Agent Workflow Platform

\n Select Workflow:\n \n \n\n \n\n

Workflow Result

— Waiting for execution —

Execution Log

— Waiting for execution —

\n \n\n \n \n\n`;\n}\n\n// Main HTTP request handler\nexport default async function(req: Request): Promise {\n globalLogger.clear(); // Clear logs for each new request\n const requestStartTime = Date.now();\n // Using import.meta.url to get the current val's URL for the view source link\n const currentValUrl = import.meta.url;\n const requestLog = globalLogger.createLogFunction(`REQ-${requestStartTime}`, \"RequestHandler\");\n\n requestLog(\"INFO\", \"RequestReceived\", `${req.method} ${req.url}`);\n\n try {\n if (req.method === \"GET\") {\n requestLog(\"DEBUG\", \"ServeHTML\", \"Serving HTML UI with workflow selector\");\n return new Response(generateHtmlWithWorkflowSelector(availableWorkflows, currentValUrl), { // Pass workflows map\n status: 200,\n headers: { \"Content-Type\": \"text/html\" },\n });\n } else if (req.method === \"POST\") {\n let requestBody: { workflowId?: string; payload?: any };\n try {\n requestBody = await req.json();\n requestLog(\"DEBUG\", \"ParsePayload\", \"Parsed JSON payload\", { body: requestBody });\n } catch (e: any) {\n requestLog(\"ERROR\", \"ParsePayload\", \"Failed to parse request JSON\", e);\n return new Response(\n JSON.stringify({\n success: false,\n error: \"Invalid JSON payload\",\n details: e.message,\n debugLog: globalLogger.getLogs(),\n }),\n { status: 400, headers: { \"Content-Type\": \"application/json\" } },\n );\n }\n\n const { workflowId, payload: initialPayload } = requestBody;\n\n if (!workflowId || !availableWorkflows[workflowId]) {\n requestLog(\"ERROR\", \"WorkflowSelection\", `Invalid or missing workflowId: ${workflowId}`);\n return new Response(\n JSON.stringify({\n success: false,\n error: \"Invalid or missing workflowId\",\n availableIds: Object.keys(availableWorkflows),\n debugLog: globalLogger.getLogs(),\n }),\n { status: 400, headers: { \"Content-Type\": \"application/json\" } },\n );\n }\n\n const workflowToExecute = availableWorkflows[workflowId];\n requestLog(\"INFO\", \"WorkflowExecute\", `Executing workflow: ${workflowToExecute.id}`);\n\n const result = await workflowEngine.execute(workflowToExecute, initialPayload || {}); // Pass empty object if payload is undefined\n\n requestLog(\"INFO\", \"WorkflowComplete\", `Workflow ${workflowToExecute.id} finished`, {\n success: result.success,\n durationMs: Date.now() - requestStartTime,\n });\n\n // Ensure the final response includes the debugLog\n const responsePayload = { ...result, debugLog: globalLogger.getLogs() };\n\n return new Response(JSON.stringify(responsePayload), {\n status: result.success ? 200 : (result.error === \"Workflow input error.\" ? 400 : 500),\n headers: { \"Content-Type\": \"application/json\", \"Access-Control-Allow-Origin\": \"*\" },\n });\n } else {\n requestLog(\"WARN\", \"MethodNotAllowed\", `Method: ${req.method}`);\n return new Response(\n JSON.stringify({ success: false, error: \"Method Not Allowed\", debugLog: globalLogger.getLogs() }),\n { status: 405, headers: { \"Content-Type\": \"application/json\", \"Allow\": \"GET, POST\" } },\n );\n }\n } catch (e: any) {\n requestLog(\"ERROR\", \"UnhandledException\", \"Unhandled error in request handler\", e);\n return new Response(\n JSON.stringify({\n success: false,\n error: \"Internal Server Error\",\n details: e.message,\n debugLog: globalLogger.getLogs(),\n }),\n { status: 500, headers: { \"Content-Type\": \"application/json\" } },\n );\n }\n}",[],[204],{"_103":211,"_95":212,"_146":213,"_148":-5,"_149":135,"_150":214,"_152":117,"_105":215,"_100":216,"_155":156,"_157":217,"_159":27,"_160":-5,"_161":218,"_98":219},"27747ef0-2ef7-11f0-84bf-569c3dd06744","ui.ts","02da3640-2c6b-11f0-8d84-569c3dd06744",92,["D",1747029950183],3,"// ui.tsx\n\nexport function generateHtmlShellV2(): string {\n return `\n\nSuperpowered Agent Platform Demo V2 (Parallel)\n\n\n\n

Superpowered Agent Platform Demo V2 (Parallel)

This demo runs the enhancedAnalysis workflow. Steps 1 (fetch) and 2 (summarize) can run in parallel. Step 3 (combine) waits for both.

