(void 0, true);
+					}
 					yield* closeReasoning();
 					if (hasEmittedTextMessageStart) {
 						yield {
@@ -817,6 +970,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				}
 				if (chunk.type === "response.output_item.added") {
 					const item = chunk.item;
+					if (item.type === "web_search_call") recordProviderWebSearchCall(item, chunk.output_index);
 					if (item.type === "reasoning") {
 						captureReasoningItem(item);
 						yield* openReasoning();
@@ -850,6 +1004,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							metadata.started = true;
 						}
 					}
+					yield* userToolChunks(item, chunk.output_index, false);
 				}
 				if (chunk.type === "response.function_call_arguments.delta" && chunk.delta) {
 					const metadata = toolCallMetadata.get(chunk.item_id);
@@ -912,6 +1067,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				}
 				if (chunk.type === "response.output_item.done") {
 					const item = chunk.item;
+					if (item.type === "web_search_call") recordProviderWebSearchCall(item, chunk.output_index);
 					if (item.type === "reasoning") {
 						captureReasoningItem(item);
 						yield* openReasoning();
@@ -973,9 +1129,17 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							metadata.pendingArguments = void 0;
 						}
 					}
+					yield* userToolChunks(item, chunk.output_index, false);
+				}
+				if (chunk.type === "response.output_text.annotation.added") {
+					const citation = readURLCitation(chunk.annotation);
+					if (citation) webSearchCitations.push(citation);
 				}
 				if (chunk.type === "response.completed") {
-					const completedText = chunk.response.output.flatMap((item) => item.type === "message" ? item.content : []).filter((part) => part.type === "output_text").map((part) => part.text).join("");
+					const responseOutput = Array.isArray(chunk.response.output) ? chunk.response.output : [];
+					const assistantMessage = responseOutput.find((item) => item.type === "message");
+					for (const [index, item] of responseOutput.entries()) if (item.type === "web_search_call") recordProviderWebSearchCall(item, index);
+					const completedText = responseOutput.flatMap((item) => item.type === "message" ? item.content : []).filter((part) => part.type === "output_text").map((part) => part.text).join("");
 					if (accumulatedContent.length === 0 && completedText.length > 0) {
 						if (!hasEmittedTextMessageStart) {
 							hasEmittedTextMessageStart = true;
@@ -998,7 +1162,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							content: accumulatedContent
 						};
 					}
-					if (Array.isArray(chunk.response.output)) for (const item of chunk.response.output) captureReasoningItem(item);
+					for (const item of responseOutput) captureReasoningItem(item);
 					if (!reasoningMessageId && (reasoningItemId || reasoningEncryptedContent)) {
 						const signature = packResponsesReasoningSignature(reasoningItemId, reasoningEncryptedContent);
 						if (closedReasoningStepId && signature) yield {
@@ -1012,11 +1176,11 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						};
 						else if (!closedReasoningStepId) yield* openReasoning();
 					}
-					for (const item of chunk.response.output) {
+					for (const [outputIndex, item] of responseOutput.entries()) {
 						if (item.type !== "function_call" || !item.id) continue;
 						const metadata = toolCallMetadata.get(item.id) ?? {
 							callId: item.call_id || item.id,
-							index: 0,
+							index: outputIndex,
 							name: item.name || "",
 							started: false
 						};
@@ -1070,6 +1234,12 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							metadata.pendingArguments = void 0;
 						}
 					}
+					yield* emitProviderWebSearchCalls(assistantMessage);
+					const shellOutputs = hostedShellCallIds(responseOutput);
+					for (const [outputIndex, item] of responseOutput.entries()) {
+						if (isRecord(item) && item.type === "shell_call" && typeof item.call_id === "string" && shellOutputs.has(item.call_id)) continue;
+						yield* userToolChunks(item, outputIndex, true);
+					}
 					yield* closeReasoning();
 					if (hasEmittedTextMessageStart) {
 						yield {
@@ -1080,7 +1250,12 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						};
 						hasEmittedTextMessageStart = false;
 					}
-					const hasFunctionCalls = chunk.response.output.some((item) => item.type === "function_call");
+					const hasFunctionCalls = responseOutput.some((item) => {
+						if (!isRecord(item)) return false;
+						if (item.type === "function_call") return true;
+						if (item.type === "shell_call" && typeof item.call_id === "string" && shellOutputs.has(item.call_id)) return false;
+						return readUserExecutedCall(item, { bareShell: true }) !== null;
+					});
 					const incompleteReason = chunk.response.incomplete_details?.reason;
 					const finishReason = hasFunctionCalls ? "tool_calls" : incompleteReason === "max_output_tokens" ? "length" : incompleteReason === "content_filter" ? "content_filter" : "stop";
 					yield {
@@ -1093,6 +1268,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						finishReason
 					};
 					runFinishedEmitted = true;
+					return;
 				}
 				if (chunk.type === "error") {
 					const code = chunk.code ?? void 0;
@@ -1112,6 +1288,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				}
 			}
 			if (!runFinishedEmitted && aguiState.hasEmittedRunStarted) {
+				yield* emitProviderWebSearchCalls(void 0, true);
 				yield* closeReasoning();
 				if (hasEmittedTextMessageStart) yield {
 					type: EventType.TEXT_MESSAGE_END,
@@ -1119,13 +1296,17 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 					model: model || options.model,
 					timestamp: Date.now()
 				};
+				const message = "Response stream ended before response.completed";
 				yield {
