@use-lattice/litmus 0.121.3

package/dist/src/responses-B2LCDCXZ.js

@@ -0,0 +1,667 @@
+import { T as state, b as getEnvInt, r as logger, x as getEnvString, y as getEnvFloat } from "./logger-C40ZGil9.js";
+import { r as importModule } from "./esm-DMVc93eh.js";
+import { i as isJavascriptFile } from "./fileExtensions-DnqA1y9x.js";
+import { d as maybeLoadToolsFromExternalFile, h as renderVarsInObject, u as maybeLoadResponseFormatFromExternalFile } from "./util-DuntT1Ga.js";
+import { h as REQUEST_TIMEOUT_MS, m as LONG_RUNNING_MODEL_TIMEOUT_MS } from "./fetch-DQckpUFz.js";
+import { r as fetchWithCache } from "./cache-BkrqU5Ba.js";
+import { t as OpenAiGenericProvider } from "./openai-DumO8WQn.js";
+import { a as calculateOpenAICost, c as getTokenUsage$1, s as formatOpenAiError } from "./util-DsS_-v4p.js";
+import path from "path";
+//#region src/providers/functionCallbackUtils.ts
+/**
+* Handles function callback execution for AI providers.
+* Provides a unified way to execute function callbacks across different provider formats.
+*/
+var FunctionCallbackHandler = class {
+	loadedCallbacks = {};
+	mcpToolNames = null;
+	constructor(mcpClient) {
+		this.mcpClient = mcpClient;
+	}
+	/**
+	* Processes a function call by executing its callback or returning the original call
+	* @param call The function call to process (can be various formats)
+	* @param callbacks Configuration mapping function names to callbacks
+	* @param context Optional context to pass to the callback
+	* @returns The result of processing
+	*/
+	async processCall(call, callbacks, context) {
+		const functionInfo = this.extractFunctionInfo(call);
+		if (this.mcpClient && functionInfo) {
+			if (this.mcpToolNames === null) {
+				const mcpTools = this.mcpClient.getAllTools();
+				this.mcpToolNames = new Set(mcpTools.map((tool) => tool.name));
+			}
+			if (this.mcpToolNames.has(functionInfo.name)) return await this.executeMcpTool(functionInfo.name, functionInfo.arguments);
+		}
+		if (!functionInfo || !callbacks || !callbacks[functionInfo.name]) return {
+			output: typeof call === "string" ? call : JSON.stringify(call),
+			isError: false
+		};
+		try {
+			return {
+				output: await this.executeCallback(functionInfo.name, functionInfo.arguments || "{}", callbacks, context),
+				isError: false
+			};
+		} catch (error) {
+			logger.debug(`Function callback failed for ${functionInfo.name}: ${error}`);
+			return {
+				output: typeof call === "string" ? call : JSON.stringify(call),
+				isError: true
+			};
+		}
+	}
+	/**
+	* Processes multiple function calls
+	* @param calls Array of calls or a single call
+	* @param callbacks Configuration mapping function names to callbacks
+	* @param context Optional context to pass to callbacks
+	* @param options Processing options
+	* @returns Processed output in appropriate format
+	*/
+	async processCalls(calls, callbacks, context, _options) {
+		if (!calls) return calls;
+		const isArray = Array.isArray(calls);
+		const callsArray = isArray ? calls : [calls];
+		const results = await Promise.all(callsArray.map((call) => this.processCall(call, callbacks, context)));
+		if (results.some((r, index) => !r.isError && r.output !== JSON.stringify(callsArray[index]))) {
+			const outputs = results.map((r) => r.output);
+			if (!isArray && outputs.length === 1) return outputs[0];
+			return outputs.every((o) => typeof o === "string") ? outputs.join("\n") : outputs;
+		}
+		if (!isArray && results.length === 1) return results[0].output;
+		return calls;
+	}
+	/**
+	* Extracts function name and arguments from various call formats
+	*/
+	extractFunctionInfo(call) {
+		if (!call || typeof call !== "object") return null;
+		if (call.name && typeof call.name === "string") return {
+			name: call.name,
+			arguments: call.arguments
+		};