\n\n

Workflow Result

— Waiting for execution —

Execution Log

— Waiting for execution —

\n\n`;\n}",[],[212],{"_103":221,"_95":222,"_146":223,"_148":-5,"_149":135,"_150":224,"_152":117,"_105":225,"_100":170,"_155":156,"_157":226,"_159":27,"_160":-5,"_161":227,"_98":228},"7d1cc9a4-2f5e-11f0-9406-569c3dd06744","workflowDefinition.ts","e211a83a-2c6a-11f0-ac69-569c3dd06744",94,["D",1747074332085],"// comprehensiveWorkflow.ts\nimport { WorkflowDefinition } from \"https://esm.town/v/salon/mandate/interfaces.ts\"; // Adjust path for Val Town\n\nexport const uiTestWorkflow: WorkflowDefinition = {\n id: \"UITestingWorkflowWithVariousInputs\",\n description: \"A workflow to test UI input generation, conditional logic, and various input sourcing patterns using existing agents.\",\n initialInputSchema: {\n type: \"object\",\n properties: {\n mainProcessingText: {\n type: \"string\",\n description: \"Main text for summarization and to be used as a query.\",\n },\n optionalSupplementaryText: {\n type: \"string\",\n description: \"Optional: Supplementary text for different processing.\",\n },\n fetchSourceURL: {\n type: \"string\",\n description: \"URL for the fetchAgent (e.g., https://jsonplaceholder.typicode.com/todos/1)\",\n default: \"https://jsonplaceholder.typicode.com/todos/1\",\n },\n enableExternalFetch: {\n type: \"boolean\",\n description: \"Tick to enable fetching external data.\",\n default: false,\n },\n enableDetailedMode: {\n type: \"boolean\",\n description: \"Tick to enable detailed mode for conditional steps.\",\n default: true,\n },\n processingChoice: { // UI will render as text input\n type: \"string\",\n description: \"Enter 'A' for one path, 'B' for another (simulates choice).\",\n default: \"A\",\n },\n itemCount: { // UI will render as text input\n type: \"string\",\n description: \"Number of items (e.g., '5') - agent might parse to number.\",\n default: \"3\",\n },\n },\n required: [\"mainProcessingText\", \"fetchSourceURL\"],\n },\n steps: [\n {\n id: \"step1_summarizeMainText\",\n agent: \"summarizerAgent\",\n input: {\n textToSummarize: { source: \"initial\", field: \"mainProcessingText\" },\n },\n // Agent output: { summary: \"...\" }\n },\n {\n id: \"step2_sentimentOnMainSummary\",\n agent: \"sentimentAnalysisAgent\",\n input: {\n articles: [ // Array literal with one object, content sourced from step1\n {\n id: \"mainTextSummary\",\n title: \"Sentiment of Main Summary\",\n rawContent: { source: \"step1_summarizeMainText\", field: \"summary\" },\n },\n ],\n },\n dependencies: [\"step1_summarizeMainText\"],\n // Agent output: { articlesWithSentiment: [{..., sentiment: \"...\"}] }\n },\n {\n id: \"step3_configureBasedOnMainText\",\n agent: \"configurationAgent\",\n input: {\n userQuery: { source: \"initial\", field: \"mainProcessingText\" }, // Use main text as query\n },\n // Agent output: { config: { topic: \"...\", keywords: [], ... } }\n },\n {\n id: \"step4_fetchExternalData\",\n agent: \"fetchAgent\", // Assumes fetchAgent takes a 'url' in its payload. If not, it uses its default.\n input: {\n url: { source: \"initial\", field: \"fetchSourceURL\" }, // Test sourcing URL from initial payload\n },\n config: { // Example of static config for a step, if fetchAgent could use it\n \"static_api_key\": \"ui-test-fetch-key\",\n \"item_limit_from_initial\": { source: \"initial\", field: \"itemCount\" } // Note: Engine currently doesn't resolve sources in step.config\n },\n condition: { source: \"initial\", field: \"enableExternalFetch\", operator: \"isTrue\" },\n // Agent output: { data: {...}, rawText: \"...\", contentType: \"...\" }\n },\n {\n id: \"step5_summarizeOptionalText\",\n agent: \"summarizerAgent\",\n input: {\n textToSummarize: { source: \"initial\", field: \"optionalSupplementaryText\" },\n },\n condition: { source: \"initial\", field: \"optionalSupplementaryText\", operator: \"isNotEmpty\" },\n // Agent output: { summary: \"...\" }\n },\n {\n id: \"step6_combineFetchedAndSummarizedData\",\n agent: \"combinerAgent\",\n input: {\n primarySummary: { source: \"step1_summarizeMainText\", field: \"summary\" },\n fetchedContentRaw: { source: \"step4_fetchExternalData\", field: \"rawText\", optional: true }, // Data from conditional fetch\n optionalSummary: { source: \"step5_summarizeOptionalText\", field: \"summary\", optional: true }, // Data from conditional summary\n staticValue: \"Default Mode\", // Literal string\n processingModeFromInput: { source: \"initial\", field: \"processingChoice\" },\n configurationTopic: { source: \"step3_configureBasedOnMainText\", field: \"config.topic\", optional: true },\n itemCountAsString: { source: \"initial\", field: \"itemCount\" }\n },\n dependencies: [\"step1_summarizeMainText\", \"step4_fetchExternalData\", \"step5_summarizeOptionalText\", \"step3_configureBasedOnMainText\"],\n // Agent output: { combined: \"...\" }\n },\n {\n id: \"step7_generateMockInsights\", // Using summarizer for simplicity to generate \"insights\"\n agent: \"summarizerAgent\",\n input: {\n textToSummarize: { source: \"step6_combineFetchedAndSummarizedData\", field: \"combined\" }\n },\n dependencies: [\"step6_combineFetchedAndSummarizedData\"],\n // Agent output: { summary: \"...