@staticn0va/wigolo 0.1.0 → 0.1.2

package/dist/search/answer-synthesis.js

@@ -0,0 +1,244 @@
+import {
+  checkSamplingSupport,
+  requestSampling,
+  extractTextFromSamplingResponse
+} from "./sampling.js";
+import { createLogger } from "../logger.js";
+const log = createLogger("search");
+const MAX_CHARS_PER_SOURCE = 3e3;
+const MAX_RESPONSE_TOKENS = 1500;
+const FALLBACK_MAX_BULLETS = 5;
+const FALLBACK_KEYPOINT_MAX_CHARS = 240;
+async function synthesizeAnswer(results, query, server) {
+  try {
+    const sourcesText = buildSourcesText(results);
+    if (!sourcesText) {
+      log.info("no content available for synthesis");
+      return {
+        fallback: true,
+        warning: "No results with content available for answer synthesis"
+      };
+    }
+    if (!checkSamplingSupport(server)) {
+      log.info("sampling not supported by client, falling back to context format");
+      return {
+        fallback: true,
+        warning: "Client does not support MCP sampling; falling back to context format"
+      };
+    }
+    const prompt = buildSynthesisPrompt(query, sourcesText);
+    const response = await requestSampling(
+      server,
+      [{ role: "user", content: { type: "text", text: prompt } }],
+      MAX_RESPONSE_TOKENS
+    );
+    const answerText = extractTextFromSamplingResponse(response);
+    if (!answerText) {
+      log.warn("sampling returned empty response");
+      return {
+        fallback: true,
+        warning: "Sampling returned empty response; falling back to context format"
+      };
+    }
+    const citations = extractCitations(answerText, results);
+    log.info("answer synthesis complete", {
+      answerLength: answerText.length,
+      citationCount: citations.length
+    });
+    return {
+      answer: answerText,
+      citations,
+      fallback: false
+    };
+  } catch (err) {
+    log.error("answer synthesis failed", { error: String(err) });
+    return {
+      fallback: true,
+      warning: `Answer synthesis failed: ${err instanceof Error ? err.message : String(err)}; falling back to context format`
+    };
+  }
+}
+function buildSourcesText(results) {
+  try {
+    const blocks = [];
+    let sourceIndex = 1;
+    for (const result of results) {
+      const content = result.markdown_content || result.snippet || "";
+      if (!content.trim()) continue;
+      const truncated = content.length > MAX_CHARS_PER_SOURCE ? content.slice(0, MAX_CHARS_PER_SOURCE) : content;
+      blocks.push(`[${sourceIndex}] ${result.title} (${result.url})
+${truncated}`);
+      sourceIndex++;
+    }
+    if (blocks.length === 0) return "";
+    return blocks.join("\n\n---\n\n");
+  } catch (err) {
+    log.error("buildSourcesText failed", { error: String(err) });
+    return "";
+  }
+}
+function buildSynthesisPrompt(query, sourcesText) {
+  return `Based on the following sources, provide a concise and direct answer to the question. Use numbered citations like [1], [2] to reference specific sources.
+
+Question: ${query}
+
+Sources:
+${sourcesText}
+
+Instructions:
+- Be concise and direct. Answer in 2-4 paragraphs maximum.
+- Cite sources using [1], [2], etc. matching the source numbers above.
+- If sources contain conflicting information, note the discrepancy.
+- If the sources don't adequately answer the question, say so.
+- Do not include information not found in the provided sources.`;
+}
+function buildStructuredFallback(results, query) {
+  const bullets = [];
+  const citations = [];
+  let n = 0;
+  for (const r of results) {
+    if (n >= FALLBACK_MAX_BULLETS) break;
+    const body = r.markdown_content && r.markdown_content.trim() || r.snippet && r.snippet.trim() || "";
+    if (!body) continue;
+    const keypoint = extractKeypoint(body, FALLBACK_KEYPOINT_MAX_CHARS);
+    if (!keypoint) continue;
+    n += 1;
+    bullets.push(`- **${r.title}** \u2014 ${keypoint} [${n}]`);
+    citations.push({ index: n, url: r.url, title: r.title, snippet: r.snippet });
+  }
+  if (bullets.length === 0) {
+    return { answer: "", citations: [], warning: "No sampling server available; no content to summarize" };
+  }
+  const q = query && query.trim() ? query.trim() : "this query";
+  const answer = `Based on the top ${bullets.length} sources for "${q}":
+
+${bullets.join("\n")}
+
+Sources:
+${citations.map((c) => `[${c.index}] ${c.title} \u2014 ${c.url}`).join("\n")}`;
+  return {
+    answer,
+    citations,
+    warning: "Client does not support MCP sampling; returning heuristic key-point summary instead of synthesized answer"
+  };
+}
+function extractKeypoint(body, maxChars) {
+  const trimmed = body.trim();
+  if (!trimmed) return "";
+  const firstPara = trimmed.split(/\n\s*\n/)[0].trim();
+  if (!firstPara) return "";
+  const stripped = firstPara.replace(/^#+\s*/, "").trim();
+  if (!stripped) return "";
+  if (stripped.length <= maxChars) return stripped;
+  const window = stripped.slice(0, maxChars);
+  const lastStop = Math.max(window.lastIndexOf(". "), window.lastIndexOf("! "), window.lastIndexOf("? "));
+  if (lastStop > maxChars * 0.6) {
+    return window.slice(0, lastStop + 1);
+  }
+  return window + "\u2026";
+}
+async function runSynthesis(input) {
+  const { query, results, samplingServer } = input;
+  if (!results || results.length === 0) {
+    return {
+      ok: false,
+      error: "no_content",
+      error_reason: "No sources returned content for this query",
+      stage: "synthesize",
+      hint: "Broaden the query, increase max_results, or remove restrictive filters"
+    };
+  }
+  if (samplingServer) {
+    try {
+      const r = await synthesizeAnswer(results, query, samplingServer);
+      if (r.answer && !r.fallback) {
+        return {
+          ok: true,
+          data: {
+            answer: r.answer,
+            citations: r.citations ?? [],
+            warning: r.warning,
+            fallback_level: 1
+          }
+        };
+      }
+    } catch (err) {
+      log.warn("synthesis level-1 sampling failed, falling through to heuristic", { error: String(err) });
+    }
+  }
+  const fb = buildStructuredFallback(results, query);
+  if (fb.answer && fb.answer.length > 0) {
+    return {
+      ok: true,
+      data: {
+        answer: fb.answer,
+        citations: fb.citations,
+        warning: fb.warning,
+        fallback_level: 2
+      }
+    };
+  }
+  const top = results.slice(0, 5);
+  const citations = top.map((r, i) => ({
+    index: i + 1,
+    url: r.url,
+    title: r.title,
+    snippet: r.snippet
+  }));
+  if (citations.length > 0) {
+    const lines = citations.map((c) => `[${c.index}] ${c.title} \u2014 ${c.url}
+${c.snippet ?? ""}`);
+    return {
+      ok: true,
+      data: {
+        answer: `Evidence for "${query}" (no synthesis available \u2014 content too sparse to summarize):
+
+${lines.join("\n\n")}`,
+        citations,
+        warning: "sparse_content: returned raw evidence dump instead of synthesized answer",
+        fallback_level: 3
+      }
+    };
+  }
+  return {
+    ok: false,
+    error: "no_content",
+    error_reason: "Sources returned but contained no usable text",
+    stage: "synthesize"
+  };
+}
+function extractCitations(answer, results) {
+  try {
+    if (!answer || results.length === 0) return [];
+    const citationRegex = /\[(\d+)\]/g;
+    const seen = /* @__PURE__ */ new Set();
+    const citations = [];
+    let match;
+    while ((match = citationRegex.exec(answer)) !== null) {
+      const index = parseInt(match[1], 10);
+      if (isNaN(index) || index < 1 || index > results.length) continue;
+      if (seen.has(index)) continue;
+      seen.add(index);
+      const result = results[index - 1];
+      citations.push({
+        index,
+        url: result.url,
+        title: result.title,
+        snippet: result.snippet
+      });
+    }
+    return citations;
+  } catch (err) {
+    log.error("citation extraction failed", { error: String(err) });
+    return [];
+  }
+}
+export {
+  buildSourcesText,
+  buildStructuredFallback,
+  buildSynthesisPrompt,
+  extractCitations,
+  runSynthesis,
+  synthesizeAnswer
+};
+//# sourceMappingURL=answer-synthesis.js.map

