[{"_1":2,"_170":-5,"_171":-5},"loaderData",{"_3":4,"_91":92},"root",{"_5":6,"_34":35,"_64":65},"ctx",{"_7":8,"_23":-5,"_24":25,"_26":27,"_28":20,"_29":20,"_30":20,"_31":32,"_33":20},"auth",{"_9":-5,"_10":-5,"_11":-5,"_12":-5,"_13":-5,"_14":-5,"_15":-5,"_16":-5,"_17":-5,"_18":-5,"_19":20,"_21":22},"sessionClaims","sessionId","sessionStatus","userId","actor","orgId","orgRole","orgSlug","orgPermissions","factorVerificationAge","isAuthenticated",false,"tokenType","session_token","user","preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","notificationCount",0,"flagDenols","ENV",{"_36":37,"_38":39,"_40":41,"_42":43,"_44":45,"_46":47,"_48":49,"_50":51,"_52":53,"_54":55,"_56":57,"_58":59,"_60":61,"_62":63},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","VAL_RUNTIME_NAME","prod","RENDER_GIT_COMMIT","2f607f7b74b4e52529fb75176d403d8b5b284eb1","RENDER_GIT_BRANCH","main","CLERK_PUBLISHABLE_KEY","pk_live_Y2xlcmsudmFsLnRvd24k","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","API_URL","https://api.val.town","clerkState",{"_66":67},"__internal_clerk_state",{"_68":69,"_70":45,"_71":72,"_73":72,"_74":20,"_75":72,"_76":72,"_77":72,"_78":72,"_79":80,"_89":20,"_90":20},"__clerk_ssr_state",{"_21":22,"_9":-5,"_10":-5,"_11":-5,"_12":-5,"_13":-5,"_14":-5,"_15":-5,"_16":-5,"_17":-5,"_18":-5,"_19":20},"__publishableKey","__proxyUrl","","__domain","__isSatellite","__signInUrl","__signUpUrl","__afterSignInUrl","__afterSignUpUrl","__clerk_debug",{"_81":20,"_19":20,"_82":72,"_83":84,"_85":72,"_86":45,"_87":20,"_88":72},"isSignedIn","proxyUrl","reason","session-token-and-uat-missing","message","publishableKey","isSatellite","domain","__telemetryDisabled","__telemetryDebug","routes/embed.x.$handle.$name.$",{"_93":94,"_95":96,"_97":98,"_99":100,"_148":149,"_167":168,"_169":-5},"handle","salon","name","mandate","path","agent_ecosystem.ts","project",{"_101":102,"_95":96,"_103":104,"_105":-5,"_106":20,"_107":108,"_109":-5,"_110":-5,"_12":111,"_112":113,"_114":115,"_116":117,"_118":32,"_14":119,"_120":121,"_122":121,"_123":124,"_125":124,"_126":124,"_127":121,"_128":121,"_23":129,"_132":133,"_135":136,"_147":20},"id","5dfdba8e-2c6a-11f0-8699-569c3dd06744","createdAt",["D",1746749580352],"deletedAt","legacyEnvVars","privacy","public","description","imageUrl","b12b39d0-1fe4-11f0-9001-4fc1852e0e06","pinnedFileId","80211736-2c70-11f0-b8c8-569c3dd06744","pinPreview",true,"screenshotVersion",96,"likeCount","45a0d482-0d4d-4887-9bc4-d08243be47a9","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm",{"_93":94,"_130":131},"profileImageUrl","https://img.clerk.com/eyJ0eXBlIjoiZGVmYXVsdCIsImlpZCI6Imluc18yTm10cGhQUklwU0tqZHNiOHNnNGd2ZjJURFUiLCJyaWQiOiJ1c2VyXzJ3NnI5NzZ4ajBmUUZoTUxYdExrMmZYWnAzdyJ9","org",{"_93":94,"_134":131},"avatar","branch",{"_137":117,"_95":43,"_101":138,"_139":102,"_140":-5,"_141":-5,"_142":-5,"_143":-5,"_144":-5,"_145":96,"_146":94},"version","5dfe05ac-2c6a-11f0-8699-569c3dd06744","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","isOwned","file",{"_101":150,"_95":98,"_151":152,"_153":-5,"_154":138,"_155":156,"_157":115,"_103":158,"_137":159,"_160":161,"_162":163,"_164":20,"_165":-5,"_97":166},"83a3552c-2ef0-11f0-9d91-569c3dd06744","fileId","077c9e6c-2d16-11f0-8497-569c3dd06744","parentId","branchId","branchVersion",90,"latest",["D",1747027098390],13,"fileType","script","content","// agent_ecosystem.ts\nimport { XMLParser } from \"https://esm.sh/fast-xml-parser@4.2.5\"; // For parsing RSS/Atom feeds\n// import { combinerAgent, fetchAgent, summarizerAgent } from \"https://esm.town/v/salon/mandate/agents.ts\";\nimport { AgentContext, AgentInput, AgentOutput } from \"https://esm.town/v/salon/mandate/interfaces.ts\";\nimport { fetch } from \"https://esm.town/v/std/fetch\";\nimport { OpenAI } from \"https://esm.town/v/std/openai\";\n// If Deno.env is used, ensure it's for Val Town secrets if possible or clearly noted.\n// For direct Deno.env.get('OPENAI_API_KEY'), it relies on the Val Town environment variable being set.\n// The std/openai library usually handles API key abstraction using user's Val Town secrets.\n\n// Your provided summarizerAgent\nexport async function summarizerAgent(\n input: AgentInput<{ textToSummarize: string }>,\n context: AgentContext,\n): Promise> {\n const { mandateId, taskId, payload } = input;\n const { log } = context;\n log(\"INFO\", \"SummarizerAgent\", `Start`);\n\n try {\n if (!payload || typeof payload.textToSummarize !== \"string\") {\n throw new Error(\"Invalid input: \\\"textToSummarize\\\" missing or not a string.\");\n }\n if (payload.textToSummarize.trim() === \"\") {\n log(\"WARN\", \"SummarizerAgent\", \"Input text empty, returning empty summary.\");\n return { mandateId, correlationId: taskId, payload: { summary: \"\" } };\n }\n\n const openai = new OpenAI(); // Using OpenAI client as per your example\n\n log(\"INFO\", \"SummarizerAgent\", \"Generating summary with OpenAI...\");\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\", // Explicitly setting a model\n messages: [\n {\n role: \"system\",\n content: \"You are an expert summarization assistant. Provide a concise summary.\",\n },\n { role: \"user\", content: `Please summarize the following text:\\n\\n${payload.textToSummarize}` },\n ],\n temperature: 0.5,\n });\n\n const summary = completion.choices[0]?.message?.content?.trim() ?? \"Could not generate summary.\";\n\n if (summary === \"Could not generate summary.\") {\n log(\"WARN\", \"SummarizerAgent\", \"OpenAI did not return a valid summary content.\");\n }\n\n log(\"SUCCESS\", \"SummarizerAgent\", `Complete`);\n return { mandateId, correlationId: taskId, payload: { summary } };\n } catch (e: any) {\n log(\"ERROR\", \"SummarizerAgent\", `Failed: ${e.message}`, e);\n return { mandateId, correlationId: taskId, payload: { summary: \"\" }, error: e.message };\n }\n}\n\n// Your provided fetchAgent (Note: \"Workspaceing\" seems like a typo, changed to \"Fetching\")\nexport async function fetchAgent(\n input: AgentInput<{ url_from_input?: string; maxHeadlines?: number }>,\n context: AgentContext,\n): Promise<\n AgentOutput<{\n data: {\n headlines: Array<{ title: string; link: string; summary: string; publishedDate: string }>;\n feedTitle: string;\n sourceUrl: string;\n message?: string;\n };\n rawText?: string;\n contentType?: string | null;\n }>\n> {\n const { mandateId, taskId, payload } = input;\n const { log, config } = context;\n\n const DEFAULT_RSS_URL = \"https://feeds.npr.org/1001/rss.xml\";\n const urlToFetch = payload?.url_from_input ?? config?.rssFeedUrl ?? DEFAULT_RSS_URL;\n const maxHeadlines = Number(payload?.maxHeadlines ?? config?.maxHeadlines ?? 