+		if (call.type === "function" && call.function?.name) return {
+			name: call.function.name,
+			arguments: call.function.arguments
+		};
+		return null;
+	}
+	/**
+	* Executes a function callback
+	*/
+	async executeCallback(functionName, args, callbacks, context) {
+		let callback = this.loadedCallbacks[functionName];
+		if (!callback) {
+			const callbackConfig = callbacks[functionName];
+			if (typeof callbackConfig === "string") if (callbackConfig.startsWith("file://")) callback = await this.loadExternalFunction(callbackConfig);
+			else callback = new Function("return " + callbackConfig)();
+			else if (typeof callbackConfig === "function") callback = callbackConfig;
+			else throw new Error(`Invalid callback configuration for ${functionName}`);
+			this.loadedCallbacks[functionName] = callback;
+		}
+		const result = await callback(args, context);
+		return typeof result === "string" ? result : JSON.stringify(result);
+	}
+	/**
+	* Loads a function from an external file
+	*/
+	async loadExternalFunction(fileRef) {
+		let filePath = fileRef.slice(7);
+		let functionName;
+		if (filePath.includes(":")) {
+			const splits = filePath.split(":");
+			if (splits[0] && isJavascriptFile(splits[0])) [filePath, functionName] = splits;
+		}
+		try {
+			const resolvedPath = path.resolve(state.basePath || "", filePath);
+			logger.debug(`Loading function from ${resolvedPath}${functionName ? `:${functionName}` : ""}`);
+			const mod = await importModule(resolvedPath);
+			const func = functionName && mod[functionName] ? mod[functionName] : mod.default || mod;
+			if (typeof func !== "function") throw new Error(`Expected ${resolvedPath}${functionName ? `:${functionName}` : ""} to export a function, got ${typeof func}`);
+			return func;
+		} catch (error) {
+			throw new Error(`Failed to load function from ${fileRef}: ${error}`);
+		}
+	}
+	/**
+	* Executes an MCP tool
+	*/
+	async executeMcpTool(toolName, args) {
+		try {
+			if (!this.mcpClient) throw new Error("MCP client not available");
+			const parsedArgs = args == null || args === "" ? {} : typeof args === "string" ? JSON.parse(args) : args;
+			const result = await this.mcpClient.callTool(toolName, parsedArgs);
+			if (result?.error) return {
+				output: `MCP Tool Error (${toolName}): ${result.error}`,
+				isError: true
+			};
+			const normalizeContent = (content) => {
+				if (content == null) return "";
+				if (typeof content === "string") return content;
+				if (Array.isArray(content)) return content.map((part) => {
+					if (typeof part === "string") return part;
+					if (part && typeof part === "object") {
+						if ("text" in part && part.text != null) return String(part.text);
+						if ("json" in part) return JSON.stringify(part.json);
+						if ("data" in part) return JSON.stringify(part.data);
+						return JSON.stringify(part);
+					}
+					return String(part);
+				}).join("\n");
+				return JSON.stringify(content);
+			};
+			return {
+				output: `MCP Tool Result (${toolName}): ${normalizeContent(result?.content)}`,
+				isError: false
+			};
+		} catch (error) {
+			const errorMessage = error instanceof Error ? error.message : String(error);
+			logger.debug(`MCP tool execution failed for ${toolName}: ${errorMessage}`);
+			return {
+				output: `MCP Tool Error (${toolName}): ${errorMessage}`,
+				isError: true
+			};
+		}
+	}
+	/**
+	* Sets the MCP client, preserving any loaded callbacks
+	*/
+	setMcpClient(client) {
+		this.mcpClient = client;
+		this.mcpToolNames = null;
+	}
+	/**
+	* Clears the cached callbacks
+	*/
+	clearCache() {
+		this.loadedCallbacks = {};
+	}
+};
+//#endregion
+//#region src/providers/responses/processor.ts
+/**
+* Extract user-facing metadata from response data.
+* Only includes fields that are useful for users viewing eval results.