package/dist/search/answer-synthesis.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/search/answer-synthesis.ts"],"sourcesContent":["import type { SearchResultItem, Citation, StageResult } from '../types.js';\nimport type { SamplingCapableServer } from './sampling.js';\nimport {\n  checkSamplingSupport,\n  requestSampling,\n  extractTextFromSamplingResponse,\n} from './sampling.js';\nimport { createLogger } from '../logger.js';\n\nconst log = createLogger('search');\n\nconst MAX_CHARS_PER_SOURCE = 3000;\nconst MAX_RESPONSE_TOKENS = 1500;\nconst FALLBACK_MAX_BULLETS = 5;\nconst FALLBACK_KEYPOINT_MAX_CHARS = 240;\n\nexport interface SynthesisResult {\n  answer?: string;\n  citations?: Citation[];\n  fallback: boolean;\n  warning?: string;\n}\n\nexport async function synthesizeAnswer(\n  results: SearchResultItem[],\n  query: string,\n  server: SamplingCapableServer,\n): Promise<SynthesisResult> {\n  try {\n    const sourcesText = buildSourcesText(results);\n    if (!sourcesText) {\n      log.info('no content available for synthesis');\n      return {\n        fallback: true,\n        warning: 'No results with content available for answer synthesis',\n      };\n    }\n\n    if (!checkSamplingSupport(server)) {\n      log.info('sampling not supported by client, falling back to context format');\n      return {\n        fallback: true,\n        warning: 'Client does not support MCP sampling; falling back to context format',\n      };\n    }\n\n    const prompt = buildSynthesisPrompt(query, sourcesText);\n\n    const response = await requestSampling(\n      server,\n      [{ role: 'user', content: { type: 'text', text: prompt } }],\n      MAX_RESPONSE_TOKENS,\n    );\n\n    const answerText = extractTextFromSamplingResponse(response);\n\n    if (!answerText) {\n      log.warn('sampling returned empty response');\n      return {\n        fallback: true,\n        warning: 'Sampling returned empty response; falling back to context format',\n      };\n    }\n\n    const citations = extractCitations(answerText, results);\n\n    log.info('answer synthesis complete', {\n      answerLength: answerText.length,\n      citationCount: citations.length,\n    });\n\n    return {\n      answer: answerText,\n      citations,\n      fallback: false,\n    };\n  } catch (err) {\n    log.error('answer synthesis failed', { error: String(err) });\n    return {\n      fallback: true,\n      warning: `Answer synthesis failed: ${err instanceof Error ? err.message : String(err)}; falling back to context format`,\n    };\n  }\n}\n\nexport function buildSourcesText(results: SearchResultItem[]): string {\n  try {\n    const blocks: string[] = [];\n    let sourceIndex = 1;\n\n    for (const result of results) {\n      const content = result.markdown_content || result.snippet || '';\n      if (!content.trim()) continue;\n\n      const truncated = content.length > MAX_CHARS_PER_SOURCE\n        ? content.slice(0, MAX_CHARS_PER_SOURCE)\n        : content;\n\n      blocks.push(`[${sourceIndex}] ${result.title} (${result.url})\\n${truncated}`);\n      sourceIndex++;\n    }\n\n    if (blocks.length === 0) return '';\n\n    return blocks.join('\\n\\n---\\n\\n');\n  } catch (err) {\n    log.error('buildSourcesText failed', { error: String(err) });\n    return '';\n  }\n}\n\nexport function buildSynthesisPrompt(query: string, sourcesText: string): string {\n  return `Based on the following sources, provide a concise and direct answer to the question. Use numbered citations like [1], [2] to reference specific sources.\n\nQuestion: ${query}\n\nSources:\n${sourcesText}\n\nInstructions:\n- Be concise and direct. Answer in 2-4 paragraphs maximum.