5);\n\n log(\"INFO\", \"FetchAgent\", `Workspaceing headlines from ${urlToFetch}, max: ${maxHeadlines}`);\n\n try {\n const resp = await fetch(urlToFetch, { headers: { \"User-Agent\": \"NewsAnalysisBot/1.0 (ValTown)\" } }); // Added User-Agent\n const fetchedContentType = resp.headers.get(\"content-type\")?.toLowerCase() || \"\";\n const rawResponseText = await resp.text();\n\n if (!resp.ok) {\n let errorBody = rawResponseText;\n if (errorBody.length > 500) errorBody = errorBody.substring(0, 500) + \"...\";\n throw new Error(`Workspace failed: ${resp.status} ${resp.statusText}. URL: ${urlToFetch}. Body: ${errorBody}`);\n }\n\n let headlines: Array<{ title: string; link: string; summary: string; publishedDate: string }> = [];\n let feedTitle = \"N/A\";\n let parsedSuccessfully = false;\n let parsingMessage: string | undefined = undefined;\n\n if (\n fetchedContentType.includes(\"xml\") || fetchedContentType.includes(\"rss\") || urlToFetch.endsWith(\".xml\")\n || urlToFetch.endsWith(\".rss\") || urlToFetch === DEFAULT_RSS_URL\n ) {\n try {\n const parser = new XMLParser({\n ignoreAttributes: false,\n htmlEntities: true,\n removeNSPrefix: true,\n });\n const feedJson = parser.parse(rawResponseText);\n\n if (feedJson.rss && feedJson.rss.channel) {\n feedTitle = typeof feedJson.rss.channel.title === \"string\"\n ? feedJson.rss.channel.title\n : JSON.stringify(feedJson.rss.channel.title);\n const items = Array.isArray(feedJson.rss.channel.item)\n ? feedJson.rss.channel.item\n : (feedJson.rss.channel.item ? [feedJson.rss.channel.item] : []);\n for (const item of items.slice(0, maxHeadlines)) {\n headlines.push({\n title: (item.title || \"No title\").toString().replace(/<[^>]+>/g, \"\").trim(),\n link: (item.link || (typeof item.link === \"object\" ? item.link[\"@_href\"] : item.link) || \"#\").toString(),\n summary: (item.description || \"\").toString(), // Raw summary, might contain HTML\n publishedDate: (item.pubDate || item.published || item.updated || \"\").toString(),\n });\n }\n parsedSuccessfully = headlines.length > 0;\n } else if (feedJson.feed) { // Atom\n feedTitle = (typeof feedJson.feed.title === \"object\" ? feedJson.feed.title[\"#text\"] : feedJson.feed.title)\n || \"N/A\";\n const entries = Array.isArray(feedJson.feed.entry)\n ? feedJson.feed.entry\n : (feedJson.feed.entry ? [feedJson.feed.entry] : []);\n for (const entry of entries.slice(0, maxHeadlines)) {\n let link = \"#\";\n if (Array.isArray(entry.link)) {\n const alternateLink = entry.link.find(l_1 => l_1[\"@_rel\"] === \"alternate\" || !l_1[\"@_rel\"]);\n link = alternateLink ? alternateLink[\"@_href\"] : (entry.link[0] ? entry.link[0][\"@_href\"] : \"#\");\n } else if (entry.link && typeof entry.link === \"object\") {\n link = entry.link[\"@_href\"] || \"#\";\n } else if (typeof entry.link === \"string\") { link = entry.link; }\n headlines.push({\n title: ((typeof entry.title === \"object\" ? entry.title[\"#text\"] : entry.title) || \"No title\").toString()\n .replace(/<[^>]+>/g, \"\").trim(),\n link: link.toString(),\n summary: ((entry.summary\n ? (typeof entry.summary === \"object\" ? entry.summary[\"#text\"] : entry.summary)\n : (entry.content ? (typeof entry.content === \"object\" ? entry.content[\"#text\"] : entry.content) : \"\"))\n || \"\").toString(),\n publishedDate: (entry.updated || entry.published || \"\").toString(),\n });\n }\n parsedSuccessfully = headlines.length > 0;\n }\n if (parsedSuccessfully)\n log(\"INFO\", \"FetchAgent\", `Successfully parsed ${headlines.length} headlines from \"${feedTitle}\".`);\n else {\n parsingMessage = \"RSS/Atom structure not as expected or no items found.\";\n log(\"WARN\", \"FetchAgent\", parsingMessage);\n }\n } catch (parseError: any) {\n parsingMessage = `Failed to parse XML/RSS content. Error: ${parseError.message}`;\n log(\"WARN\", \"FetchAgent\", `${parsingMessage} from URL: ${urlToFetch}`);\n }\n } else {\n parsingMessage =\n `Content type \"${fetchedContentType}\" is not XML/RSS. Not attempting RSS parse. Raw text will be available.`;\n log(\"INFO\", \"FetchAgent\", parsingMessage);\n }\n const outputData = {\n headlines,\n feedTitle,\n sourceUrl: urlToFetch,\n ...(parsingMessage && !parsedSuccessfully ? { message: parsingMessage } : {}),\n };\n return {\n mandateId,\n correlationId: taskId,\n payload: { data: outputData, rawText: rawResponseText, contentType: fetchedContentType },\n };\n } catch (e: any) {\n log(\"ERROR\", \"FetchAgent\", `Workspace or processing failed: ${e.message}`, e);\n return {\n mandateId,\n correlationId: taskId,\n payload: {\n data: {\n headlines: [],\n feedTitle: \"Error\",\n sourceUrl: urlToFetch,\n message: `Workspace or processing failed: ${e.message}`,\n },\n rawText: undefined,\n contentType: undefined,\n },\n error: e.message,\n };\n }\n}\n\n// Your provided combinerAgent (not directly used in the 12-step orchestrator, but kept for reference)\nexport async function combinerAgent(\n // ... (combinerAgent code as you provided, also updating its OpenAI call) ...\n input: AgentInput<{\n summary?: string;\n externalData?: {\n headlines?: Array<{ title: string; link: string; summary: string; publishedDate: string }>;\n feedTitle?: string;\n sourceUrl?: string;\n message?: string;\n };\n externalDataRawText?: string;\n externalDataContentType?: string | null;\n }>,\n context: AgentContext,\n): Promise> {\n const { mandateId, taskId, payload } = input;\n const { log } = context;\n log(\"INFO\", \"CombinerAgent\", \"Combining results\");\n try {\n const summaryText = payload.summary ?? \"N/A\";\n let externalDataDescription = \"External Data: Not Fetched/Available or not in expected format.