+*/
+function extractMetadata(data, processedOutput) {
+	const metadata = {};
+	if (typeof data.id === "string" && data.id) metadata.responseId = data.id;
+	if (typeof data.model === "string" && data.model) metadata.model = data.model;
+	if (Array.isArray(processedOutput.annotations) && processedOutput.annotations.length > 0) metadata.annotations = processedOutput.annotations;
+	return metadata;
+}
+/**
+* Extract token usage from response data, handling both OpenAI Chat Completions format
+* (prompt_tokens, completion_tokens) and Azure Responses format (input_tokens, output_tokens)
+*/
+function getTokenUsage(data, cached) {
+	if (data.usage) if (cached) {
+		const totalTokens = data.usage.total_tokens || (data.usage.input_tokens || 0) + (data.usage.output_tokens || 0);
+		return {
+			cached: totalTokens,
+			total: totalTokens,
+			numRequests: 1
+		};
+	} else {
+		const promptTokens = data.usage.prompt_tokens || data.usage.input_tokens || 0;
+		const completionTokens = data.usage.completion_tokens || data.usage.output_tokens || 0;
+		return {
+			total: data.usage.total_tokens || promptTokens + completionTokens,
+			prompt: promptTokens,
+			completion: completionTokens,
+			numRequests: 1,
+			...data.usage.completion_tokens_details ? { completionDetails: {
+				reasoning: data.usage.completion_tokens_details.reasoning_tokens,
+				acceptedPrediction: data.usage.completion_tokens_details.accepted_prediction_tokens,
+				rejectedPrediction: data.usage.completion_tokens_details.rejected_prediction_tokens
+			} } : {}
+		};
+	}
+	return {};
+}
+/**
+* Shared response processor for OpenAI and Azure Responses APIs.
+* Handles all response types with identical logic to ensure feature parity.
+*/
+var ResponsesProcessor = class {
+	constructor(config) {
+		this.config = config;
+	}
+	async processResponseOutput(data, requestConfig, cached) {
+		logger.debug(`Processing ${this.config.providerType} responses output`, {
+			responseId: data.id,
+			model: data.model
+		});
+		if (data.error) return { error: formatOpenAiError(data) };
+		try {
+			const context = {
+				config: requestConfig,
+				cached,
+				data
+			};
+			const processedOutput = await this.processOutput(data.output, context);
+			if (processedOutput.isRefusal) return {
+				output: processedOutput.refusal,
+				tokenUsage: getTokenUsage(data, cached),
+				isRefusal: true,
+				cached,
+				cost: this.config.costCalculator(this.config.modelName, data.usage, requestConfig),
+				raw: data,
+				metadata: extractMetadata(data, processedOutput)
+			};
+			let finalOutput = processedOutput.result;
+			if (requestConfig.response_format?.type === "json_schema" && typeof finalOutput === "string") try {
+				finalOutput = JSON.parse(finalOutput);
+			} catch (error) {
+				logger.error(`Failed to parse JSON output: ${error}`);
+			}
+			const result = {
+				output: finalOutput,
+				tokenUsage: getTokenUsage(data, cached),
+				cached,
+				cost: this.config.costCalculator(this.config.modelName, data.usage, requestConfig),
+				raw: data,
+				metadata: extractMetadata(data, processedOutput)
+			};
+			if (processedOutput.annotations && processedOutput.annotations.length > 0) result.raw = {
+				...data,
+				annotations: processedOutput.annotations
+			};
+			return result;
+		} catch (err) {
+			return { error: `Error parsing response: ${String(err)}\nResponse: ${JSON.stringify(data)}` };
+		}
+	}
+	async processOutput(output, context) {
+		if (this.config.modelName.includes("deep-research")) logger.debug(`Deep research response structure: ${JSON.stringify(context.data, null, 2)}`);
+		if (!output || !Array.isArray(output) || output.length === 0) throw new Error("Invalid response format: Missing output array");
+		let result = "";
+		let refusal = "";
+		let isRefusal = false;
+		const annotations = [];
+		for (const item of output) {
+			if (!item || typeof item !== "object") {
+				logger.warn(`Skipping invalid output item: ${JSON.stringify(item)}`);
+				continue;
+			}
+			const processed = await this.processOutputItem(item, context);
+			if (processed.isRefusal) {
+				refusal = processed.content || "";
+				isRefusal = true;
+			} else if (processed.content) if (result) result += "\n" + processed.content;
+			else result = processed.content;
+			if (processed.annotations) annotations.push(...processed.annotations);
+		}
+		return {
+			result,
+			refusal,
+			isRefusal,
+			annotations: annotations.length > 0 ? annotations : void 0
+		};
+	}
+	async processOutputItem(item, context) {
+		switch (item.type) {
+			case "function_call": return await this.processFunctionCall(item, context);
+			case "message": return await this.processMessage(item, context);
+			case "tool_result": return this.processToolResult(item);
+			case "reasoning": return this.processReasoning(item);
+			case "web_search_call": return this.processWebSearch(item);
+			case "code_interpreter_call": return this.processCodeInterpreter(item);
+			case "mcp_list_tools": return this.processMcpListTools(item);
+			case "mcp_call": return this.processMcpCall(item);
+			case "mcp_approval_request": return this.processMcpApprovalRequest(item);
+			default:
+				logger.debug(`Unknown output item type: ${item.type}`);
+				return {};
+		}
+	}
+	async processFunctionCall(item, context) {
+		let functionResult;
+		if (item.arguments === "{}" && item.status === "completed") functionResult = JSON.stringify({
+			type: "function_call",
+			name: item.name,
+			status: "no_arguments_provided",
+			note: "Function called but no arguments were extracted. Consider using the correct Responses API tool format."