\n- Cite sources using [1], [2], etc. matching the source numbers above.\n- If sources contain conflicting information, note the discrepancy.\n- If the sources don't adequately answer the question, say so.\n- Do not include information not found in the provided sources.`;\n}\n\nexport interface StructuredFallbackResult {\n  answer: string;\n  citations: Citation[];\n  warning: string;\n}\n\n// Heuristic answer without LLM sampling: top-N sources as bulleted key points\n// with numeric citations. Used when client lacks sampling capability.\nexport function buildStructuredFallback(\n  results: SearchResultItem[],\n  query: string,\n): StructuredFallbackResult {\n  const bullets: string[] = [];\n  const citations: Citation[] = [];\n  let n = 0;\n\n  for (const r of results) {\n    if (n >= FALLBACK_MAX_BULLETS) break;\n    const body = (r.markdown_content && r.markdown_content.trim()) || (r.snippet && r.snippet.trim()) || '';\n    if (!body) continue;\n\n    const keypoint = extractKeypoint(body, FALLBACK_KEYPOINT_MAX_CHARS);\n    if (!keypoint) continue;\n\n    n += 1;\n    bullets.push(`- **${r.title}** — ${keypoint} [${n}]`);\n    citations.push({ index: n, url: r.url, title: r.title, snippet: r.snippet });\n  }\n\n  if (bullets.length === 0) {\n    return { answer: '', citations: [], warning: 'No sampling server available; no content to summarize' };\n  }\n\n  const q = query && query.trim() ? query.trim() : 'this query';\n  const answer = `Based on the top ${bullets.length} sources for \"${q}\":\\n\\n${bullets.join('\\n')}\\n\\nSources:\\n${citations.map(c => `[${c.index}] ${c.title} — ${c.url}`).join('\\n')}`;\n\n  return {\n    answer,\n    citations,\n    warning: 'Client does not support MCP sampling; returning heuristic key-point summary instead of synthesized answer',\n  };\n}\n\nfunction extractKeypoint(body: string, maxChars: number): string {\n  const trimmed = body.trim();\n  if (!trimmed) return '';\n\n  // First paragraph before a blank line\n  const firstPara = trimmed.split(/\\n\\s*\\n/)[0].trim();\n  if (!firstPara) return '';\n\n  // Strip markdown headings at the start\n  const stripped = firstPara.replace(/^#+\\s*/, '').trim();\n  if (!stripped) return '';\n\n  if (stripped.length <= maxChars) return stripped;\n\n  // Try to cut at sentence end within budget\n  const window = stripped.slice(0, maxChars);\n  const lastStop = Math.max(window.lastIndexOf('. '), window.lastIndexOf('! '), window.lastIndexOf('? '));\n  if (lastStop > maxChars * 0.6) {\n    return window.slice(0, lastStop + 1);\n  }\n  return window + '…';\n}\n\nexport interface SynthesisInput {\n  query: string;\n  results: SearchResultItem[];\n  samplingServer?: SamplingCapableServer;\n  /** Reserved for future per-source truncation; T6 currently relies on MAX_CHARS_PER_SOURCE. */\n  maxTotalChars: number;\n}\n\nexport interface SynthesizedAnswer {\n  answer: string;\n  citations: Citation[];\n  warning?: string;\n  fallback_level: 1 | 2 | 3;\n}\n\nexport async function runSynthesis(\n  input: SynthesisInput,\n): Promise<StageResult<SynthesizedAnswer>> {\n  const { query, results, samplingServer } = input;\n\n  if (!results || results.length === 0) {\n    return {\n      ok: false,\n      error: 'no_content',\n      error_reason: 'No sources returned content for this query',\n      stage: 'synthesize',\n      hint: 'Broaden the query, increase max_results, or remove restrictive filters',\n    };\n  }\n\n  if (samplingServer) {\n    try {\n      const r = await synthesizeAnswer(results, query, samplingServer);\n      if (r.answer && !r.fallback) {\n        return {\n          ok: true,\n          data: {\n            answer: r.answer,\n            citations: r.citations ?? [],\n            warning: r.warning,\n            fallback_level: 1,\n          },\n        };\n      }\n    } catch (err) {\n      log.warn('synthesis level-1 sampling failed, falling through to heuristic', { error: String(err) });\n    }\n  }\n\n  const fb = buildStructuredFallback(results, query);\n  if (fb.answer && fb.answer.length > 0) {\n    return {\n      ok: true,\n      data: {\n        answer: fb.answer,\n        citations: fb.citations,\n        warning: fb.warning,\n        fallback_level: 2,\n      },\n    };\n  }\n\n  const top = results.slice(0, 5);\n  const citations: Citation[] = top.map((r, i) => ({\n    index: i + 1, url: r.url, title: r.title, snippet: r.snippet,\n  }));\n  if (citations.length > 0) {\n    const lines = citations.map(c => `[${c.index}] ${c.title} — ${c.url}\\n${c.snippet ?? ''}`);\n    return {\n      ok: true,\n      data: {\n        answer: `Evidence for \"${query}\" (no synthesis available — content too sparse to summarize):\\n\\n${lines.join('\\n\\n')}`,\n        citations,\n        warning: 'sparse_content: returned raw evidence dump instead of synthesized answer',\n        fallback_level: 3,\n      },\n    };\n  }\n\n  return {\n    ok: false,\n    error: 'no_content',\n    error_reason: 'Sources returned but contained no usable text',\n    stage: 'synthesize',\n  };\n}\n\nexport function extractCitations(\n  answer: string,\n  results: SearchResultItem[],\n): Citation[] {\n  try {\n    if (!answer || results.length === 0) return [];\n\n    const citationRegex = /\\[(\\d+)\\]/g;\n    const seen = new Set<number>();\n    const citations: Citation[] = [];\n\n    let match: RegExpExecArray | null;\n    while ((match = citationRegex.exec(answer)) !== null) {\n      const index = parseInt(match[1], 10);\n      if (isNaN(index) || index < 1 || index > results.length) continue;\n      if (seen.has(index)) continue;\n      seen.add(index);\n\n      const result = results[index - 1];\n      citations.push({\n        index,\n        url: result.url,\n        title: result.title,\n        snippet: result.snippet,\n      });\n    }\n\n    return citations;\n  } catch (err) {\n    log.error('citation extraction failed', { error: String(err) });\n    return [];\n  }\n}\n"],"mappings":"AAEA;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,oBAAoB;AAE7B,MAAM,MAAM,aAAa,QAAQ;AAEjC,MAAM,uBAAuB;AAC7B,MAAM,sBAAsB;AAC5B,MAAM,uBAAuB;AAC7B,MAAM,8BAA8B;AASpC,eAAsB,iBACpB,SACA,OACA,QAC0B;AAC1B,MAAI;AACF,UAAM,cAAc,iBAAiB,OAAO;AAC5C,QAAI,CAAC,aAAa;AAChB,UAAI,KAAK,oCAAoC;AAC7C,aAAO;AAAA,QACL,UAAU;AAAA,QACV,SAAS;AAAA,MACX;AAAA,IACF;AAEA,QAAI,CAAC,qBAAqB,MAAM,GAAG;AACjC,UAAI,KAAK,kEAAkE;AAC3E,aAAO;AAAA,QACL,UAAU;AAAA,QACV,SAAS;AAAA,MACX;AAAA,IACF;AAEA,UAAM,SAAS,qBAAqB,OAAO,WAAW;AAEtD,UAAM,WAAW,MAAM;AAAA,MACrB;AAAA,MACA,CAAC,EAAE,MAAM,QAAQ,SAAS,EAAE,MAAM,QAAQ,MAAM,OAAO,EAAE,CAAC;AAAA,MAC1D;AAAA,IACF;AAEA,UAAM,aAAa,gCAAgC,QAAQ;AAE3D,QAAI,CAAC,YAAY;AACf,UAAI,KAAK,kCAAkC;AAC3C,aAAO;AAAA,QACL,UAAU;AAAA,QACV,SAAS;AAAA,MACX;AAAA,IACF;AAEA,UAAM,YAAY,iBAAiB,YAAY,OAAO;AAEtD,QAAI,KAAK,6BAA6B;AAAA,MACpC,cAAc,WAAW;AAAA,MACzB,eAAe,UAAU;AAAA,IAC3B,CAAC;AAED,WAAO;AAAA,MACL,QAAQ;AAAA,MACR;AAAA,MACA,UAAU;AAAA,IACZ;AAAA,EACF,SAAS,KAAK;AACZ,QAAI,MAAM,2BAA2B,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AAC3D,WAAO;AAAA,MACL,UAAU;AAAA,MACV,SAAS,4BAA4B,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,CAAC;AAAA,IACvF;AAAA,EACF;AACF;AAEO,SAAS,iBAAiB,SAAqC;AACpE,MAAI;AACF,UAAM,SAAmB,CAAC;AAC1B,QAAI,cAAc;AAElB,eAAW,UAAU,SAAS;AAC5B,YAAM,UAAU,OAAO,oBAAoB,OAAO,WAAW;AAC7D,UAAI,CAAC,QAAQ,KAAK,EAAG;AAErB,YAAM,YAAY,QAAQ,SAAS,uBAC/B,QAAQ,MAAM,GAAG,oBAAoB,IACrC;AAEJ,aAAO,KAAK,IAAI,WAAW,KAAK,OAAO,KAAK,KAAK,OAAO