\";\n // ... (rest of your description logic)\n if (\n payload.externalData?.headlines && Array.isArray(payload.externalData.headlines)\n && payload.externalData.headlines.length > 0\n ) {\n const headlinesArray = payload.externalData.headlines;\n const feedTitle = payload.externalData.feedTitle || \"News Feed\";\n externalDataDescription = `Recent Headlines from ${feedTitle} (Source: ${\n payload.externalData.sourceUrl || \"Unknown\"\n }):\\n`;\n headlinesArray.forEach((h, index) => {\n externalDataDescription += `${index + 1}. ${h.title}\\n`;\n if (h.link && h.link !== \"#\") externalDataDescription += ` Link: ${h.link}\\n`;\n if (h.summary) {\n const snippet = h.summary.replace(/<[^>]+>/g, \"\").substring(0, 150);\n externalDataDescription += ` Summary Snippet: ${snippet}${h.summary.length > 150 ? \"...\" : \"\"}\\n`;\n }\n });\n } else if (payload.externalData?.message) {\n externalDataDescription = `External Data Information: ${payload.externalData.message} (Source: ${\n payload.externalData.sourceUrl || \"Unknown\"\n })`;\n } else if (payload.externalData) {\n let jsonDataString = JSON.stringify(payload.externalData);\n if (jsonDataString.length > 750) jsonDataString = jsonDataString.substring(0, 750) + \"... (JSON truncated)\";\n externalDataDescription = `External Data (JSON): ${jsonDataString}`;\n } else if (payload.externalDataRawText) {\n let rawTextSnippet = payload.externalDataRawText.substring(0, 750);\n if (payload.externalDataRawText.length > 750) rawTextSnippet += \"... (text truncated)\";\n externalDataDescription = `External Data (Content-Type: ${\n payload.externalDataContentType || \"unknown\"\n }): ${rawTextSnippet}`;\n }\n\n const openai = new OpenAI(); // Using OpenAI client as per your example\n log(\"INFO\", \"CombinerAgent\", \"Combining text with OpenAI...\");\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\", // Explicitly setting a model\n messages: [\n {\n role: \"system\",\n content:\n \"You are an expert content synthesizer. Combine the provided summary and external data (often news headlines) into a coherent and informative paragraph. Highlight key insights or connections if possible.\",\n },\n {\n role: \"user\",\n content:\n `Please synthesize the following information:\\n\\nInternal Summary:\\n${summaryText}\\n\\nInformation from External Source:\\n${externalDataDescription}`,\n },\n ],\n temperature: 0.7,\n });\n const combined = completion.choices[0]?.message?.content?.trim() ?? \"Could not combine information.\";\n if (combined === \"Could not combine information.\")\n log(\"WARN\", \"CombinerAgent\", \"OpenAI did not return valid combined content.\");\n log(\"SUCCESS\", \"CombinerAgent\", \"Combined successfully\");\n return { mandateId, correlationId: taskId, payload: { combined } };\n } catch (e: any) {\n log(\"ERROR\", \"CombinerAgent\", `Combining failed: ${e.message}`, e);\n return { mandateId, correlationId: taskId, payload: { combined: \"\" }, error: e.message };\n }\n}\n\n// --- Interfaces for Ecosystem Data Objects ---\ninterface AnalysisConfig {\n topic: string;\n keywords: string[];\n suggestedNewsFeeds: { name: string; url: string }[]; // Changed from discoveredFeeds to suggested\n maxHeadlinesPerFeed: number;\n analysisDepth: \"cursory\" | \"standard\" | \"deep\";\n}\n\ninterface FetchedArticle {\n id: string;\n title: string;\n link: string;\n summary: string; // Raw summary from RSS, may contain HTML\n publishedDate: string;\n sourceFeedName: string;\n sourceFeedUrl: string;\n rawContent?: string; // Full article text, if fetched\n cleanedSummary?: string; // HTML-stripped summary for processing\n isRelevant?: boolean;\n relevanceRationale?: string;\n sentiment?: \"positive\" | \"negative\" | \"neutral\" | \"mixed\";\n sentimentScore?: number; // e.g., -1.0 to 1.0\n themes?: string[];\n entities?: Array<{ text: string; type: \"PERSON\" | \"ORG\" | \"LOC\" | \"MISC\" }>;\n}\n\ninterface WorkflowData {\n initialQuery?: string;\n config?: AnalysisConfig;\n discoveredFeeds?: { name: string; url: string }[]; // Feeds to be used\n fetchedArticles?: FetchedArticle[]; // Raw from fetch (multiple feeds)\n processedArticles?: FetchedArticle[]; // Cleaned articles\n relevantArticles?: FetchedArticle[]; // Filtered by relevance\n articlesWithSentiment?: FetchedArticle[];\n articlesWithThemes?: FetchedArticle[];\n trendReport?: any;\n anomalyReport?: any;\n insights?: string[];\n finalReport?: string;\n alertMessage?: string;\n}\n\n// --- AGENT DEFINITIONS for 12-STEP WORKFLOW ---\n\n// 1. ConfigurationAgent\nexport async function configurationAgent(\n input: AgentInput<{ userQuery: string }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { userQuery } = input.payload;\n log(\"INFO\", \"ConfigurationAgent\", `Processing query: ${userQuery}`);\n const openai = new OpenAI();\n\n const systemPrompt =\n \"You are a configuration assistant. Based on the user's query, extract the main topic, 3-5 relevant keywords, and suggest 2-3 news feed categories (e.g., 'general world news', 'technology news', 'finance news', 'health news', 'science news'). Output a JSON object with keys: 'topic' (string), 'keywords' (array of strings), 'feedCategories' (array of strings).\";\n let parsedLlmResponse = {\n topic: \"Default Topic Analysis\",\n keywords: [\"default\", \"analysis\"],\n feedCategories: [\"general world news\"],\n };\n\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: systemPrompt }, { role: \"user\", content: userQuery }],\n response_format: { type: \"json_object\" },\n temperature: 0.3,\n max_tokens: 250,\n });\n const llmResponseContent = completion.choices[0]?.message?.content?.trim();\n if (llmResponseContent) parsedLlmResponse = JSON.parse(llmResponseContent);\n } catch (e) {\n log(\n \"ERROR\",\n \"ConfigurationAgent\",\n `Failed to parse LLM config response or LLM error: ${e.message}. Using defaults.