@tanstack/openai-base 0.10.16 → 0.11.1

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

@@ -3,7 +3,8 @@ 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";
@@ -639,6 +640,46 @@ 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();
@@ -850,6 +891,7 @@ 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);
@@ -973,6 +1015,7 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							metadata.pendingArguments = void 0;
 						}
 					}
+					yield* userToolChunks(item, chunk.output_index, false);
 				}
 				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("");
@@ -1012,11 +1055,11 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						};
 						else if (!closedReasoningStepId) yield* openReasoning();
 					}
-					for (const item of chunk.response.output) {
+					for (const [outputIndex, item] of chunk.response.output.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 +1113,11 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 							metadata.pendingArguments = void 0;
 						}
 					}
+					const shellOutputs = hostedShellCallIds(chunk.response.output);
+					for (const [outputIndex, item] of chunk.response.output.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 +1128,12 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						};
 						hasEmittedTextMessageStart = false;
 					}
-					const hasFunctionCalls = chunk.response.output.some((item) => item.type === "function_call");
+					const hasFunctionCalls = chunk.response.output.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 {
@@ -1198,8 +1251,16 @@ 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 +1271,16 @@ 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,
@@ -1224,13 +1291,28 @@ var OpenAIBaseResponsesTextAdapter = class extends BaseTextAdapter {
 						}] : []
 					});
 				}
+				const canPairReasoning = reasoningCandidates === 0 || reasoningCandidates === 1 && emittedReasoning === 1;
 				if (message.toolCalls && message.toolCalls.length > 0) for (const toolCall of message.toolCalls) {
 					const argumentsString = typeof toolCall.function.arguments === "string" ? toolCall.function.arguments : JSON.stringify(toolCall.function.arguments);
+					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 = toolCall.metadata?.itemId;
 					result.push({
 						type: "function_call",
 						call_id: toolCall.id,
-						...itemId && { id: itemId },
+						...itemId && canPairReasoning && { id: itemId },
 						name: toolCall.function.name,
 						arguments: argumentsString
 					});
@@ -1270,6 +1352,14 @@ 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 +1374,13 @@ 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 +1395,14 @@ 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,