,GAAG;AAAA,EAAM,SAAS,EAAE;AAC5E;AAAA,IACF;AAEA,QAAI,OAAO,WAAW,EAAG,QAAO;AAEhC,WAAO,OAAO,KAAK,aAAa;AAAA,EAClC,SAAS,KAAK;AACZ,QAAI,MAAM,2BAA2B,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AAC3D,WAAO;AAAA,EACT;AACF;AAEO,SAAS,qBAAqB,OAAe,aAA6B;AAC/E,SAAO;AAAA;AAAA,YAEG,KAAK;AAAA;AAAA;AAAA,EAGf,WAAW;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAQb;AAUO,SAAS,wBACd,SACA,OAC0B;AAC1B,QAAM,UAAoB,CAAC;AAC3B,QAAM,YAAwB,CAAC;AAC/B,MAAI,IAAI;AAER,aAAW,KAAK,SAAS;AACvB,QAAI,KAAK,qBAAsB;AAC/B,UAAM,OAAQ,EAAE,oBAAoB,EAAE,iBAAiB,KAAK,KAAO,EAAE,WAAW,EAAE,QAAQ,KAAK,KAAM;AACrG,QAAI,CAAC,KAAM;AAEX,UAAM,WAAW,gBAAgB,MAAM,2BAA2B;AAClE,QAAI,CAAC,SAAU;AAEf,SAAK;AACL,YAAQ,KAAK,OAAO,EAAE,KAAK,aAAQ,QAAQ,KAAK,CAAC,GAAG;AACpD,cAAU,KAAK,EAAE,OAAO,GAAG,KAAK,EAAE,KAAK,OAAO,EAAE,OAAO,SAAS,EAAE,QAAQ,CAAC;AAAA,EAC7E;AAEA,MAAI,QAAQ,WAAW,GAAG;AACxB,WAAO,EAAE,QAAQ,IAAI,WAAW,CAAC,GAAG,SAAS,wDAAwD;AAAA,EACvG;AAEA,QAAM,IAAI,SAAS,MAAM,KAAK,IAAI,MAAM,KAAK,IAAI;AACjD,QAAM,SAAS,oBAAoB,QAAQ,MAAM,iBAAiB,CAAC;AAAA;AAAA,EAAS,QAAQ,KAAK,IAAI,CAAC;AAAA;AAAA;AAAA,EAAiB,UAAU,IAAI,OAAK,IAAI,EAAE,KAAK,KAAK,EAAE,KAAK,WAAM,EAAE,GAAG,EAAE,EAAE,KAAK,IAAI,CAAC;AAElL,SAAO;AAAA,IACL;AAAA,IACA;AAAA,IACA,SAAS;AAAA,EACX;AACF;AAEA,SAAS,gBAAgB,MAAc,UAA0B;AAC/D,QAAM,UAAU,KAAK,KAAK;AAC1B,MAAI,CAAC,QAAS,QAAO;AAGrB,QAAM,YAAY,QAAQ,MAAM,SAAS,EAAE,CAAC,EAAE,KAAK;AACnD,MAAI,CAAC,UAAW,QAAO;AAGvB,QAAM,WAAW,UAAU,QAAQ,UAAU,EAAE,EAAE,KAAK;AACtD,MAAI,CAAC,SAAU,QAAO;AAEtB,MAAI,SAAS,UAAU,SAAU,QAAO;AAGxC,QAAM,SAAS,SAAS,MAAM,GAAG,QAAQ;AACzC,QAAM,WAAW,KAAK,IAAI,OAAO,YAAY,IAAI,GAAG,OAAO,YAAY,IAAI,GAAG,OAAO,YAAY,IAAI,CAAC;AACtG,MAAI,WAAW,WAAW,KAAK;AAC7B,WAAO,OAAO,MAAM,GAAG,WAAW,CAAC;AAAA,EACrC;AACA,SAAO,SAAS;AAClB;AAiBA,eAAsB,aACpB,OACyC;AACzC,QAAM,EAAE,OAAO,SAAS,eAAe,IAAI;AAE3C,MAAI,CAAC,WAAW,QAAQ,WAAW,GAAG;AACpC,WAAO;AAAA,MACL,IAAI;AAAA,MACJ,OAAO;AAAA,MACP,cAAc;AAAA,MACd,OAAO;AAAA,MACP,MAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI,gBAAgB;AAClB,QAAI;AACF,YAAM,IAAI,MAAM,iBAAiB,SAAS,OAAO,cAAc;AAC/D,UAAI,EAAE,UAAU,CAAC,EAAE,UAAU;AAC3B,eAAO;AAAA,UACL,IAAI;AAAA,UACJ,MAAM;AAAA,YACJ,QAAQ,EAAE;AAAA,YACV,WAAW,EAAE,aAAa,CAAC;AAAA,YAC3B,SAAS,EAAE;AAAA,YACX,gBAAgB;AAAA,UAClB;AAAA,QACF;AAAA,MACF;AAAA,IACF,SAAS,KAAK;AACZ,UAAI,KAAK,mEAAmE,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AAAA,IACpG;AAAA,EACF;AAEA,QAAM,KAAK,wBAAwB,SAAS,KAAK;AACjD,MAAI,GAAG,UAAU,GAAG,OAAO,SAAS,GAAG;AACrC,WAAO;AAAA,MACL,IAAI;AAAA,MACJ,MAAM;AAAA,QACJ,QAAQ,GAAG;AAAA,QACX,WAAW,GAAG;AAAA,QACd,SAAS,GAAG;AAAA,QACZ,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,QAAM,MAAM,QAAQ,MAAM,GAAG,CAAC;AAC9B,QAAM,YAAwB,IAAI,IAAI,CAAC,GAAG,OAAO;AAAA,IAC/C,OAAO,IAAI;AAAA,IAAG,KAAK,EAAE;AAAA,IAAK,OAAO,EAAE;AAAA,IAAO,SAAS,EAAE;AAAA,EACvD,EAAE;AACF,MAAI,UAAU,SAAS,GAAG;AACxB,UAAM,QAAQ,UAAU,IAAI,OAAK,IAAI,EAAE,KAAK,KAAK,EAAE,KAAK,WAAM,EAAE,GAAG;AAAA,EAAK,EAAE,WAAW,EAAE,EAAE;AACzF,WAAO;AAAA,MACL,IAAI;AAAA,MACJ,MAAM;AAAA,QACJ,QAAQ,iBAAiB,KAAK;AAAA;AAAA,EAAoE,MAAM,KAAK,MAAM,CAAC;AAAA,QACpH;AAAA,QACA,SAAS;AAAA,QACT,gBAAgB;AAAA,MAClB;AAAA,IACF;AAAA,EACF;AAEA,SAAO;AAAA,IACL,IAAI;AAAA,IACJ,OAAO;AAAA,IACP,cAAc;AAAA,IACd,OAAO;AAAA,EACT;AACF;AAEO,SAAS,iBACd,QACA,SACY;AACZ,MAAI;AACF,QAAI,CAAC,UAAU,QAAQ,WAAW,EAAG,QAAO,CAAC;AAE7C,UAAM,gBAAgB;AACtB,UAAM,OAAO,oBAAI,IAAY;AAC7B,UAAM,YAAwB,CAAC;AAE/B,QAAI;AACJ,YAAQ,QAAQ,cAAc,KAAK,MAAM,OAAO,MAAM;AACpD,YAAM,QAAQ,SAAS,MAAM,CAAC,GAAG,EAAE;AACnC,UAAI,MAAM,KAAK,KAAK,QAAQ,KAAK,QAAQ,QAAQ,OAAQ;AACzD,UAAI,KAAK,IAAI,KAAK,EAAG;AACrB,WAAK,IAAI,KAAK;AAEd,YAAM,SAAS,QAAQ,QAAQ,CAAC;AAChC,gBAAU,KAAK;AAAA,QACb;AAAA,QACA,KAAK,OAAO;AAAA,QACZ,OAAO,OAAO;AAAA,QACd,SAAS,OAAO;AAAA,MAClB,CAAC;AAAA,IACH;AAEA,WAAO;AAAA,EACT,SAAS,KAAK;AACZ,QAAI,MAAM,8BAA8B,EAAE,OAAO,OAAO,GAAG,EAAE,CAAC;AAC9D,WAAO,CAAC;AAAA,EACV;AACF;","names":[]}