\" } -> we'll call this our \"insights\"\n },\n {\n id: \"step8_compileSimpleReport\",\n agent: \"reportCompilationAgent\",\n input: {\n // reportCompilationAgent expects: config, discoveredFeeds?, articlesWithThemes?, trendReport?, anomalyReport?, insights?\n config: { source: \"step3_configureBasedOnMainText\", field: \"config\", optional: true }, // Config from earlier step\n insights: [{ source: \"step7_generateMockInsights\", field: \"summary\" }], // Use summary as a mock insight (needs to be an array)\n articlesWithThemes: [ // Static array with one sourced item and one literal, to test array input resolution\n { title: \"Main Summary Used\", sentiment: { source: \"step2_sentimentOnMainSummary\", field: \"articlesWithSentiment[0].sentiment\", optional: true }, themes: [\"main\"] },\n // The path \"articlesWithSentiment[0].sentiment\" WILL NOT WORK with current resolveNestedValue.\n // For it to work, resolveNestedValue needs to handle array indexing like [0].\n // Let's simplify for now or note this as a known limitation to test.\n // Simplified:\n // { title: \"Main Summary Data\", rawContent: { source: \"step1_summarizeMainText\", field: \"summary\" } },\n // { title: \"Static Test Article\", rawContent: \"This is static content for the report.\" }\n ],\n // To make it simpler for current engine, we'll provide simpler structures for articlesWithThemes or make it empty.\n articlesWithThemes: [], // Simplest for now to avoid path resolution issues\n trendReport: { // Static object for trend report\n dominantSentiment: \"Mixed (UI Test)\",\n keyThemes: [{source: \"initial\", field: \"processingChoice\"}, \"Static Theme\"], // Mixed array\n emergingPatterns: \"N/A for UI Test\",\n },\n anomalyReport: { // Static object for anomaly report\n anomaliesFound: {source: \"initial\", field: \"enableDetailedMode\"}, // Sourced boolean\n anomalyDetails: [],\n },\n // discoveredFeeds can be omitted if optional and agent handles it, or provide empty array\n discoveredFeeds: [],\n },\n dependencies: [\"step3_configureBasedOnMainText\", \"step7_generateMockInsights\", \"step2_sentimentOnMainSummary\"],\n },\n {\n id: \"step9_formatConditionalAlert\",\n agent: \"alertFormattingAgent\",\n input: {\n // alertFormattingAgent expects: config, trendReport?, anomalyReport?, insights?\n config: { source: \"step3_configureBasedOnMainText\", field: \"config\", optional: true },\n insights: [{ source: \"step7_generateMockInsights\", field: \"summary\" }],\n trendReport: { source: \"step8_compileSimpleReport\", field: \"input.trendReport\" }, // Sourcing from another step's *resolved input*\n // The above line `field: \"input.trendReport\"` is an interesting test for resolveNestedValue.\n // It means the payload of step8 would need to contain an 'input' key.\n // Agents usually output a 'payload' which is their result, not their input.\n // Let's use the static trendReport from step8's definition for simplicity or its output.\n // Step8 (reportCompilationAgent) outputs { finalReport: \"...\" }. It does not output its input.\n // So, we must reconstruct or reuse:\n trendReport: { // Re-using the static object definition from step8 for simplicity\n dominantSentiment: \"Mixed (UI Test Alert)\",\n keyThemes: [{source: \"initial\", field: \"processingChoice\"}, \"Static Alert Theme\"],\n emergingPatterns: \"Alert Pattern for UI Test\",\n },\n anomalyReport: { anomaliesFound: {source: \"initial\", field: \"enableDetailedFetch\"}, anomalyDetails: [{description:\"Alert test\"}] },\n },\n condition: { source: \"initial\", field: \"enableDetailedMode\", operator: \"isTrue\" },\n dependencies: [\"step3_configureBasedOnMainText\", \"step7_generateMockInsights\", \"step8_compileSimpleReport\"],\n }\n ],\n outputMapping: {\n finalReportFromTest: { source: \"step8_compileSimpleReport\", field: \"finalReport\", optional: true },\n alertFromTest: { source: \"step9_formatConditionalAlert\", field: \"alertMessage\", optional: true },\n combinedDataFromTest: { source: \"step6_combineFetchedAndSummarizedData\", field: \"combined\" },\n mainSummaryFromTest: { source: \"step1_summarizeMainText\", field: \"summary\" },\n mainSentimentPayloadFromTest: { source: \"step2_sentimentOnMainSummary\", field: \"articlesWithSentiment\", optional: true },\n optionalConfigTopic: { source: \"step3_configureBasedOnMainText\", field: \"config.topic\", optional: true },\n fetchedDataRaw: { source: \"step4_fetchExternalData\", field: \"rawText\", optional: true },\n initialInputsUsed: { source: \"initial\", field: \".\"} // Echo all initial inputs\n },\n};\n\nexport const comprehensiveTopicAnalysisWorkflow: WorkflowDefinition = {\n id: \"ComprehensiveTopicAnalysisWithOptionalUserText\",\n description:\n \"Performs a complex, multi-step analysis of a topic using news feeds (fetch, clean, relevance, sentiment, themes, trends) and, in parallel, processes optional user-provided text (summarize, sentiment). Results are then synthesized into a comprehensive report and potential alerts.