@mgsoftwarebv/mg-dashboard-mcp 7.0.23 → 7.1.0

package/dist/index.js

@@ -8208,7 +8208,11 @@ function formatBytes(bytes) {
 }
 var RESPONSE_MAX_BYTES = 8192;
 var NO_FOOTER_TOOLS = /* @__PURE__ */ new Set();
-var RAW_JSON_TOOLS = /* @__PURE__ */ new Set(["get_mg_dashboard_commits"]);
+var RAW_JSON_TOOLS = /* @__PURE__ */ new Set([
+  "get_mg_dashboard_commits",
+  "extract_article",
+  "research_topic"
+]);
 var TOOL_CACHE_TTL_MS = {
   "list-servers": 6e4,
   "docker-list": 3e4
@@ -10356,6 +10360,66 @@ var TOOLS = [
       required: ["actor", "dateFrom", "dateTo"]
     }
   },
+  {
+    name: "extract_article",
+    description: "Extract readable article/content data from ONE PUBLIC URL, with a browser-render fallback for pages that block a normal fetch (403) or only render with JavaScript (ticket 2026-DASMG-039). Returns title, author, publishedAt, source, summary, facts[], optional rawText (capped), canonicalUrl, the `extractionMethod` used (http | browser | jsonld | rss | amp | metadata_only) and an explicit `limitations` array (e.g. blocked_without_browser, paywall_detected, partial_content, rawtext_truncated). PUBLIC URLs only \u2014 SSRF-guarded (private/loopback/metadata addresses rejected), no paywall bypass, no credentials; the source/canonical URL is always preserved. Use it to turn a news/article link into safe factual source material for agents; it never auto-publishes or rewrites. For several sources + cross-checking, use research_topic.",
+    inputSchema: {
+      type: "object",
+      properties: {
+        url: {
+          type: "string",
+          description: "Public http(s) URL of the article/page to extract."
+        },
+        includeRawText: {
+          type: "boolean",
+          description: "Include the readable body as `rawText` (default true). Set false for a compact metadata-only result."
+        },
+        maxRawTextChars: {
+          type: "number",
+          description: "Hard cap for `rawText` length (default 20000, max 100000). Truncation adds a `rawtext_truncated` limitation."
+        }
+      },
+      required: ["url"]
+    }
+  },
+  {
+    name: "research_topic",
+    description: "Turn SEVERAL public sources on one topic into a single cross-checked, sourced 'evidence pack' (ticket 2026-DASMG-040). Composes extract_article per URL and aggregates: collect, extract, cross-check, then hand a downstream agent a safe base to write ORIGINAL content from. Modes: curated (`urls` + optional `topic`), hybrid (`urls` + `discover`), discovery (`topic` + `discover.enabled`). Discovery is currently DEFERRED \u2014 when requested it returns a `discovery_unavailable` limitation while curated `urls` still work. Returns `sources` (id,url,canonicalUrl,source,title,publishedAt,authorityTier,extractionMethod,limitations), `verifiedFacts` (claim + corroboratedBy + confidence + authorityTier + asOf), `conflicts` (variants + suggestedResolution), `uniqueClaims` (single-source), `openQuestions`, `verbatimWarnings`, and pack-level `limitations`. Structured EVIDENCE ONLY \u2014 never ready-to-publish prose or brand voice; every claim traces back to a source id. PUBLIC URLs only (inherits the SSRF guard). Persistence/reuse is out of scope (2026-DASMG-041).",
+    inputSchema: {
+      type: "object",
+      properties: {
+        topic: {
+          type: "string",
+          description: "The topic the evidence pack is about (used for context/labelling)."
+        },
+        urls: {
+          type: "array",
+          items: { type: "string" },
+          description: "Curated list of public http(s) URLs to extract and cross-check (max 20)."