package/dist/search/context-formatter.d.ts

@@ -0,0 +1,3 @@
+import type { SearchResultItem } from '../types.js';
+export declare function formatSearchContext(results: SearchResultItem[], maxTotalChars: number): string;
+//# sourceMappingURL=context-formatter.d.ts.map

package/dist/search/context-formatter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"context-formatter.d.ts","sourceRoot":"","sources":["../../src/search/context-formatter.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAIpD,wBAAgB,mBAAmB,CAAC,OAAO,EAAE,gBAAgB,EAAE,EAAE,aAAa,EAAE,MAAM,GAAG,MAAM,CA0D9F"}

package/dist/search/context-formatter.js

@@ -0,0 +1,56 @@
+import { createLogger } from "../logger.js";
+const log = createLogger("search");
+function formatSearchContext(results, maxTotalChars) {
+  if (results.length === 0 || maxTotalChars <= 0) {
+    return "";
+  }
+  const blocks = [];
+  let charsUsed = 0;
+  for (let i = 0; i < results.length; i++) {
+    const result = results[i];
+    const content = result.markdown_content || result.snippet || "";
+    const header = `Source: ${result.title} (${result.url})`;
+    const separator = "\n";
+    const blockSeparator = i > 0 ? "\n\n" : "";
+    const minBlockSize = blockSeparator.length + header.length + separator.length;
+    if (i > 0 && charsUsed + minBlockSize >= maxTotalChars) {
+      log.debug("context formatter stopping: next header exceeds budget", {
+        resultIndex: i,
+        charsUsed,
+        maxTotalChars
+      });
+      break;
+    }
+    const remainingBudget = maxTotalChars - charsUsed - blockSeparator.length - header.length - separator.length;
+    let truncatedContent;
+    if (remainingBudget <= 0) {
+      const totalAvailable = maxTotalChars - charsUsed - blockSeparator.length;
+      if (totalAvailable <= 0 && i > 0) break;
+      const headerTruncated = totalAvailable > 3 ? header.slice(0, totalAvailable - 3) + "..." : header.slice(0, Math.max(totalAvailable, 0));
+      blocks.push(blockSeparator + headerTruncated);
+      charsUsed += blockSeparator.length + headerTruncated.length;
+      break;
+    } else if (content.length > remainingBudget) {
+      const truncLen = Math.max(remainingBudget - 3, 0);
+      truncatedContent = truncLen > 0 ? content.slice(0, truncLen) + "..." : "...";
+      log.debug("context formatter truncated result content", {
+        resultIndex: i,
+        originalLength: content.length,
+        truncatedTo: truncatedContent.length
+      });
+    } else {
+      truncatedContent = content;
+    }
+    const block = blockSeparator + header + separator + truncatedContent;
+    blocks.push(block);
+    charsUsed += block.length;
+    if (charsUsed >= maxTotalChars) {
+      break;
+    }
+  }
+  return blocks.join("").trimEnd();
+}
+export {
+  formatSearchContext
+};
+//# sourceMappingURL=context-formatter.js.map

package/dist/search/context-formatter.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/search/context-formatter.ts"],"sourcesContent":["import { createLogger } from '../logger.js';\nimport type { SearchResultItem } from '../types.js';\n\nconst log = createLogger('search');\n\nexport function formatSearchContext(results: SearchResultItem[], maxTotalChars: number): string {\n  if (results.length === 0 || maxTotalChars <= 0) {\n    return '';\n  }\n\n  const blocks: string[] = [];\n  let charsUsed = 0;\n\n  for (let i = 0; i < results.length; i++) {\n    const result = results[i];\n    const content = result.markdown_content || result.snippet || '';\n    const header = `Source: ${result.title} (${result.url})`;\n    const separator = '\\n';\n    const blockSeparator = i > 0 ? '\\n\\n' : '';\n\n    const minBlockSize = blockSeparator.length + header.length + separator.length;\n\n    if (i > 0 && charsUsed + minBlockSize >= maxTotalChars) {\n      log.debug('context formatter stopping: next header exceeds budget', {\n        resultIndex: i,\n        charsUsed,\n        maxTotalChars,\n      });\n      break;\n    }\n\n    const remainingBudget = maxTotalChars - charsUsed - blockSeparator.length - header.length - separator.length;\n\n    let truncatedContent: string;\n    if (remainingBudget <= 0) {\n      const totalAvailable = maxTotalChars - charsUsed - blockSeparator.length;\n      if (totalAvailable <= 0 && i > 0) break;\n      const headerTruncated = totalAvailable > 3 ? header.slice(0, totalAvailable - 3) + '...' : header.slice(0, Math.max(totalAvailable, 0));\n      blocks.push(blockSeparator + headerTruncated);\n      charsUsed += blockSeparator.length + headerTruncated.length;\n      break;\n    } else if (content.length > remainingBudget) {\n      const truncLen = Math.max(remainingBudget - 3, 0);\n      truncatedContent = truncLen > 0 ? content.slice(0, truncLen) + '...' : '...';\n      log.debug('context formatter truncated result content', {\n        resultIndex: i,\n        originalLength: content.length,\n        truncatedTo: truncatedContent.length,\n      });\n    } else {\n      truncatedContent = content;\n    }\n\n    const block = blockSeparator + header + separator + truncatedContent;\n    blocks.push(block);\n    charsUsed += block.length;\n\n    if (charsUsed >= maxTotalChars) {\n      break;\n    }\n  }\n\n  return blocks.join('').trimEnd();\n}\n"],"mappings":"AAAA,SAAS,oBAAoB;AAG7B,MAAM,MAAM,aAAa,QAAQ;AAE1B,SAAS,oBAAoB,SAA6B,eAA+B;AAC9F,MAAI,QAAQ,WAAW,KAAK,iBAAiB,GAAG;AAC9C,WAAO;AAAA,EACT;AAEA,QAAM,SAAmB,CAAC;AAC1B,MAAI,YAAY;AAEhB,WAAS,IAAI,GAAG,IAAI,QAAQ,QAAQ,KAAK;AACvC,UAAM,SAAS,QAAQ,CAAC;AACxB,UAAM,UAAU,OAAO,oBAAoB,OAAO,WAAW;AAC7D,UAAM,SAAS,WAAW,OAAO,KAAK,KAAK,OAAO,GAAG;AACrD,UAAM,YAAY;AAClB,UAAM,iBAAiB,IAAI,IAAI,SAAS;AAExC,UAAM,eAAe,eAAe,SAAS,OAAO,SAAS,UAAU;AAEvE,QAAI,IAAI,KAAK,YAAY,gBAAgB,eAAe;AACtD,UAAI,MAAM,0DAA0D;AAAA,QAClE,aAAa;AAAA,QACb;AAAA,QACA;AAAA,MACF,CAAC;AACD;AAAA,IACF;AAEA,UAAM,kBAAkB,gBAAgB,YAAY,eAAe,SAAS,OAAO,SAAS,UAAU;AAEtG,QAAI;AACJ,QAAI,mBAAmB,GAAG;AACxB,YAAM,iBAAiB,gBAAgB,YAAY,eAAe;AAClE,UAAI,kBAAkB,KAAK,IAAI,EAAG;AAClC,YAAM,kBAAkB,iBAAiB,IAAI,OAAO,MAAM,GAAG,iBAAiB,CAAC,IAAI,QAAQ,OAAO,MAAM,GAAG,KAAK,IAAI,gBAAgB,CAAC,CAAC;AACtI,aAAO,KAAK,iBAAiB,eAAe;AAC5C,mBAAa,eAAe,SAAS,gBAAgB;AACrD;AAAA,IACF,WAAW,QAAQ,SAAS,iBAAiB;AAC3C,YAAM,WAAW,KAAK,IAAI,kBAAkB,GAAG,CAAC;AAChD,yBAAmB,WAAW,IAAI,QAAQ,MAAM,GAAG,QAAQ,IAAI,QAAQ;AACvE,UAAI,MAAM,8CAA8C;AAAA,QACtD,aAAa;AAAA,QACb,gBAAgB,QAAQ;AAAA,QACxB,aAAa,iBAAiB;AAAA,MAChC,CAAC;AAAA,IACH,OAAO;AACL,yBAAmB;AAAA,IACrB;AAEA,UAAM,QAAQ,iBAAiB,SAAS,YAAY;AACpD,WAAO,KAAK,KAAK;AACjB,iBAAa,MAAM;AAEnB,QAAI,aAAa,eAAe;AAC9B;AAAA,IACF;AAAA,EACF;AAEA,SAAO,OAAO,KAAK,EAAE,EAAE,QAAQ;AACjC;","names":[]}