\",\n initialInputSchema: {\n type: \"object\",\n properties: {\n userQuery: {\n type: \"string\",\n description: \"The main query or topic for analysis.\",\n },\n optionalUserTextToAnalyze: {\n type: \"string\",\n description: \"Optional supplementary text provided by the user for analysis.\",\n },\n enableAlerting: {\n type: \"boolean\",\n description: \"Flag to enable or disable alert generation.\",\n default: false,\n },\n // You might add other initial global configs here if not derived by configurationAgent\n // e.g., maxHeadlinesPerFeed, analysisDepth, defaultNewsSources\n },\n required: [\"userQuery\"],\n },\n steps: [\n {\n id: \"step1_configureAnalysis\",\n agent: \"configurationAgent\", // This agent MUST output { config: { topic: ..., keywords: ..., ...} }\n input: { userQuery: { source: \"initial\", field: \"userQuery\" } },\n dependencies: [],\n },\n {\n id: \"step2A_validateSources\",\n agent: \"sourceValidationAgent\",\n input: { config: { source: \"step1_configureAnalysis\", field: \"config\" } }, // Relies on step1_configureAnalysis.config\n dependencies: [\"step1_configureAnalysis\"],\n },\n {\n id: \"step2A_fetchNews\",\n agent: \"parallelFetchAgent\",\n input: {\n feedsToFetch: { source: \"step2A_validateSources\", field: \"discoveredFeeds\" },\n // Assumes maxHeadlinesPerFeed is part of the config object from step1\n maxHeadlinesPerFeed: { source: \"step1_configureAnalysis\", field: \"config.maxHeadlinesPerFeed\" },\n },\n dependencies: [\"step1_configureAnalysis\", \"step2A_validateSources\"],\n },\n {\n id: \"step2A_cleanArticles\",\n agent: \"articleCleaningAgent\",\n input: { articles: { source: \"step2A_fetchNews\", field: \"fetchedArticles\" } },\n dependencies: [\"step2A_fetchNews\"],\n },\n {\n id: \"step2A_assessRelevance\",\n agent: \"relevanceAssessmentAgent\",\n input: {\n articles: { source: \"step2A_cleanArticles\", field: \"processedArticles\" },\n topic: { source: \"step1_configureAnalysis\", field: \"config.topic\" },\n keywords: { source: \"step1_configureAnalysis\", field: \"config.keywords\" },\n },\n dependencies: [\"step1_configureAnalysis\", \"step2A_cleanArticles\"],\n },\n {\n id: \"step2A_extractContent\",\n agent: \"contentExtractionAgent\",\n input: {\n articles: { source: \"step2A_assessRelevance\", field: \"relevantArticles\" },\n analysisDepth: { source: \"step1_configureAnalysis\", field: \"config.analysisDepth\" },\n },\n dependencies: [\"step1_configureAnalysis\", \"step2A_assessRelevance\"],\n },\n {\n id: \"step2A_newsSentiment\",\n agent: \"sentimentAnalysisAgent\",\n input: { articles: { source: \"step2A_extractContent\", field: \"articlesWithContent\" } },\n dependencies: [\"step2A_extractContent\"],\n },\n {\n id: \"step2A_newsThemes\",\n agent: \"keyThemeExtractionAgent\",\n input: {\n articles: { source: \"step2A_newsSentiment\", field: \"articlesWithSentiment\" },\n topic: { source: \"step1_configureAnalysis\", field: \"config.topic\" },\n },\n dependencies: [\"step1_configureAnalysis\", \"step2A_newsSentiment\"],\n },\n {\n id: \"step2B_summarizeUserText\",\n agent: \"summarizerAgent\",\n input: { textToSummarize: { source: \"initial\", field: \"optionalUserTextToAnalyze\" } },\n condition: { source: \"initial\", field: \"optionalUserTextToAnalyze\", operator: \"isNotEmpty\" },\n dependencies: [],\n },\n {\n id: \"step2B_userTextSentiment\",\n agent: \"sentimentAnalysisAgent\",\n input: { // Input needs to be an array of \"article-like\" objects for sentimentAnalysisAgent\n articles: [{\n id: \"userTextItemForSentiment\", // Static ID for this pseudo-article\n title: \"User Provided Text Sentiment Analysis\", // Static title\n // The content for sentiment analysis is the summary from the previous step\n rawContent: { source: \"step2B_summarizeUserText\", field: \"summary\" },\n // Add other fields if your sentiment agent expects them, e.g., sourceUrl: null\n }],\n },\n // Condition ensures this step only runs if summarization happened and produced a summary\n condition: { source: \"step2B_summarizeUserText\", field: \"summary\", operator: \"isNotEmpty\" },\n dependencies: [\"step2B_summarizeUserText\"],\n },\n {\n id: \"step3_newsTrendsAnomalies\",\n agent: \"trendAndAnomalyDetectionAgent\",\n input: {\n articlesWithThemes: { source: \"step2A_newsThemes\", field: \"articlesWithThemes\" },\n topic: { source: \"step1_configureAnalysis\", field: \"config.topic\" },\n },\n dependencies: [\"step1_configureAnalysis\", \"step2A_newsThemes\"],\n },\n {\n id: \"step3_generateNewsInsights\",\n agent: \"insightGenerationAgent\",\n input: {\n trendReport: { source: \"step3_newsTrendsAnomalies\", field: \"trendReport\" },\n anomalyReport: { source: \"step3_newsTrendsAnomalies\", field: \"anomalyReport\" },\n config: { source: \"step1_configureAnalysis\", field: \"config\" },\n articlesCount: { source: \"step2A_assessRelevance\", field: \"relevantArticles.length\" }, // Assuming relevantArticles is an array\n },\n dependencies: [\"step1_configureAnalysis\", \"step3_newsTrendsAnomalies\", \"step2A_assessRelevance\"],\n },\n {\n id: \"step4_compileReport\",\n agent: \"reportCompilationAgent\",\n input: {\n config: { source: \"step1_configureAnalysis\", field: \"config\" },\n discoveredFeeds: { source: \"step2A_validateSources\", field: \"discoveredFeeds\", optional: true },\n articlesWithThemes: { source: \"step2A_newsThemes\", field: \"articlesWithThemes\", optional: true },\n trendReport: { source: \"step3_newsTrendsAnomalies\", field: \"trendReport\", optional: true },\n anomalyReport: { source: \"step3_newsTrendsAnomalies\", field: \"anomalyReport\", optional: true },\n insights: { source: \"step3_generateNewsInsights\", field: \"insights\", optional: true },\n userTextSummaryResult: { source: \"step2B_summarizeUserText\", field: \"summary\", optional: true },\n userTextSentimentResult: { // Changed from userTextSentimentArticles for clarity, assuming agent outputs sentiment directly for the item\n source: \"step2B_userTextSentiment\",\n field: \"articlesWithSentiment\", // Assuming sentiment agent returns an array, take first item, or agent adapts\n optional: true,\n },\n },\n dependencies: [\n \"step1_configureAnalysis\",\n \"step2A_validateSources\", // Ensures feed discovery attempt\n \"step2A_newsThemes\", // Ensures news analysis branch completes if data available\n \"step3_newsTrendsAnomalies\",\n \"step3_generateNewsInsights\",\n \"step2B_summarizeUserText\", // Ensures user text summarization attempt\n \"step2B_userTextSentiment\", // Ensures user text sentiment analysis attempt\n ],\n },\n {\n id: \"step4_formatAlert\",\n agent: \"alertFormattingAgent\",\n input: {\n config: { source: \"step1_configureAnalysis\", field: \"config\" },\n trendReport: { source: \"step3_newsTrendsAnomalies\", field: \"trendReport\", optional: true }, // Optional as trends might not always be present\n anomalyReport: { source: \"step3_newsTrendsAnomalies\", field: \"anomalyReport\", optional: true }, // Optional\n insights: { source: \"step3_generateNewsInsights\", field: \"insights\", optional: true }, // Optional\n // You might also want to pass specific user text sentiment if it's part of alert criteria\n // userTextSentiment: { source: \"step2B_userTextSentiment\", field: \"articlesWithSentiment[0].sentiment\", optional: true }\n },\n condition: { source: \"initial\", field: \"enableAlerting\", operator: \"isTrue\" }, // Explicitly check for true\n dependencies: [\"step1_configureAnalysis\", \"step3_newsTrendsAnomalies\", \"step3_generateNewsInsights\"],\n },\n ],\n outputMapping: {\n report: { source: \"step4_compileReport\", field: \"compiledReport\" }, // Assuming reportCompilationAgent outputs { compiledReport: ... }\n alert: { source: \"step4_formatAlert\", field: \"alertMessage\", optional: true }, // Assuming alertFormattingAgent outputs { alertData: ... }\n configurationUsed: { source: \"step1_configureAnalysis\", field: \"config\" }, // To see what config was used\n },\n};\n\nexport const enhancedAnalysisWorkflow: WorkflowDefinition = {\n id: \"enhancedAnalysis\",\n description: \"Fetches data (conditionally), summarizes text, and combines results with parallel potential.\",\n initialInputSchema: {\n type: \"object\",\n properties: {\n userText: { type: \"string\" },\n enableExternalFetch: { type: \"boolean\", default: false }, // Added default\n },\n required: [\"userText\"],\n },\n steps: [\n {\n id: \"step1_fetchData\",\n agent: \"fetcher\", // Assuming this agent outputs { data: ..., rawText: ..., contentType: ... }\n config: { defaultUrl: \"https://feeds.npr.org/1001/rss.xml\" },\n condition: { source: \"initial\", field: \"enableExternalFetch\", operator: \"isTrue\" }, // Explicitly check for true\n dependencies: [],\n },\n {\n id: \"step2_summarize\",\n agent: \"summarizer\", // Assuming this agent outputs { summary: \"...\" }\n input: { textToSummarize: { source: \"initial\", field: \"userText\" } },\n dependencies: [],\n },\n {\n id: \"step3_combine\",\n agent: \"combiner\", // This agent must handle optional externalData fields\n input: {\n summary: { source: \"step2_summarize\", field: \"summary\" },\n externalData: { source: \"step1_fetchData\", field: \"data\", optional: true },\n externalDataRawText: { source: \"step1_fetchData\", field: \"rawText\", optional: true },\n externalDataContentType: { source: \"step1_fetchData\", field: \"contentType\", optional: true },\n },\n dependencies: [\"step1_fetchData\", \"step2_summarize\"], // Dependencies ensure order if conditional step runs\n },\n ],\n outputMapping: {\n finalCombinedOutput: { source: \"step3_combine\", field: \"combined\" }, // Assuming combinerAgent outputs { combined: ... }\n originalSummaryProvided: { source: \"step2_summarize\", field: \"summary\" },\n fetchStepRan: { source: \"initial\", field: \"enableExternalFetch\" }, // To confirm if fetch was attempted\n fetchedData: { source: \"step1_fetchData\", field: \"data\", optional: true }, // Optionally output fetched data if it ran\n },\n};",[],[222],{"_103":230,"_95":231,"_146":232,"_148":-5,"_149":135,"_150":233,"_152":117,"_105":234,"_100":235,"_155":156,"_157":236,"_159":27,"_160":-5,"_161":237,"_98":238},"866602e4-2ef7-11f0-aa16-569c3dd06744","workflowEngine.ts","b78ab462-2c6a-11f0-bb53-569c3dd06744",93,["D",1747030109501],6,"// workflowEngine.ts\nimport { AgentRegistry } from \"https://esm.town/v/salon/mandate/agentRegistry.ts\";\nimport {\n AgentContext,\n AgentInput,\n AgentOutput,\n InputMapping, // Keep this type for explicit mappings\n WorkflowDefinition,\n WorkflowResult,\n WorkflowStep,\n} from \"https://esm.town/v/salon/mandate/interfaces.ts\";\nimport { MemoryLogger } from \"https://esm.town/v/salon/mandate/logger.ts\";\n\nexport class WorkflowEngine {\n constructor(private registry: AgentRegistry, private logger: MemoryLogger) {}\n\n private resolveNestedValue(obj: any, fieldPath: string): any {\n if (obj === undefined || obj === null || !fieldPath) return undefined; // Guard against null/undefined obj\n return fieldPath.split(\".\").reduce(\n (acc, part) => (acc && typeof acc === \"object\" && acc[part] !== undefined) ? acc[part] : undefined,\n obj,\n );\n }\n\n // New recursive function to resolve any kind of input structure\n private resolveInputStructure(\n structure: any,\n initialPayload: any,\n stepResults: Map>,\n contextStepId?: string, // Optional: for more detailed logging\n contextInputKey?: string, // Optional: for more detailed logging\n ): any {\n const log = this.logger.createLogFunction(\n contextStepId ? `M-UnknownStep-${contextStepId}` : `M-UnknownWorkflow`, // Placeholder mandateId for this internal log\n \"WorkflowEngine:InputResolver\",\n );\n\n // Case 1: It's an InputMapping object { source: \"...\", field: \"...\" }\n if (\n typeof structure === \"object\"\n && structure !== null\n && typeof structure.source === \"string\"\n && typeof structure.field === \"string\"\n && Object.keys(structure).length >= 2 && Object.keys(structure).length <= 3 // Allow for 'optional'\n ) {\n const sourceData = structure.source === \"initial\"\n ? initialPayload\n : stepResults.get(structure.source)?.payload;\n\n if (sourceData === undefined && structure.source !== \"initial\" && !stepResults.has(structure.source)) {\n // This can happen if a source step was skipped and the input wasn't marked optional in the workflow definition\n // or if this input mapping is for an output that an agent didn't produce.\n if (contextStepId && contextInputKey) {\n log(\n \"WARN\",\n `Step:${contextStepId}`,\n `Source step \"${structure.source}\" for input \"${contextInputKey}\" did not produce output or was not found.`,\n );\n }\n return undefined; // Or handle based on structure.optional if you implement that\n }\n const value = this.resolveNestedValue(sourceData, structure.field);\n if (value === undefined && contextStepId && contextInputKey) {\n // Log if a specific mapping resolves to undefined, helps debug missing fields\n log(\n \"WARN\",\n `Step:${contextStepId}`,\n `Input mapping for \"${contextInputKey}\" resolved to undefined from ${structure.source}.${structure.field}`,\n );\n }\n return value;\n }\n // Case 2: It's an array\n else if (Array.isArray(structure)) {\n return structure.map(item =>\n this.resolveInputStructure(item, initialPayload, stepResults, contextStepId, contextInputKey)\n );\n }\n // Case 3: It's a non-InputMapping, non-array object\n else if (typeof structure === \"object\" && structure !== null) {\n const resolvedObject: Record = {};\n for (const [key, value] of Object.entries(structure)) {\n resolvedObject[key] = this.resolveInputStructure(\n value,\n initialPayload,\n stepResults,\n contextStepId,\n contextInputKey ? `${contextInputKey}.${key}` : key,\n );\n }\n return resolvedObject;\n }\n // Case 4: It's a literal (string, number, boolean, null)\n return structure;\n }\n\n // The original resolveInputValue can be kept specifically for OutputMapping\n // as output mappings are expected to be simple {source, field}\n private resolveOutputValue(\n mapping: InputMapping, // OutputMappingValue is essentially an InputMapping\n initialPayload: any,\n stepResults: Map>,\n ): any {\n if (!mapping || typeof mapping.source !== \"string\" || typeof mapping.field !== \"string\") {\n this.logger.record({\n mandateId: \"WorkflowEngine\", // Generic mandateId for this type of error\n level: \"ERROR\",\n component: \"WorkflowEngine:OutputResolver\",\n message: \"Invalid output mapping structure\",\n details: { mapping },\n });\n return undefined;\n }\n const sourceData = mapping.source === \"initial\" ? initialPayload : stepResults.get(mapping.source)?.payload;\n // Check if the source step's payload exists, especially if the step might have been skipped or failed\n if (sourceData === undefined) {\n if (mapping.optional) { // Assuming OutputMappingValue might have an 'optional' field\n return undefined; // It's optional, so undefined is acceptable\n }\n this.logger.record({\n mandateId: \"WorkflowEngine\",\n level: \"WARN\",\n component: \"WorkflowEngine:OutputResolver\",\n message: `Source data for output mapping from step \"${mapping.source}\" is undefined.