+		});
+		else functionResult = await this.config.functionCallbackHandler.processCalls(item, context.config.functionToolCallbacks);
+		return { content: functionResult };
+	}
+	async processMessage(item, context) {
+		if (item.role !== "assistant") return {};
+		let content = "";
+		let isRefusal = false;
+		let refusal = "";
+		const annotations = [];
+		if (item.content) for (const contentItem of item.content) {
+			if (!contentItem || typeof contentItem !== "object") {
+				logger.warn(`Skipping invalid content item: ${JSON.stringify(contentItem)}`);
+				continue;
+			}
+			if (contentItem.type === "output_text") {
+				content += contentItem.text;
+				if (Array.isArray(contentItem.annotations) && contentItem.annotations.length > 0) annotations.push(...contentItem.annotations);
+			} else if (contentItem.type === "tool_use" || contentItem.type === "function_call") content = await this.config.functionCallbackHandler.processCalls(contentItem, context.config.functionToolCallbacks);
+			else if (contentItem.type === "refusal") {
+				refusal = contentItem.refusal;
+				isRefusal = true;
+			}
+		}
+		else if (item.refusal) {
+			refusal = item.refusal;
+			isRefusal = true;
+		}
+		return {
+			content: isRefusal ? refusal : content,
+			isRefusal,
+			annotations: annotations.length > 0 ? annotations : void 0
+		};
+	}
+	processToolResult(item) {
+		return Promise.resolve({ content: JSON.stringify(item) });
+	}
+	processReasoning(item) {
+		if (!item.summary || !item.summary.length) return Promise.resolve({});
+		const reasoningText = `Reasoning: ${item.summary.map((s) => s.text).join("\n")}`;
+		return Promise.resolve({ content: reasoningText });
+	}
+	processWebSearch(item) {
+		let content = "";
+		const action = item.action;
+		if (action) if (action.type === "search") content = `Web Search: "${action.query}"`;
+		else if (action.type === "open_page") content = `Opening page: ${action.url}`;
+		else if (action.type === "find_in_page") content = `Finding in page: "${action.query}"`;
+		else content = `Web action: ${action.type}`;
+		else content = `Web Search Call (status: ${item.status || "unknown"})`;
+		if (item.status === "failed" && item.error) content += ` (Error: ${item.error})`;
+		return Promise.resolve({ content });
+	}
+	processCodeInterpreter(item) {
+		let content = `Code Interpreter: ${item.code || "Running code..."}`;
+		if (item.status === "failed" && item.error) content += ` (Error: ${item.error})`;
+		return Promise.resolve({ content });
+	}
+	processMcpListTools(item) {
+		const content = `MCP Tools from ${item.server_label}: ${JSON.stringify(item.tools, null, 2)}`;
+		return Promise.resolve({ content });
+	}
+	processMcpCall(item) {
+		let content;
+		if (item.error) content = `MCP Tool Error (${item.name}): ${item.error}`;
+		else content = `MCP Tool Result (${item.name}): ${item.output}`;
+		return Promise.resolve({ content });
+	}
+	processMcpApprovalRequest(item) {
+		const content = `MCP Approval Required for ${item.server_label}.${item.name}: ${item.arguments}`;
+		return Promise.resolve({ content });
+	}
+};
+//#endregion
+//#region src/providers/openai/responses.ts
+var OpenAiResponsesProvider = class extends OpenAiGenericProvider {
+	functionCallbackHandler = new FunctionCallbackHandler();
+	processor;
+	static OPENAI_RESPONSES_MODEL_NAMES = [