package/dist/search/dedup.d.ts

@@ -5,6 +5,7 @@ export interface MergedSearchResult {
     snippet: string;
     relevance_score: number;
     engines: string[];
+    published_date?: string;
 }
 export declare function deduplicateResults(results: RawSearchResult[]): MergedSearchResult[];
 //# sourceMappingURL=dedup.d.ts.map

package/dist/search/dedup.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"dedup.d.ts","sourceRoot":"","sources":["../../src/search/dedup.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AAEnD,MAAM,WAAW,kBAAkB;IACjC,KAAK,EAAE,MAAM,CAAC;IACd,GAAG,EAAE,MAAM,CAAC;IACZ,OAAO,EAAE,MAAM,CAAC;IAChB,eAAe,EAAE,MAAM,CAAC;IACxB,OAAO,EAAE,MAAM,EAAE,CAAC;CACnB;AAED,wBAAgB,kBAAkB,CAAC,OAAO,EAAE,eAAe,EAAE,GAAG,kBAAkB,EAAE,CAkCnF"}
+{"version":3,"file":"dedup.d.ts","sourceRoot":"","sources":["../../src/search/dedup.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,aAAa,CAAC;AAGnD,MAAM,WAAW,kBAAkB;IACjC,KAAK,EAAE,MAAM,CAAC;IACd,GAAG,EAAE,MAAM,CAAC;IACZ,OAAO,EAAE,MAAM,CAAC;IAChB,eAAe,EAAE,MAAM,CAAC;IACxB,OAAO,EAAE,MAAM,EAAE,CAAC;IAClB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,wBAAgB,kBAAkB,CAAC,OAAO,EAAE,eAAe,EAAE,GAAG,kBAAkB,EAAE,CAwCnF"}

package/dist/search/dedup.js

@@ -1,35 +1,43 @@
-import { normalizeUrl } from '../cache/store.js';
-export function deduplicateResults(results) {
-    const urlMap = new Map();
-    for (const result of results) {
-        let normalized;
-        try {
-            normalized = normalizeUrl(result.url);
-        }
-        catch {
-            normalized = result.url;
-        }
-        const existing = urlMap.get(normalized);
-        if (existing) {
-            if (result.relevance_score > existing.relevance_score) {
-                existing.relevance_score = result.relevance_score;
-                existing.title = result.title;
-                existing.snippet = result.snippet;
-            }
-            if (!existing.engines.includes(result.engine)) {
-                existing.engines.push(result.engine);
-            }
-        }
-        else {
-            urlMap.set(normalized, {
-                title: result.title,
-                url: result.url,
-                snippet: result.snippet,
-                relevance_score: result.relevance_score,
-                engines: [result.engine],
-            });
-        }
+import { normalizeUrl } from "../cache/store.js";
+import { normalizeResultUrl } from "./url-unwrap.js";
+function deduplicateResults(results) {
+  const urlMap = /* @__PURE__ */ new Map();
+  for (const raw of results) {
+    const unwrappedUrl = normalizeResultUrl(raw.url);
+    const result = unwrappedUrl !== raw.url ? { ...raw, url: unwrappedUrl } : raw;
+    let normalized;
+    try {
+      normalized = normalizeUrl(result.url);
+    } catch {
+      normalized = result.url;
     }
-    return [...urlMap.values()].sort((a, b) => b.relevance_score - a.relevance_score);
+    const existing = urlMap.get(normalized);
+    if (existing) {
+      if (result.relevance_score > existing.relevance_score) {
+        existing.relevance_score = result.relevance_score;
+        existing.title = result.title;
+        existing.snippet = result.snippet;
+      }
+      if (result.published_date && !existing.published_date) {
+        existing.published_date = result.published_date;
+      }
+      if (!existing.engines.includes(result.engine)) {
+        existing.engines.push(result.engine);
+      }
+    } else {
+      urlMap.set(normalized, {
+        title: result.title,
+        url: result.url,
+        snippet: result.snippet,
+        relevance_score: result.relevance_score,
+        engines: [result.engine],
+        ...result.published_date ? { published_date: result.published_date } : {}
+      });
+    }
+  }
+  return [...urlMap.values()].sort((a, b) => b.relevance_score - a.relevance_score);
 }
+export {
+  deduplicateResults
+};
 //# sourceMappingURL=dedup.js.map