`,\n details: { mapping },\n });\n return undefined;\n }\n return this.resolveNestedValue(sourceData, mapping.field);\n }\n\n async execute(\n definition: WorkflowDefinition,\n initialPayload: InitialPayload,\n ): Promise> {\n const mandateId = `M-${Date.now()}-${definition.id}`;\n const log = this.logger.createLogFunction(mandateId, \"WorkflowEngine\");\n const stepResults = new Map>();\n const completedSteps = new Set();\n const runningSteps = new Set();\n\n log(\"INFO\", `Workflow:${definition.id}`, `Execution starting`, { initialPayload });\n\n // --- Initial Input Validation (simplified for brevity, your existing logic is fine) ---\n if (definition.initialInputSchema) {\n // ... (your validation logic - seems okay)\n if (false /* replace with actual validation failure */) {\n // ... return error ...\n }\n log(\"DEBUG\", `Workflow:${definition.id}`, `Initial input validation passed.`);\n }\n\n try {\n const stepLookup = new Map(definition.steps.map(step => [step.id, step]));\n\n while (completedSteps.size < definition.steps.length) {\n const readyToExecuteSteps: WorkflowStep[] = [];\n for (const step of definition.steps) {\n if (completedSteps.has(step.id) || runningSteps.has(step.id)) continue;\n\n const dependenciesMet = (step.dependencies ?? []).every(depId => completedSteps.has(depId));\n if (!dependenciesMet) continue;\n\n if (step.condition) {\n // Condition mapping should also use the new resolveInputStructure\n // Assuming step.condition is a standard InputMapping {source, field, operator}\n // For simplicity, if step.condition is just {source, field}, resolveInputValue is okay\n // but ideally, condition resolution might need to be more flexible too.\n // For now, let's assume current condition resolution is working based on logs.\n // Current logic for condition value:\n const conditionSourceValue = this.resolveInputStructure(\n step.condition,\n initialPayload,\n stepResults,\n step.id,\n \"conditionSource\",\n );\n\n let conditionResult = false;\n if (step.condition.operator === \"isNotEmpty\") {\n conditionResult = conditionSourceValue !== undefined && conditionSourceValue !== null\n && conditionSourceValue !== \"\";\n if (Array.isArray(conditionSourceValue)) conditionResult = conditionSourceValue.length > 0;\n } else if (step.condition.operator === \"isTrue\") {\n conditionResult = !!conditionSourceValue;\n } else { // Default: truthy\n conditionResult = !!conditionSourceValue;\n }\n\n if (!conditionResult) {\n log(\"INFO\", `Step:${step.id}`, `Skipped due to condition not met`, {\n conditionField: step.condition.field,\n value: conditionSourceValue,\n });\n completedSteps.add(step.id);\n continue;\n }\n log(\"DEBUG\", `Step:${step.id}`, `Condition met`, {\n conditionField: step.condition.field,\n value: conditionSourceValue,\n });\n }\n readyToExecuteSteps.push(step);\n }\n\n if (readyToExecuteSteps.length === 0 && completedSteps.size < definition.steps.length) {\n // ... (your deadlock detection - seems okay) ...\n throw new Error(`Workflow deadlock or circular dependency.`);\n }\n if (readyToExecuteSteps.length === 0 && completedSteps.size === definition.steps.length) break;\n\n const executionPromises = readyToExecuteSteps.map(async (step) => {\n runningSteps.add(step.id);\n const agent = this.registry.getAgent(step.agent);\n if (!agent) throw new Error(`Agent \"${step.agent}\" for step \"${step.id}\" not found.`);\n\n const agentInputPayload: Record = {};\n if (step.input) {\n for (const [key, inputStructureToResolve] of Object.entries(step.input)) {\n // Use the new recursive resolver\n agentInputPayload[key] = this.resolveInputStructure(\n inputStructureToResolve,\n initialPayload,\n stepResults,\n step.id, // Pass stepId for context\n key, // Pass inputKey for context\n );\n }\n }\n // ... (rest of agent execution logic: context, agent call, error handling, stepResults.set) ...\n const taskId = `T-${Date.now()}-${step.id}`;\n const agentContext: AgentContext = {\n /* ... your context setup ... */\n mandateId,\n taskId,\n log: this.logger.createLogFunction(mandateId, `Agent:${step.agent}`), // Simplified logger for agent\n config: step.config,\n };\n const agentInput: AgentInput = { mandateId, taskId, payload: agentInputPayload };\n\n log(\"DEBUG\", `Step:${step.id}`, `Executing agent \"${step.agent}\"`, { input: agentInputPayload }); // This log will now be more accurate\n try {\n const output = await agent(agentInput, agentContext);\n if (output.error) throw new Error(`Agent \"${step.agent}\" (Step \"${step.id}\") failed: ${output.error}`);\n // ... (correlationId check) ...\n log(\"DEBUG\", `Step:${step.id}`, `Agent \"${step.agent}\" completed`, { output: output.payload });\n stepResults.set(step.id, output); // Store the full AgentOutput\n return { stepId: step.id, success: true, error: null };\n } catch (e: any) {\n log(\"ERROR\", `Step:${step.id}`, `Agent \"${step.agent}\" execution error`, {\n message: e.message,\n stack: e.stack,\n });\n return { stepId: step.id, success: false, error: e };\n } finally {\n runningSteps.delete(step.id);\n }\n }); // End of map for executionPromises\n\n const results = await Promise.allSettled(executionPromises);\n // ... (your results processing and error handling - seems okay) ...