+		"gpt-4o",
+		"gpt-4o-2024-08-06",
+		"gpt-4o-2024-11-20",
+		"gpt-4o-2024-05-13",
+		"gpt-4o-2024-07-18",
+		"gpt-4o-mini",
+		"gpt-4o-mini-2024-07-18",
+		"gpt-4.1",
+		"gpt-4.1-2025-04-14",
+		"gpt-4.1-mini",
+		"gpt-4.1-mini-2025-04-14",
+		"gpt-4.1-nano",
+		"gpt-4.1-nano-2025-04-14",
+		"gpt-5",
+		"gpt-5-2025-08-07",
+		"gpt-5-chat",
+		"gpt-5-chat-latest",
+		"gpt-5-nano",
+		"gpt-5-nano-2025-08-07",
+		"gpt-5-mini",
+		"gpt-5-mini-2025-08-07",
+		"gpt-5-pro",
+		"gpt-5-pro-2025-10-06",
+		"gpt-5.1",
+		"gpt-5.1-2025-11-13",
+		"gpt-5.1-mini",
+		"gpt-5.1-nano",
+		"gpt-5.1-codex",
+		"gpt-5.1-codex-max",
+		"gpt-5.1-chat-latest",
+		"gpt-5.2",
+		"gpt-5.2-2025-12-11",
+		"gpt-5.2-chat-latest",
+		"gpt-5.2-codex",
+		"gpt-5.2-pro",
+		"gpt-5.2-pro-2025-12-11",
+		"gpt-5.3-chat-latest",
+		"gpt-5.3-codex",
+		"gpt-5.3-codex-spark",
+		"gpt-5.4",
+		"gpt-5.4-2026-03-05",
+		"gpt-5.4-mini",
+		"gpt-5.4-mini-2026-03-17",
+		"gpt-5.4-nano",
+		"gpt-5.4-nano-2026-03-17",
+		"gpt-5.4-pro",
+		"gpt-5.4-pro-2026-03-05",
+		"gpt-audio",
+		"gpt-audio-2025-08-28",
+		"gpt-audio-mini",
+		"gpt-audio-mini-2025-10-06",
+		"computer-use-preview",
+		"computer-use-preview-2025-03-11",
+		"o1",
+		"o1-2024-12-17",
+		"o1-preview",
+		"o1-preview-2024-09-12",
+		"o1-mini",
+		"o1-mini-2024-09-12",
+		"o1-pro",
+		"o1-pro-2025-03-19",
+		"o3-pro",
+		"o3-pro-2025-06-10",
+		"o3",
+		"o3-2025-04-16",
+		"o4-mini",
+		"o4-mini-2025-04-16",
+		"o3-mini",
+		"o3-mini-2025-01-31",
+		"codex-mini-latest",
+		"gpt-5-codex",
+		"o3-deep-research",
+		"o3-deep-research-2025-06-26",
+		"o4-mini-deep-research",
+		"o4-mini-deep-research-2025-06-26"
+	];
+	config;
+	constructor(modelName, options = {}) {
+		super(modelName, options);
+		this.config = options.config || {};
+		this.processor = new ResponsesProcessor({
+			modelName: this.modelName,
+			providerType: "openai",
+			functionCallbackHandler: this.functionCallbackHandler,
+			costCalculator: (modelName, usage, config) => calculateOpenAICost(modelName, config, usage?.input_tokens, usage?.output_tokens, 0, 0) ?? 0
+		});
+	}
+	isGPT5Model() {
+		return this.modelName.startsWith("gpt-5") || this.modelName.includes("/gpt-5");
+	}
+	isReasoningModel() {
+		return this.modelName.startsWith("o1") || this.modelName.startsWith("o3") || this.modelName.startsWith("o4") || this.modelName.includes("/o1") || this.modelName.includes("/o3") || this.modelName.includes("/o4") || this.modelName === "codex-mini-latest" || this.isGPT5Model();
+	}
+	supportsTemperature() {
+		return !this.isReasoningModel();
+	}
+	async getOpenAiBody(prompt, context, _callApiOptions) {
+		const config = {
+			...this.config,
+			...context?.prompt?.config
+		};
+		let input;
+		try {
+			const parsedJson = JSON.parse(prompt);
+			if (Array.isArray(parsedJson)) input = parsedJson;
+			else input = prompt;
+		} catch {
+			input = prompt;
+		}
+		const isReasoningModel = this.isReasoningModel();
+		const maxOutputTokensDefault = config.omitDefaults ? getEnvString("OPENAI_MAX_TOKENS") === void 0 ? void 0 : getEnvInt("OPENAI_MAX_TOKENS") : getEnvInt("OPENAI_MAX_TOKENS", 1024);
+		const reasoningMaxOutputTokensDefault = getEnvInt("OPENAI_MAX_COMPLETION_TOKENS") ?? getEnvInt("OPENAI_MAX_TOKENS");