`,\n );\n }\n\n const feedMap = { // Simplified map; a real version might use a more extensive categorized list or API\n \"general world news\": { name: \"NPR News\", url: \"https://feeds.npr.org/1001/rss.xml\" },\n \"technology news\": { name: \"TechCrunch\", url: \"https://techcrunch.com/feed/\" },\n \"finance news\": { name: \"Yahoo Finance News\", url: \"https://finance.yahoo.com/news/rssindex\" },\n \"health news\": { name: \"NIH News\", url: \"https://www.nih.gov/news-events/news-releases/feed\" }, // Example\n \"science news\": { name: \"ScienceDaily\", url: \"https://www.sciencedaily.com/rss/all.xml\" }, // Example\n };\n\n const suggestedNewsFeeds = (parsedLlmResponse.feedCategories || [\"general world news\"])\n .map(category => feedMap[category.toLowerCase().trim()] || feedMap[\"general world news\"])\n .filter((value, index, self) => self.findIndex(t => t.url === value.url) === index) // Deduplicate\n .slice(0, 3); // Limit to 3 feeds for example\n\n const config: AnalysisConfig = {\n topic: parsedLlmResponse.topic,\n keywords: parsedLlmResponse.keywords,\n suggestedNewsFeeds,\n maxHeadlinesPerFeed: 5,\n analysisDepth: \"standard\",\n };\n log(\n \"SUCCESS\",\n \"ConfigurationAgent\",\n `Configuration generated for topic: ${config.topic} with ${config.suggestedNewsFeeds.length} feeds.`,\n );\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { config } };\n}\n\n// 2. SourceValidationAgent (was SourceSuggestionAgent - focuses on validating/using config's suggestions)\nexport async function sourceValidationAgent(\n input: AgentInput<{ config: AnalysisConfig }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const feeds = input.payload.config.suggestedNewsFeeds;\n log(\"INFO\", \"SourceValidationAgent\", `Validating ${feeds.length} feeds for topic: ${input.payload.config.topic}`);\n // TODO: Add actual validation if a feed URL is responsive (e.g., quick HEAD request)\n // For now, just passes them through.\n log(\"SUCCESS\", \"SourceValidationAgent\", `Using ${feeds.length} feeds.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { discoveredFeeds: feeds } };\n}\n\n// 3. ParallelFetchAgent (This agent will call the user-provided fetchAgent for each feed)\nexport async function parallelFetchAgent(\n input: AgentInput<{ feedsToFetch: { name: string; url: string }[]; maxHeadlinesPerFeed: number }>,\n context: AgentContext, // This context will be passed to the individual fetchAgent calls\n): Promise> {\n const { log } = context; // Main orchestrator log\n const { feedsToFetch, maxHeadlinesPerFeed } = input.payload;\n let allArticles: FetchedArticle[] = [];\n\n log(\"INFO\", \"ParallelFetchAgent\", `Starting parallel fetch from ${feedsToFetch.length} feeds.`);\n\n for (const feed of feedsToFetch) {\n log(\"INFO\", \"ParallelFetchAgent\", `Dispatching fetch for ${feed.name} (${feed.url})`);\n // Call the user-provided fetchAgent\n const singleFetchOutput = await fetchAgent(\n {\n mandateId: input.mandateId,\n taskId: `Workspace-${feed.name}`,\n payload: { url_from_input: feed.url, maxHeadlines: maxHeadlinesPerFeed },\n },\n {\n ...context,\n log: (level, agentName, message, ...data) =>\n context.log(level, `WorkspaceAgentInstance[${feed.name}] via ${agentName}`, message, ...data),\n }, // Pass context, maybe wrap log\n );\n\n if (singleFetchOutput.payload?.data?.headlines) {\n const transformedArticles = singleFetchOutput.payload.data.headlines.map(h => ({\n id: h.link || `${feed.url}-${h.title}-${Date.now()}`, // Ensure an ID\n title: h.title,\n link: h.link,\n summary: h.summary, // This is raw summary from RSS\n publishedDate: h.publishedDate,\n sourceFeedName: feed.name,\n sourceFeedUrl: feed.url,\n }));\n allArticles.push(...transformedArticles);\n log(\"INFO\", \"ParallelFetchAgent\", `Workspaceed ${transformedArticles.length} articles from ${feed.name}.`);\n } else {\n log(\n \"WARN\",\n \"ParallelFetchAgent\",\n `No headlines fetched or error from ${feed.name}. Error: ${\n singleFetchOutput.error || singleFetchOutput.payload?.data?.message\n }`,\n );\n }\n }\n log(\"SUCCESS\", \"ParallelFetchAgent\", `Total articles fetched from all feeds: ${allArticles.length}.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { fetchedArticles: allArticles } };\n}\n\n// 4. ArticleCleaningAgent\nexport async function articleCleaningAgent(\n input: AgentInput<{ articles: FetchedArticle[] }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n log(\"INFO\", \"ArticleCleaningAgent\", `Cleaning ${input.payload.articles.length} articles.`);\n const processedArticles = input.payload.articles.map(article => ({\n ...article,\n title: article.title.replace(/\\s+/g, \" \").trim(),\n // Basic HTML stripping from summary. A more robust stripper might be needed for complex HTML.\n cleanedSummary: article.summary.replace(/<[^>]+>/g, \" \").replace(/ /g, \" \").replace(/\\s+/g, \" \").trim(),\n }));\n log(\"SUCCESS\", \"ArticleCleaningAgent\", `Cleaned ${processedArticles.length} articles.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { processedArticles } };\n}\n\n// 5. RelevanceAssessmentAgent\nexport async function relevanceAssessmentAgent(\n input: AgentInput<{ articles: FetchedArticle[]; topic: string; keywords: string[] }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { articles, topic, keywords } = input.payload;\n const openai = new OpenAI();\n let relevantArticles: FetchedArticle[] = [];\n log(\"INFO\", \"RelevanceAssessmentAgent\", `Assessing relevance for ${articles.length} articles on topic: ${topic}`);\n\n for (const article of articles) {\n const contentToAssess = `Title: ${article.title}\\nSummary: ${article.cleanedSummary || article.summary}`;\n const prompt = `Is the following article highly relevant to the topic \"${topic}\" and related to keywords like [${\n keywords.join(\", \")\n }]? Respond with a JSON object: {\"isRelevant\": boolean, \"rationale\": \"brief explanation for relevance or irrelevance\"}.\\n\\nArticle:\\n${\n contentToAssess.substring(0, 1500)\n }`;\n let assessment = { isRelevant: false, rationale: \"LLM assessment failed or unclear.\" };\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"You are a relevance assessor. Output JSON.