-					type: EventType.RUN_FINISHED,
-					runId: aguiState.runId,
-					threadId: aguiState.threadId,
+					type: EventType.RUN_ERROR,
 					model: model || options.model,
 					timestamp: Date.now(),
-					finishReason: toolCallMetadata.size > 0 ? "tool_calls" : "stop"
+					message,
+					code: "incomplete-stream",
+					error: {
+						message,
+						code: "incomplete-stream"
+					}
 				};
 			}
 		} catch (error) {
@@ -1155,6 +1336,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 	*/
 	mapOptionsToRequest(options) {
 		const input = this.convertMessagesToInput(options.messages);
+		if (this.strictFallbackWarning) warnStrictFallback(options.tools, options.logger);
 		const tools = options.tools ? convertToolsToResponsesFormat(options.tools, this.makeStructuredOutputCompatible.bind(this)) : void 0;
 		const modelOptions = options.modelOptions;
 		const combinedSchema = options.outputSchema;
@@ -1198,8 +1380,16 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 	*/
 	convertMessagesToInput(messages) {
 		const result = [];
+		const seenReasoningIds = /* @__PURE__ */ new Set();
+		const userToolCalls = /* @__PURE__ */ new Map();
 		for (const message of messages) {
 			if (message.role === "tool") {
+				const owner = message.toolCallId ? userToolCalls.get(message.toolCallId) : void 0;
+				const userOutput = owner ? userToolResultItem(owner, message.content) : null;
+				if (userOutput) {
+					result.push(userOutput);
+					continue;
+				}
 				const toolContent = message.content;
 				const output = Array.isArray(toolContent) ? toolContent.map((part) => this.convertContentPartToInput(part)) : typeof toolContent === "string" ? toolContent : JSON.stringify(toolContent);
 				result.push({
@@ -1210,10 +1400,16 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				continue;
 			}
 			if (message.role === "assistant") {
+				let reasoningCandidates = 0;
+				let emittedReasoning = 0;
 				if (message.thinking) for (const thinking of message.thinking) {
 					if (!thinking.signature) continue;
 					const packed = unpackResponsesReasoningSignature(thinking.signature);
 					if (!packed?.id) continue;
+					reasoningCandidates++;
+					if (seenReasoningIds.has(packed.id)) continue;
+					seenReasoningIds.add(packed.id);
+					emittedReasoning++;
 					result.push({
 						type: "reasoning",
 						id: packed.id,
@@ -1224,18 +1420,44 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						}] : []
 					});
 				}
+				const canPairReasoning = reasoningCandidates === 0 || reasoningCandidates === 1 && emittedReasoning === 1;
+				let rawAssistantMessage;
 				if (message.toolCalls && message.toolCalls.length > 0) for (const toolCall of message.toolCalls) {
+					const metadata = toolCall.metadata;
+					if (metadata?.providerExecuted) {
+						const webSearchCall = metadata.openai?.webSearchCall;
+						if (webSearchCall && canPairReasoning) {
+							result.push(webSearchCall);
+							rawAssistantMessage ??= metadata.openai?.assistantMessage;
+						}
+						continue;
+					}
 					const argumentsString = typeof toolCall.function.arguments === "string" ? toolCall.function.arguments : JSON.stringify(toolCall.function.arguments);
-					const itemId = toolCall.metadata?.itemId;
+					const replayCall = {
+						id: toolCall.id,
+						function: {
+							name: toolCall.function.name,
+							arguments: argumentsString
+						},
+						...toolCall.metadata !== void 0 ? { metadata: toolCall.metadata } : {}
+					};
+					const userItem = userToolRequestItem(replayCall);
+					if (userItem) {
+						userToolCalls.set(toolCall.id, replayCall);
+						result.push(userItem);
+						continue;
+					}
+					const itemId = metadata?.itemId;
 					result.push({
 						type: "function_call",
 						call_id: toolCall.id,
-						...itemId && { id: itemId },
+						...itemId && canPairReasoning && { id: itemId },
 						name: toolCall.function.name,
 						arguments: argumentsString
 					});
 				}
-				if (message.content) {
+				if (rawAssistantMessage) result.push(rawAssistantMessage);
+				else if (message.content) {
 					const contentStr = this.extractTextContent(message.content);
 					if (contentStr) result.push({
 						type: "message",
@@ -1270,6 +1492,14 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 			};
 			case "image": {
 				const imageMetadata = part.metadata;
+				if (isFileSource(part.source)) {
+					if (this.supportsFileSources !== true) throw unsupportedFileSourceError(this.name);
+					return {
+						type: "input_image",
+						file_id: fileReferenceFor(part.source, this.name),
+						detail: imageMetadata?.detail || "auto"
+					};
+				}
 				if (part.source.type === "url") return {
 					type: "input_image",
 					image_url: part.source.value,
@@ -1284,6 +1514,13 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 				};
 			}
 			case "audio": {
+				if (isFileSource(part.source)) {
+					if (this.supportsFileSources !== true) throw unsupportedFileSourceError(this.name);
+					return {
+						type: "input_file",
+						file_id: fileReferenceFor(part.source, this.name)
+					};
+				}
 				if (part.source.type === "url") return {
 					type: "input_file",
 					file_url: part.source.value
@@ -1298,6 +1535,14 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 			case "document": {
 				const documentMetadata = part.metadata;
 				const documentDetail = documentMetadata?.detail !== void 0 ? { detail: documentMetadata.detail } : {};
+				if (isFileSource(part.source)) {
+					if (this.supportsFileSources !== true) throw unsupportedFileSourceError(this.name);
+					return {
+						type: "input_file",
+						file_id: fileReferenceFor(part.source, this.name),
+						...documentDetail
+					};
+				}
 				if (part.source.type === "url") return {
 					type: "input_file",
 					file_url: part.source.value,