+		const maxOutputTokens = config.max_output_tokens ?? (isReasoningModel ? reasoningMaxOutputTokensDefault : maxOutputTokensDefault);
+		const temperatureDefault = config.omitDefaults ? getEnvString("OPENAI_TEMPERATURE") === void 0 ? void 0 : getEnvFloat("OPENAI_TEMPERATURE") : getEnvFloat("OPENAI_TEMPERATURE", 0);
+		const temperature = this.supportsTemperature() ? config.temperature ?? temperatureDefault : void 0;
+		const reasoningEffort = isReasoningModel ? renderVarsInObject(config.reasoning_effort, context?.vars) : void 0;
+		const instructions = config.instructions;
+		const responseFormat = maybeLoadResponseFormatFromExternalFile(config.response_format, context?.vars);
+		let textFormat;
+		if (responseFormat) if (responseFormat.type === "json_object") textFormat = { format: { type: "json_object" } };
+		else if (responseFormat.type === "json_schema") {
+			const schema = responseFormat.schema || responseFormat.json_schema?.schema;
+			textFormat = { format: {
+				type: "json_schema",
+				name: responseFormat.json_schema?.name || responseFormat.name || "response_schema",
+				schema,
+				strict: true
+			} };
+		} else textFormat = { format: { type: "text" } };
+		else textFormat = { format: { type: "text" } };
+		if (this.isGPT5Model() && config.verbosity) textFormat = {
+			...textFormat,
+			verbosity: config.verbosity
+		};
+		const loadedTools = config.tools ? await maybeLoadToolsFromExternalFile(config.tools, context?.vars) : void 0;
+		const body = {
+			model: this.modelName,
+			input,
+			...maxOutputTokens === void 0 ? {} : { max_output_tokens: maxOutputTokens },
+			...reasoningEffort ? { reasoning: { effort: reasoningEffort } } : {},
+			...temperature === void 0 ? {} : { temperature },
+			...instructions ? { instructions } : {},
+			...(!reasoningEffort || reasoningEffort === "none") && (config.top_p !== void 0 || getEnvString("OPENAI_TOP_P")) ? { top_p: config.top_p ?? getEnvFloat("OPENAI_TOP_P", 1) } : {},
+			...loadedTools ? { tools: loadedTools } : {},
+			...config.tool_choice ? { tool_choice: config.tool_choice } : {},
+			...config.max_tool_calls ? { max_tool_calls: config.max_tool_calls } : {},
+			...config.previous_response_id ? { previous_response_id: config.previous_response_id } : {},
+			text: textFormat,
+			...config.truncation ? { truncation: config.truncation } : {},
+			...config.metadata ? { metadata: config.metadata } : {},
+			..."parallel_tool_calls" in config ? { parallel_tool_calls: Boolean(config.parallel_tool_calls) } : {},
+			...config.stream ? { stream: config.stream } : {},
+			..."store" in config ? { store: Boolean(config.store) } : {},
+			...config.background ? { background: config.background } : {},
+			...config.webhook_url ? { webhook_url: config.webhook_url } : {},
+			...config.user ? { user: config.user } : {},
+			...config.passthrough || {}
+		};
+		if (config.reasoning && this.isReasoningModel()) body.reasoning = config.reasoning;
+		return {
+			body,
+			config: {
+				...config,
+				tools: loadedTools,
+				response_format: responseFormat
+			}
+		};
+	}
+	async callApi(prompt, context, callApiOptions) {
+		if (this.requiresApiKey() && !this.getApiKey()) throw new Error(this.getMissingApiKeyErrorMessage());
+		const { body, config } = await this.getOpenAiBody(prompt, context, callApiOptions);
+		const isDeepResearchModel = this.modelName.includes("deep-research");