\" }, {\n role: \"user\",\n content: prompt,\n }],\n response_format: { type: \"json_object\" },\n temperature: 0.2,\n max_tokens: 150,\n });\n const llmResponse = completion.choices[0]?.message?.content?.trim();\n if (llmResponse) assessment = JSON.parse(llmResponse);\n } catch (e) {\n log(\n \"WARN\",\n \"RelevanceAssessmentAgent\",\n `LLM relevance assessment error for article \"${\n article.title.substring(0, 30)\n }\": ${e.message}. Defaulting to keyword check.`,\n );\n assessment.isRelevant = keywords.some(k =>\n article.title.toLowerCase().includes(k.toLowerCase())\n || (article.cleanedSummary || \"\").toLowerCase().includes(k.toLowerCase())\n );\n assessment.rationale = assessment.isRelevant ? \"Keyword match fallback.\" : \"No keyword match (fallback).\";\n }\n if (assessment.isRelevant) {\n relevantArticles.push({ ...article, isRelevant: true, relevanceRationale: assessment.rationale });\n }\n }\n log(\n \"SUCCESS\",\n \"RelevanceAssessmentAgent\",\n `Found ${relevantArticles.length} relevant articles out of ${articles.length}.`,\n );\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { relevantArticles } };\n}\n\n// 6. ContentExtractionAgent (Simplified: mainly uses cleaned summary, conceptual for full text)\nexport async function contentExtractionAgent(\n input: AgentInput<{ articles: FetchedArticle[]; analysisDepth: \"cursory\" | \"standard\" | \"deep\" }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { articles, analysisDepth } = input.payload;\n log(\"INFO\", \"ContentExtractionAgent\", `Content extraction. Depth: ${analysisDepth}. Articles: ${articles.length}`);\n // Simulating full content: For \"deep\" analysis, one might try to fetch article.link.\n // Here, we'll just ensure `rawContent` is populated from `cleanedSummary` for LLM processing consistency.\n const updatedArticles = articles.map(article => ({\n ...article,\n // Max 4000 chars for rawContent to avoid excessive token usage later.\n rawContent: (article.cleanedSummary || article.summary).substring(0, 4000),\n }));\n log(\"SUCCESS\", \"ContentExtractionAgent\", `Content prepared for ${updatedArticles.length} articles.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { articlesWithContent: updatedArticles } };\n}\n\n// 7. SentimentAnalysisAgent\nexport async function sentimentAnalysisAgent(\n input: AgentInput<{ articles: FetchedArticle[] }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const openai = new OpenAI();\n let articlesWithSentiment: FetchedArticle[] = [];\n if (!input.payload.articles)\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { articlesWithSentiment } };\n log(\"INFO\", \"SentimentAnalysisAgent\", `Analyzing sentiment for ${input.payload.articles.length} articles.`);\n\n for (const article of input.payload.articles) {\n const textToAnalyze = (article.rawContent || article.cleanedSummary || \"\").substring(0, 3000);\n const prompt =\n `Analyze the sentiment of the following text. Respond with a JSON object: {\"sentiment\": \"positive\" | \"negative\" | \"neutral\" | \"mixed\", \"score\": float_between_-1_and_1 (optional, e.g. 0.7 for positive)}.\\n\\nText:\\n${textToAnalyze}`;\n let sentimentResult: FetchedArticle[\"sentiment\"] = \"neutral\";\n let sentimentScore: FetchedArticle[\"sentimentScore\"] = 0.0;\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"You are a sentiment analysis expert. Output JSON.\" }, {\n role: \"user\",\n content: prompt,\n }],\n response_format: { type: \"json_object\" },\n temperature: 0.3,\n max_tokens: 100,\n });\n const llmResponse = completion.choices[0]?.message?.content?.trim();\n if (llmResponse) {\n const parsed = JSON.parse(llmResponse);\n sentimentResult = parsed.sentiment;\n sentimentScore = parsed.score;\n }\n } catch (e) {\n log(\n \"WARN\",\n \"SentimentAnalysisAgent\",\n `LLM sentiment error for article \"${article.title.substring(0, 30)}\": ${e.message}`,\n );\n }\n articlesWithSentiment.push({ ...article, sentiment: sentimentResult, sentimentScore });\n }\n log(\"SUCCESS\", \"SentimentAnalysisAgent\", `Analyzed sentiment for ${articlesWithSentiment.length} articles.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { articlesWithSentiment } };\n}\n\n// 8. KeyThemeExtractionAgent\nexport async function keyThemeExtractionAgent(\n input: AgentInput<{ articles: FetchedArticle[]; topic: string }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const openai = new OpenAI();\n let articlesWithThemes: FetchedArticle[] = [];\n log(\n \"INFO\",\n \"KeyThemeExtractionAgent\",\n `Extracting themes for ${input.payload.articles.length} articles related to ${input.payload.topic}.`,\n );\n\n for (const article of input.payload.articles) {\n const textToAnalyze = (article.rawContent || article.cleanedSummary || \"\").substring(0, 3000);\n const prompt =\n `For the text (topic: \"${input.payload.topic}\"), identify 2-4 key themes and prominent named entities (people, orgs). JSON: {\"themes\": [\"theme1\", ...], \"entities\": [{\"text\": \"EntityName\", \"type\": \"PERSON\"|\"ORG\"|\"LOC\"|\"MISC\"}]}.\\nText:\\n${textToAnalyze}`;\n let themeResult: { themes: string[]; entities: FetchedArticle[\"entities\"] } = { themes: [], entities: [] };\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"You are a thematic/NER expert. Output JSON.\" }, {\n role: \"user\",\n content: prompt,\n }],\n response_format: { type: \"json_object\" },\n temperature: 0.4,\n max_tokens: 250,\n });\n const llmResponse = completion.choices[0]?.message?.content?.trim();\n if (llmResponse) themeResult = JSON.parse(llmResponse);\n } catch (e) {\n log(\n \"WARN\",\n \"KeyThemeExtractionAgent\",\n `LLM theme error for article \"${article.title.substring(0, 30)}\": ${e.message}`,\n );\n }\n articlesWithThemes.push({ ...article, themes: themeResult.themes, entities: themeResult.entities });\n }\n log(\"SUCCESS\", \"KeyThemeExtractionAgent\", `Extracted themes for ${articlesWithThemes.length} articles.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { articlesWithThemes } };\n}\n\n// 9. TrendAndAnomalyDetectionAgent\nexport async function trendAndAnomalyDetectionAgent(\n input: AgentInput<{ articlesWithThemes: FetchedArticle[]; topic: string; historicalContextSummary?: string }>, // historicalContext is optional\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { articlesWithThemes, topic, historicalContextSummary } = input.payload;\n const openai = new OpenAI();\n log(\n \"INFO\",\n \"TrendAndAnomalyDetectionAgent\",\n `Analyzing trends/anomalies for ${articlesWithThemes.length} articles on: ${topic}.