+        },
+        discover: {
+          type: "object",
+          description: "Discovery options (currently deferred). enabled=true emits a discovery_unavailable limitation; curated urls still work.",
+          properties: {
+            enabled: { type: "boolean" },
+            maxSources: { type: "number" },
+            recencyDays: { type: "number" },
+            locale: { type: "string" },
+            allowDomains: { type: "array", items: { type: "string" } },
+            blockDomains: { type: "array", items: { type: "string" } }
+          }
+        },
+        includeRawText: {
+          type: "boolean",
+          description: "Reserved for compatibility; claim mining always reads body text up to maxRawTextChars."
+        },
+        maxRawTextChars: {
+          type: "number",
+          description: "Per-source body cap used for claim mining (default 4000, max 20000)."
+        }
+      }
+    }
+  },
   {
     name: "search-team-memory",
     description: "Search the team's ENTIRE past Cursor history (every developer, every project) for how something was handled before. Use this FIRST, before investigating from scratch, whenever you: hit a non-trivial bug or error, are about to build something that may have been done before, need a project-specific convention/gotcha, or the user asks 'have we done X', 'how did we fix Y', or 'did we solve this already'. Hybrid semantic + keyword search over mirrored conversations. Returns ranked past chats with repo, date, a solution snippet and a similarity score. Phrase the query as the problem in natural language (e.g. 'release pipeline PM2 deploy fails with module not found').",
@@ -10585,6 +10649,80 @@ async function executeToolCall(name, a, _serverId) {
           content: [{ type: "text", text: JSON.stringify(data, null, 2) }]
         };
       }
+      // ----- Public-web extraction (article tools) -----
+      case "extract_article": {
+        const url = typeof a.url === "string" ? a.url.trim() : "";
+        if (!url) {
+          return { content: [{ type: "text", text: "Error: url is required" }] };
+        }
+        const res = await fetch(`${dashboardBaseUrl}/api/tools/extract-article`, {
+          method: "POST",
+          headers: {
+            "content-type": "application/json",
+            authorization: `Bearer ${apiKey}`
+          },
+          body: JSON.stringify({
+            url,
+            includeRawText: typeof a.includeRawText === "boolean" ? a.includeRawText : void 0,
+            maxRawTextChars: typeof a.maxRawTextChars === "number" ? a.maxRawTextChars : void 0
+          })
+        });
+        if (!res.ok) {
+          const detail = await res.text().catch(() => "");
+          return {
+            content: [
+              {
+                type: "text",
+                text: `Error: extract_article failed (${res.status}). ${detail.slice(0, 300)}`
+              }
+            ]
+          };
+        }
+        const data = await res.json();
+        return {
+          content: [{ type: "text", text: JSON.stringify(data, null, 2) }]
+        };
+      }
+      case "research_topic": {
+        const urls = Array.isArray(a.urls) ? a.urls.filter((u) => typeof u === "string") : void 0;
+        const topic = typeof a.topic === "string" ? a.topic.trim() : void 0;
+        if ((!urls || urls.length === 0) && !topic) {
+          return {
+            content: [
+              { type: "text", text: "Error: provide `urls` (curated) and/or a `topic`" }
+            ]
+          };
+        }
+        const res = await fetch(`${dashboardBaseUrl}/api/tools/research-topic`, {
+          method: "POST",
+          headers: {
+            "content-type": "application/json",
+            authorization: `Bearer ${apiKey}`
+          },
+          body: JSON.stringify({
+            topic,
+            urls,
+            discover: a.discover && typeof a.discover === "object" ? a.discover : void 0,
+            includeRawText: typeof a.includeRawText === "boolean" ? a.includeRawText : void 0,
+            maxRawTextChars: typeof a.maxRawTextChars === "number" ? a.maxRawTextChars : void 0
+          })
+        });
+        if (!res.ok) {
+          const detail = await res.text().catch(() => "");
+          return {
+            content: [
+              {
+                type: "text",
+                text: `Error: research_topic failed (${res.status}). ${detail.slice(0, 300)}`
+              }
+            ]
+          };
+        }
+        const data = await res.json();
+        return {
+          content: [{ type: "text", text: JSON.stringify(data, null, 2) }]
+        };
+      }
       // ----- Team memory -----
       case "search-team-memory": {
         const query = typeof a.query === "string" ? a.query.trim() : "";