\n let cycleError: Error | null = null;\n results.forEach(result => {\n if (result.status === \"fulfilled\" && result.value) {\n const { stepId, success, error } = result.value;\n if (success) {\n completedSteps.add(stepId);\n } else {\n if (!cycleError && error) cycleError = error;\n completedSteps.add(stepId); // Mark as completed even if failed to prevent re-execution\n }\n } else if (result.status === \"rejected\") {\n if (!cycleError) cycleError = result.reason;\n log(\"ERROR\", `Workflow:${definition.id}`, `Critical error in step execution promise.`, {\n reason: result.reason,\n });\n }\n });\n if (cycleError) throw cycleError;\n } // End while loop\n\n let finalPayload: any = {};\n if (definition.outputMapping) {\n for (const [key, mapping] of Object.entries(definition.outputMapping)) {\n // Use the specific resolveOutputValue for output mappings\n const value = this.resolveOutputValue(mapping as InputMapping, initialPayload, stepResults);\n // Only add to finalPayload if value is not undefined, or handle as per your needs\n // This will ensure keys with undefined values (due to failed resolution or optional not found) are omitted\n // which matches the previous logging behavior.\n if (value !== undefined) {\n finalPayload[key] = value;\n } else {\n log(\n \"WARN\",\n `Workflow:${definition.id}`,\n `Output mapping for key \"${key}\" resolved to undefined. Source: ${mapping.source}, Field: ${mapping.field}. It will be omitted from final payload.`,\n );\n }\n }\n } else if (definition.steps.length > 0) {\n // ... (your default output logic - seems okay, but ensure it handles undefined payloads gracefully) ...\n }\n\n log(\"SUCCESS\", `Workflow:${definition.id}`, `Execution finished successfully`, { finalPayload });\n return {\n mandateId,\n workflowId: definition.id,\n success: true,\n finalPayload: finalPayload as FinalPayload,\n debugLog: this.logger.getLogs(),\n };\n } catch (error: any) {\n log(\"ERROR\", `Workflow:${definition.id}`, `Execution failed`, { message: error.message, stack: error.stack });\n return {\n // ... (your error return structure - seems okay) ...\n mandateId,\n workflowId: definition.id,\n success: false,\n error: `Workflow \"${definition.id}\" failed.`,\n details: error?.message ?? String(error),\n debugLog: this.logger.getLogs(),\n };\n }\n }\n}",[],[231],"files",[143,164,174,184,193,202,210,220,229],"counts",{"_243":31,"_244":245,"_246":31,"_247":31},"pulls","branches",1,"forks","envVars","contents",[],"file",{"_103":185,"_95":99,"_146":186,"_148":-5,"_149":135,"_150":187,"_152":117,"_105":252,"_100":189,"_155":156,"_157":190,"_159":27,"_160":-5,"_98":253},["D",1747026465572],[99],"upstream","emptyEnvVars","isNew","routes/_app.x.$handle.$name.code.$",{"_93":94,"_95":96,"_97":41,"_100":-7,"_98":99,"_101":259,"_141":263,"_239":291,"_250":292,"_248":294,"_254":-5,"_295":27,"_296":-5},{"_103":104,"_95":96,"_105":260,"_107":-5,"_108":27,"_109":110,"_111":-5,"_112":-5,"_12":113,"_114":115,"_116":117,"_118":119,"_120":31,"_121":122,"_123":122,"_124":125,"_126":125,"_127":125,"_128":122,"_129":27,"_21":261,"_133":262},["D",1746749580352],{"_93":94,"_131":132},{"_100":119,"_95":41,"_103":135,"_136":104,"_137":-5,"_138":-5,"_139":-5,"_140":-5},[264,267,270,273,276,279,282,285,288],{"_103":144,"_95":145,"_146":147,"_148":-5,"_149":135,"_150":151,"_152":117,"_105":153,"_100":154,"_155":156,"_157":158,"_159":27,"_160":-5,"_161":265,"_98":266},[],[145],{"_103":165,"_95":166,"_146":167,"_148":-5,"_149":135,"_150":168,"_152":117,"_105":169,"_100":170,"_155":156,"_157":171,"_159":27,"_160":-5,"_161":268,"_98":269},[],[166],{"_103":175,"_95":176,"_146":177,"_148":-5,"_149":135,"_150":178,"_152":117,"_105":179,"_100":180,"_155":156,"_157":181,"_159":27,"_160":-5,"_161":271,"_98":272},[],[176],{"_103":185,"_95":99,"_146":186,"_148":-5,"_149":135,"_150":187,"_152":117,"_105":188,"_100":189,"_155":156,"_157":190,"_159":27,"_160":-5,"_161":274,"_98":275},[],[99],{"_103":194,"_95":195,"_146":196,"_148":-5,"_149":135,"_150":197,"_152":117,"_105":198,"_100":154,"_155":156,"_157":199,"_159":27,"_160":-5,"_161":277,"_98":278},[],[195],{"_103":203,"_95":204,"_146":115,"_148":-5,"_149":135,"_150":119,"_152":117,"_105":205,"_100":151,"_155":206,"_157":207,"_159":27,"_160":-5,"_161":280,"_98":281},[],[204],{"_103":211,"_95":212,"_146":213,"_148":-5,"_149":135,"_150":214,"_152":117,"_105":215,"_100":216,"_155":156,"_157":217,"_159":27,"_160":-5,"_161":283,"_98":284},[],[212],{"_103":221,"_95":222,"_146":223,"_148":-5,"_149":135,"_150":224,"_152":117,"_105":225,"_100":170,"_155":156,"_157":226,"_159":27,"_160":-5,"_161":286,"_98":287},[],[222],{"_103":230,"_95":231,"_146":232,"_148":-5,"_149":135,"_150":233,"_152":117,"_105":234,"_100":235,"_155":156,"_157":236,"_159":27,"_160":-5,"_161":289,"_98":290},[],[231],[264,267,270,273,276,279,282,285,288],{"_103":185,"_95":99,"_146":186,"_148":-5,"_149":135,"_150":187,"_152":117,"_105":252,"_100":189,"_155":156,"_157":190,"_159":27,"_160":-5,"_98":293},[99],[],"showCompareCallout","pull","routes/_app","actionData","errors"]