`,\n );\n\n const themesCounter: { [key: string]: number } = {};\n const sentimentCounter: { [key: string]: number } = { positive: 0, negative: 0, neutral: 0, mixed: 0 };\n articlesWithThemes.forEach(article => {\n (article.themes || []).forEach(theme => themesCounter[theme] = (themesCounter[theme] || 0) + 1);\n if (article.sentiment) sentimentCounter[article.sentiment] = (sentimentCounter[article.sentiment] || 0) + 1;\n });\n const topThemes = Object.entries(themesCounter).sort((a, b) => b[1] - a[1]).slice(0, 5);\n const aggregatedDataSummary = `Analyzed ${articlesWithThemes.length} articles. Top themes: ${\n topThemes.map(t => `${t[0]} (${t[1]})`).join(\", \")\n }. Sentiment: Positive(${sentimentCounter.positive}), Negative(${sentimentCounter.negative}), Neutral(${sentimentCounter.neutral}), Mixed(${sentimentCounter.mixed}).`;\n\n const trendPrompt = `Topic: \"${topic}\". Data: ${aggregatedDataSummary.substring(0, 1000)}. ${\n historicalContextSummary ? `Historical: \"${historicalContextSummary.substring(0, 500)}\"` : \"\"\n } Identify dominant sentiment, key themes, and emerging patterns. JSON: {\"dominantSentiment\": \"string\", \"keyThemes\": [\"theme1\", ...], \"emergingPatterns\": \"description\"}`;\n let trendReport: any = { dominantSentiment: \"N/A\", keyThemes: [], emergingPatterns: \"Not analyzed.\" };\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"Trend analysis expert. Output JSON.\" }, {\n role: \"user\",\n content: trendPrompt,\n }],\n response_format: { type: \"json_object\" },\n temperature: 0.5,\n max_tokens: 400,\n });\n const response = completion.choices[0]?.message?.content?.trim();\n if (response) trendReport = JSON.parse(response);\n } catch (e) {\n log(\"ERROR\", \"TrendAndAnomalyDetectionAgent\", `Trend report LLM error: ${e.message}`);\n }\n\n const anomalyPrompt = `Topic: \"${topic}\". Current data: ${aggregatedDataSummary.substring(0, 1000)}. ${\n historicalContextSummary ? `Historical: \"${historicalContextSummary.substring(0, 500)}\"` : \"\"\n } Are there surprising findings or deviations? Describe up to 2 anomalies. JSON: {\"anomaliesFound\": boolean, \"anomalyDetails\": [{\"description\": \"string\", \"severity\": \"low|medium|high\"}]}`;\n let anomalyReport: any = { anomaliesFound: false, anomalyDetails: [] };\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"Anomaly detection expert. Output JSON.\" }, {\n role: \"user\",\n content: anomalyPrompt,\n }],\n response_format: { type: \"json_object\" },\n temperature: 0.6,\n max_tokens: 400,\n });\n const response = completion.choices[0]?.message?.content?.trim();\n if (response) anomalyReport = JSON.parse(response);\n } catch (e) {\n log(\"ERROR\", \"TrendAndAnomalyDetectionAgent\", `Anomaly report LLM error: ${e.message}`);\n }\n\n log(\"SUCCESS\", \"TrendAndAnomalyDetectionAgent\", \"Trend/anomaly analysis complete.\");\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { trendReport, anomalyReport } };\n}\n\n// 10. InsightGenerationAgent\nexport async function insightGenerationAgent(\n input: AgentInput<{ trendReport: any; anomalyReport: any; config: AnalysisConfig; articlesCount: number }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { trendReport, anomalyReport, config, articlesCount } = input.payload;\n const openai = new OpenAI();\n log(\"INFO\", \"InsightGenerationAgent\", `Generating insights for topic: ${config.topic}`);\n const contextSummary =\n `Analysis of ${articlesCount} articles on \"${config.topic}\". Trends: Dominant sentiment ${trendReport.dominantSentiment}, Key themes ${\n trendReport.keyThemes?.join(\", \") || \"N/A\"\n }. Anomalies found: ${anomalyReport.anomaliesFound}. Details: ${\n anomalyReport.anomalyDetails?.map(d => d.description).join(\"; \") || \"N/A\"\n }`;\n const prompt = `Context: ${\n contextSummary.substring(0, 1500)\n }. What are 2-3 key strategic insights or actionable takeaways? Be concise.`;\n let insights: string[] = [\"No specific insights generated.\"];\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"Strategic analyst providing concise insights.\" }, {\n role: \"user\",\n content: prompt,\n }],\n temperature: 0.7,\n max_tokens: 500,\n });\n const response = completion.choices[0]?.message?.content?.trim();\n if (response) insights = response.split(\"\\n\").map(i => i.replace(/^- |^\\* |^\\d+\\. /, \"\").trim()).filter(Boolean);\n if (insights.length === 0 && response) insights = [response];\n } catch (e) {\n log(\"ERROR\", \"InsightGenerationAgent\", `Insight LLM error: ${e.message}`);\n }\n log(\"SUCCESS\", \"InsightGenerationAgent\", `Generated ${insights.length} insights.`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { insights } };\n}\n\n// Define an interface for the flat structure of inputs expected by this agent,\n// matching the 'input' mapping in the workflow definition for step4_compileReport.\n// Also, ensure AnalysisConfig and other types are appropriately defined or imported.\ninterface AnalysisConfig {\n topic: string;\n keywords: string[];\n // ... other properties of config\n}\n\ninterface ArticleForReport {\n title: string;\n sentiment?: string;\n // ... other properties if needed\n}\n\ninterface DiscoveredFeed {\n name: string;\n // ... other properties\n}\n\ninterface TrendReportData {\n dominantSentiment?: string;\n keyThemes?: string[];\n emergingPatterns?: string;\n // ... other properties\n}\n\ninterface AnomalyDetail {\n description: string;\n severity: string;\n // ... other properties\n}\n\ninterface AnomalyReportData {\n anomaliesFound: boolean;\n anomalyDetails?: AnomalyDetail[];\n // ... other properties\n}\n\n// This should match the keys provided by the workflow step's input mapping\ninterface ReportCompilationAgentPayload {\n config: AnalysisConfig;\n discoveredFeeds?: DiscoveredFeed[];\n articlesWithThemes?: ArticleForReport[]; // Changed from relevantArticles to match typical workflow