+		if (isDeepResearchModel) {
+			if (!config.tools?.some((tool) => tool.type === "web_search_preview")) return { error: `Deep research model ${this.modelName} requires the web_search_preview tool to be configured. Add it to your provider config:\ntools:\n  - type: web_search_preview` };
+			const mcpTools = config.tools?.filter((tool) => tool.type === "mcp") || [];
+			for (const mcpTool of mcpTools) if (mcpTool.require_approval !== "never") return { error: `Deep research model ${this.modelName} requires MCP tools to have require_approval: 'never'. Update your MCP tool configuration:\ntools:\n  - type: mcp\n    require_approval: never` };
+		}
+		let timeout = REQUEST_TIMEOUT_MS;
+		const isGpt5ProModel = /(^|\/)gpt-5(?:\.\d+)?-pro(?:-|$)/.test(this.modelName);
+		if (isDeepResearchModel || isGpt5ProModel) {
+			const evalTimeout = getEnvInt("PROMPTFOO_EVAL_TIMEOUT_MS", 0);
+			timeout = evalTimeout > 0 ? evalTimeout : LONG_RUNNING_MODEL_TIMEOUT_MS;
+			logger.debug(`Using timeout of ${timeout}ms for long-running model ${this.modelName}`);
+		}
+		let data;
+		let status;
+		let statusText;
+		let cached = false;
+		let deleteFromCache;
+		let responseHeaders;
+		try {
+			({data, cached, status, statusText, deleteFromCache, headers: responseHeaders} = await fetchWithCache(`${this.getApiUrl()}/responses`, {
+				method: "POST",
+				headers: {
+					"Content-Type": "application/json",
+					...this.getApiKey() ? { Authorization: `Bearer ${this.getApiKey()}` } : {},
+					...this.getOrganization() ? { "OpenAI-Organization": this.getOrganization() } : {},
+					...config.headers
+				},
+				body: JSON.stringify(body)
+			}, timeout, "json", context?.bustCache ?? context?.debug, this.config.maxRetries));
+			if (status < 200 || status >= 300) {
+				const errorMessage = `API error: ${status} ${statusText}\n${typeof data === "string" ? data : JSON.stringify(data)}`;
+				if (typeof data === "object" && data?.error?.code === "invalid_prompt") return {
+					output: errorMessage,
+					tokenUsage: data?.usage ? getTokenUsage$1(data, cached) : void 0,
+					isRefusal: true,
+					metadata: { http: {
+						status,
+						statusText,
+						headers: responseHeaders ?? {}
+					} }
+				};
+				return {
+					error: errorMessage,
+					metadata: { http: {
+						status,
+						statusText,
+						headers: responseHeaders ?? {}
+					} }
+				};
+			}
+		} catch (err) {
+			logger.error(`API call error: ${String(err)}`);
+			await deleteFromCache?.();
+			return {
+				error: `API call error: ${String(err)}`,
+				metadata: { http: {
+					status: 0,
+					statusText: "Error",
+					headers: responseHeaders ?? {}
+				} }
+			};
+		}
+		if (data.error?.message) {
+			await deleteFromCache?.();
+			return {
+				error: formatOpenAiError(data),
+				metadata: { http: {
+					status,
+					statusText,
+					headers: responseHeaders ?? {}
+				} }
+			};
+		}
+		const result = await this.processor.processResponseOutput(data, config, cached);
+		return {
+			...result,
+			metadata: {
+				...result.metadata,
+				http: {
+					status,
+					statusText,
+					headers: responseHeaders ?? {}
+				}
+			}
+		};
+	}
+};
+//#endregion
+export { ResponsesProcessor as n, FunctionCallbackHandler as r, OpenAiResponsesProvider as t };
+
+//# sourceMappingURL=responses-B2LCDCXZ.js.map