output\n trendReport?: TrendReportData;\n anomalyReport?: AnomalyReportData;\n insights?: string[];\n // Add userTextSummaryResult and userTextSentimentResult if the agent needs them\n // userTextSummaryResult?: string;\n // userTextSentimentResult?: any; // Define more specific type if possible\n}\n\n// Assuming AgentInput and AgentContext and AgentOutput are defined elsewhere\ninterface AgentInput {\n mandateId: string;\n taskId: string;\n payload: T;\n}\ninterface AgentContext {\n log: (level: string, component: string, message: string, details?: any) => void;\n // ... other context properties\n}\ninterface AgentOutput {\n mandateId: string;\n correlationId: string;\n payload: T;\n}\n\n// 11. ReportCompilationAgent\nexport async function reportCompilationAgent(\n input: AgentInput, // Use the new flat payload type\n context: AgentContext,\n): Promise> {\n const { log } = context;\n // Destructure directly from input.payload, which now holds the flat structure\n const {\n config,\n discoveredFeeds,\n articlesWithThemes, // Using this name, ensure it matches workflow output used for articles\n trendReport,\n anomalyReport,\n insights,\n } = input.payload;\n\n // CRITICAL CHECK: Ensure 'config' is present, as it's not optional in the workflow step.\n if (!config) {\n const errorMessage = \"CRITICAL: 'config' object not found in input payload for ReportCompilationAgent.\";\n log(\"ERROR\", \"ReportCompilationAgent\", errorMessage);\n return {\n mandateId: input.mandateId,\n correlationId: input.taskId,\n payload: { finalReport: `Report generation failed: ${errorMessage}` },\n };\n }\n\n log(\"INFO\", \"ReportCompilationAgent\", `Compiling report for topic: ${config.topic}`); // Directly use 'config'\n const openai = new OpenAI(); // Ensure API key is configured via environment variables or client options\n\n // The agent code used 'relevantArticles'. If 'articlesWithThemes' is the correct data source\n // from the workflow, use it. If 'relevantArticles' is a distinct piece of data,\n // ensure it's mapped in the workflow's input for this step.\n // Assuming 'articlesWithThemes' is what was intended for \"relevant articles sample\":\n const relevantArticlesSample =\n articlesWithThemes?.slice(0, 3).map(a => `- ${a.title} (Sentiment: ${a.sentiment || \"N/A\"})`).join(\"\\n\")\n || \"None selected for report.\";\n const articlesAnalyzedCount = articlesWithThemes?.length || 0;\n\n const prompt = `Generate a markdown analysis report for topic \"${config.topic}\".\nKeywords: ${config.keywords?.join(\", \") || \"N/A\"}. Date: ${new Date().toLocaleDateString()}.\nSources: ${discoveredFeeds?.map(f => f.name).join(\", \") || \"N/A\"}. Articles Analyzed: ${articlesAnalyzedCount}.\nExecutive Summary (incorporate insights: ${insights?.join(\"; \") || \"N/A\"}): [LLM to fill]\nKey Relevant Articles (Sample):\n${relevantArticlesSample}\nTrend Analysis: Dominant Sentiment: ${trendReport?.dominantSentiment || \"N/A\"}. Key Themes: ${\n trendReport?.keyThemes?.join(\", \") || \"N/A\"\n }. Emerging Patterns: ${trendReport?.emergingPatterns || \"N/A\"}.\nAnomalies: ${anomalyReport?.anomaliesFound ? \"Yes\" : \"No\"}. Details: ${\n anomalyReport?.anomalyDetails?.map(d => `- ${d.description} (Severity: ${d.severity})`).join(\"\\n\")\n || \"None reported.\"\n }\nKey Insights:\n${insights?.map(i => `- ${i}`).join(\"\\n\") || \"No specific insights.\"}\nStructure the report professionally with clear sections.\n`;\n let finalReport = \"Report generation failed.\";\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [\n { role: \"system\", content: \"Professional report writer using markdown.\" },\n { role: \"user\", content: prompt.substring(0, 15000) }, // Ensure prompt isn't excessively long\n ],\n temperature: 0.5,\n max_tokens: 3000,\n });\n finalReport = completion.choices[0]?.message?.content?.trim() || \"No report content generated by LLM.\";\n } catch (e: any) { // Catch specific error type if possible\n log(\"ERROR\", \"ReportCompilationAgent\", `Report LLM error: ${e.message}`, e);\n finalReport = `Report generation encountered an error: ${e.message}`;\n }\n log(\"SUCCESS\", \"ReportCompilationAgent\", \"Final report compiled.\");\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { finalReport } };\n}\n\n// Ensure the insightGenerationAgent is also consistent with its input expectations\n// The insightGenerationAgent seems fine as it destructures its expected inputs directly from input.payload.\n// export async function insightGenerationAgent(\n// input: AgentInput<{ trendReport: any, anomalyReport: any, config: AnalysisConfig, articlesCount: number }>,\n// ...)\n// 12. AlertFormattingAgent\nexport async function alertFormattingAgent(\n input: AgentInput<{ anomalyReport: any; insights: string[]; config: AnalysisConfig }>,\n context: AgentContext,\n): Promise> {\n const { log } = context;\n const { anomalyReport, insights, config } = input.payload;\n const openai = new OpenAI();\n let criticalInfo = \"\";\n if (anomalyReport?.anomaliesFound && anomalyReport.anomalyDetails.length > 0) {\n criticalInfo += `Anomalies for \"${config.topic}\": ${\n anomalyReport.anomalyDetails.map(d => `${d.description}(${d.severity})`).join(\"; \")\n }. `;\n }\n const potentInsights = (insights || []).filter(i => i && i.toLowerCase() !== \"no specific insights generated.\");\n if (potentInsights.length > 0) criticalInfo += `Key Insight: ${potentInsights[0]}`;\n\n if (!criticalInfo.trim()) {\n log(\"INFO\", \"AlertFormattingAgent\", \"No critical info for alert.\");\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { alertMessage: undefined } };\n }\n log(\"INFO\", \"AlertFormattingAgent\", `Formatting alert for: ${config.topic}`);\n const prompt = `Generate a concise alert (Subject & Body) for: ${\n criticalInfo.substring(0, 1000)\n }. Format as: Subject: [Alert Subject]\\nBody:\\n[Alert Body]`;\n let alertMessage: string | undefined = undefined;\n try {\n const completion = await openai.chat.completions.create({\n model: \"gpt-4o-mini\",\n messages: [{ role: \"system\", content: \"Alert generation specialist.\" }, { role: \"user\", content: prompt }],\n temperature: 0.4,\n max_tokens: 300,\n });\n alertMessage = completion.choices[0]?.message?.content?.trim();\n } catch (e) {\n log(\"ERROR\", \"AlertFormattingAgent\", `Alert LLM error: ${e.message}`);\n }\n log(\"SUCCESS\", \"AlertFormattingAgent\", `Alert message formatted: ${alertMessage ? \"Yes\" : \"No\"}`);\n return { mandateId: input.mandateId, correlationId: input.taskId, payload: { alertMessage } };\n}\n\n// --- Orchestrator Agent ---\nexport async function analysisWorkflowOrchestrator(\n initialUserQuery: string,\n baseContext?: Partial, // Allow passing a base context, e.g., for top-level config/secrets\n): Promise {\n const mandateId = \"workflow-\" + Date.now();\n const defaultLog = (level, agentName, message, ...data) =>\n console.log(`[${level}] (${mandateId}) ${agentName}: ${message}`, data.length > 0 ? data : \"\");\n const context: AgentContext = {\n log: baseContext?.log || defaultLog,\n config: baseContext?.config || {},\n secrets: baseContext?.secrets || { get: async (key) => Deno.env.get(key) }, // Basic secrets access\n };\n\n let workflowData: WorkflowData = { initialQuery: initialUserQuery };\n context.log(\"INFO\", \"Orchestrator\", `Starting workflow for query: \"${initialUserQuery}\"`);\n\n try {\n const configOut = await configurationAgent({\n mandateId,\n taskId: \"config-1\",\n payload: { userQuery: initialUserQuery },\n }, context);\n workflowData.config = configOut.payload.config;\n if (configOut.error) throw new Error(`ConfigAgent Error: ${configOut.error}`);\n\n const sourcesOut = await sourceValidationAgent({\n mandateId,\n taskId: \"sources-2\",\n payload: { config: workflowData.config! },\n }, context);\n workflowData.discoveredFeeds = sourcesOut.payload.discoveredFeeds;\n if (sourcesOut.error) throw new Error(`SourceValidationAgent Error: ${sourcesOut.error}`);\n\n const pFetchOut = await parallelFetchAgent({\n mandateId,\n taskId: \"pfetch-3\",\n payload: {\n feedsToFetch: workflowData.discoveredFeeds!,\n maxHeadlinesPerFeed: workflowData.config!.maxHeadlinesPerFeed,\n },\n }, context);\n workflowData.fetchedArticles = pFetchOut.payload.fetchedArticles;\n if (pFetchOut.error) throw new Error(`ParallelFetchAgent Error: ${pFetchOut.error}`);\n\n const cleanOut = await articleCleaningAgent({\n mandateId,\n taskId: \"clean-4\",\n payload: { articles: workflowData.fetchedArticles! },\n }, context);\n workflowData.processedArticles = cleanOut.payload.processedArticles;\n if (cleanOut.error) throw new Error(`ArticleCleaningAgent Error: ${cleanOut.error}`);\n\n const relevanceOut = await relevanceAssessmentAgent({\n mandateId,\n taskId: \"relevance-5\",\n payload: {\n articles: workflowData.processedArticles!,\n topic: workflowData.config!.topic,\n keywords: workflowData.config!.keywords,\n },\n }, context);\n workflowData.relevantArticles = relevanceOut.payload.relevantArticles;\n if (relevanceOut.error) throw new Error(`RelevanceAssessmentAgent Error: ${relevanceOut.error}`);\n\n const contentExtOut = await contentExtractionAgent({\n mandateId,\n taskId: \"content-6\",\n payload: { articles: workflowData.relevantArticles!, analysisDepth: workflowData.config!.analysisDepth },\n }, context);\n workflowData.relevantArticles = contentExtOut.payload.articlesWithContent;\n if (contentExtOut.error) throw new Error(`ContentExtractionAgent Error: ${contentExtOut.error}`);\n\n const sentimentOut = await sentimentAnalysisAgent({\n mandateId,\n taskId: \"sentiment-7\",\n payload: { articles: workflowData.relevantArticles! },\n }, context);\n workflowData.articlesWithSentiment = sentimentOut.payload.articlesWithSentiment;\n if (sentimentOut.error) throw new Error(`SentimentAnalysisAgent Error: ${sentimentOut.error}`);\n\n const themesOut = await keyThemeExtractionAgent({\n mandateId,\n taskId: \"themes-8\",\n payload: { articles: workflowData.articlesWithSentiment!, topic: workflowData.config!.topic },\n }, context);\n workflowData.articlesWithThemes = themesOut.payload.articlesWithThemes;\n if (themesOut.error) throw new Error(`KeyThemeExtractionAgent Error: ${themesOut.error}`);\n\n const trendAnomalyOut = await trendAndAnomalyDetectionAgent({\n mandateId,\n taskId: \"trendanomaly-9\",\n payload: {\n articlesWithThemes: workflowData.articlesWithThemes!,\n topic: workflowData.config!.topic,\n historicalContextSummary: undefined,\n },\n }, context); // Historical context can be added\n workflowData.trendReport = trendAnomalyOut.payload.trendReport;\n workflowData.anomalyReport = trendAnomalyOut.payload.anomalyReport;\n if (trendAnomalyOut.error) throw new Error(`TrendAndAnomalyDetectionAgent Error: ${trendAnomalyOut.error}`);\n\n const insightsOut = await insightGenerationAgent({\n mandateId,\n taskId: \"insights-10\",\n payload: {\n trendReport: workflowData.trendReport,\n anomalyReport: workflowData.anomalyReport,\n config: workflowData.config!,\n articlesCount: workflowData.relevantArticles!.length,\n },\n }, context);\n workflowData.insights = insightsOut.payload.insights;\n if (insightsOut.error) throw new Error(`InsightGenerationAgent Error: ${insightsOut.error}`);\n\n const reportOut = await reportCompilationAgent(\n { mandateId, taskId: \"report-11\", payload: { workflowData } },\n context,\n );\n workflowData.finalReport = reportOut.payload.finalReport;\n if (reportOut.error) throw new Error(`ReportCompilationAgent Error: ${reportOut.error}`);\n\n const alertOut = await alertFormattingAgent({\n mandateId,\n taskId: \"alert-12\",\n payload: {\n anomalyReport: workflowData.anomalyReport,\n insights: workflowData.insights!,\n config: workflowData.config!,\n },\n }, context);\n workflowData.alertMessage = alertOut.payload.alertMessage;\n if (alertOut.error) throw new Error(`AlertFormattingAgent Error: ${alertOut.error}`);\n\n context.log(\"SUCCESS\", \"Orchestrator\", \"Workflow completed successfully.\");\n } catch (error) {\n context.log(\"CRITICAL\", \"Orchestrator\", `Workflow failed: ${error.message}`, error);\n }\n return workflowData;\n}","deleted","label",[98],"permalink","https://www.val.town/x/salon/mandate/code/agent_ecosystem.ts","extension","actionData","errors"]