package/dist/search/dedup.js.map

@@ -1 +1 @@
-{"version":3,"file":"dedup.js","sourceRoot":"","sources":["../../src/search/dedup.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAC;AAWjD,MAAM,UAAU,kBAAkB,CAAC,OAA0B;IAC3D,MAAM,MAAM,GAAG,IAAI,GAAG,EAA8B,CAAC;IAErD,KAAK,MAAM,MAAM,IAAI,OAAO,EAAE,CAAC;QAC7B,IAAI,UAAkB,CAAC;QACvB,IAAI,CAAC;YACH,UAAU,GAAG,YAAY,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACxC,CAAC;QAAC,MAAM,CAAC;YACP,UAAU,GAAG,MAAM,CAAC,GAAG,CAAC;QAC1B,CAAC;QAED,MAAM,QAAQ,GAAG,MAAM,CAAC,GAAG,CAAC,UAAU,CAAC,CAAC;QAExC,IAAI,QAAQ,EAAE,CAAC;YACb,IAAI,MAAM,CAAC,eAAe,GAAG,QAAQ,CAAC,eAAe,EAAE,CAAC;gBACtD,QAAQ,CAAC,eAAe,GAAG,MAAM,CAAC,eAAe,CAAC;gBAClD,QAAQ,CAAC,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC;gBAC9B,QAAQ,CAAC,OAAO,GAAG,MAAM,CAAC,OAAO,CAAC;YACpC,CAAC;YACD,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,CAAC;gBAC9C,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;YACvC,CAAC;QACH,CAAC;aAAM,CAAC;YACN,MAAM,CAAC,GAAG,CAAC,UAAU,EAAE;gBACrB,KAAK,EAAE,MAAM,CAAC,KAAK;gBACnB,GAAG,EAAE,MAAM,CAAC,GAAG;gBACf,OAAO,EAAE,MAAM,CAAC,OAAO;gBACvB,eAAe,EAAE,MAAM,CAAC,eAAe;gBACvC,OAAO,EAAE,CAAC,MAAM,CAAC,MAAM,CAAC;aACzB,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED,OAAO,CAAC,GAAG,MAAM,CAAC,MAAM,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,eAAe,GAAG,CAAC,CAAC,eAAe,CAAC,CAAC;AACpF,CAAC"}
+{"version":3,"sources":["../../src/search/dedup.ts"],"sourcesContent":["import { normalizeUrl } from '../cache/store.js';\nimport type { RawSearchResult } from '../types.js';\nimport { normalizeResultUrl } from './url-unwrap.js';\n\nexport interface MergedSearchResult {\n  title: string;\n  url: string;\n  snippet: string;\n  relevance_score: number;\n  engines: string[];\n  published_date?: string;\n}\n\nexport function deduplicateResults(results: RawSearchResult[]): MergedSearchResult[] {\n  const urlMap = new Map<string, MergedSearchResult>();\n\n  for (const raw of results) {\n    const unwrappedUrl = normalizeResultUrl(raw.url);\n    const result = unwrappedUrl !== raw.url ? { ...raw, url: unwrappedUrl } : raw;\n    let normalized: string;\n    try {\n      normalized = normalizeUrl(result.url);\n    } catch {\n      normalized = result.url;\n    }\n\n    const existing = urlMap.get(normalized);\n\n    if (existing) {\n      if (result.relevance_score > existing.relevance_score) {\n        existing.relevance_score = result.relevance_score;\n        existing.title = result.title;\n        existing.snippet = result.snippet;\n      }\n      if (result.published_date && !existing.published_date) {\n        existing.published_date = result.published_date;\n      }\n      if (!existing.engines.includes(result.engine)) {\n        existing.engines.push(result.engine);\n      }\n    } else {\n      urlMap.set(normalized, {\n        title: result.title,\n        url: result.url,\n        snippet: result.snippet,\n        relevance_score: result.relevance_score,\n        engines: [result.engine],\n        ...(result.published_date ? { published_date: result.published_date } : {}),\n      });\n    }\n  }\n\n  return [...urlMap.values()].sort((a, b) => b.relevance_score - a.relevance_score);\n}\n"],"mappings":"AAAA,SAAS,oBAAoB;AAE7B,SAAS,0BAA0B;AAW5B,SAAS,mBAAmB,SAAkD;AACnF,QAAM,SAAS,oBAAI,IAAgC;AAEnD,aAAW,OAAO,SAAS;AACzB,UAAM,eAAe,mBAAmB,IAAI,GAAG;AAC/C,UAAM,SAAS,iBAAiB,IAAI,MAAM,EAAE,GAAG,KAAK,KAAK,aAAa,IAAI;AAC1E,QAAI;AACJ,QAAI;AACF,mBAAa,aAAa,OAAO,GAAG;AAAA,IACtC,QAAQ;AACN,mBAAa,OAAO;AAAA,IACtB;AAEA,UAAM,WAAW,OAAO,IAAI,UAAU;AAEtC,QAAI,UAAU;AACZ,UAAI,OAAO,kBAAkB,SAAS,iBAAiB;AACrD,iBAAS,kBAAkB,OAAO;AAClC,iBAAS,QAAQ,OAAO;AACxB,iBAAS,UAAU,OAAO;AAAA,MAC5B;AACA,UAAI,OAAO,kBAAkB,CAAC,SAAS,gBAAgB;AACrD,iBAAS,iBAAiB,OAAO;AAAA,MACnC;AACA,UAAI,CAAC,SAAS,QAAQ,SAAS,OAAO,MAAM,GAAG;AAC7C,iBAAS,QAAQ,KAAK,OAAO,MAAM;AAAA,MACrC;AAAA,IACF,OAAO;AACL,aAAO,IAAI,YAAY;AAAA,QACrB,OAAO,OAAO;AAAA,QACd,KAAK,OAAO;AAAA,QACZ,SAAS,OAAO;AAAA,QAChB,iBAAiB,OAAO;AAAA,QACxB,SAAS,CAAC,OAAO,MAAM;AAAA,QACvB,GAAI,OAAO,iBAAiB,EAAE,gBAAgB,OAAO,eAAe,IAAI,CAAC;AAAA,MAC3E,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO,CAAC,GAAG,OAAO,OAAO,CAAC,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,kBAAkB,EAAE,eAAe;AAClF;","names":[]}

package/dist/search/engines/arxiv.d.ts

@@ -0,0 +1,7 @@
+import type { SearchEngine, SearchEngineOptions, RawSearchResult } from '../../types.js';
+export declare class ArxivEngine implements SearchEngine {
+    name: string;
+    search(query: string, options?: SearchEngineOptions): Promise<RawSearchResult[]>;
+    private parseAtom;
+}
+//# sourceMappingURL=arxiv.d.ts.map

package/dist/search/engines/arxiv.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"arxiv.d.ts","sourceRoot":"","sources":["../../../src/search/engines/arxiv.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,YAAY,EAAE,mBAAmB,EAAE,eAAe,EAAE,MAAM,gBAAgB,CAAC;AAWzF,qBAAa,WAAY,YAAW,YAAY;IAC9C,IAAI,SAAW;IAET,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,GAAE,mBAAwB,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC;IAoB1F,OAAO,CAAC,SAAS;CAgClB"}

package/dist/search/engines/arxiv.js

@@ -0,0 +1,70 @@
+import { parseHTML } from "linkedom";
+import { createLogger } from "../../logger.js";
+const log = createLogger("search");
+const SNIPPET_LIMIT = 200;
+function textContent(el) {
+  return (el?.textContent ?? "").replace(/\s+/g, " ").trim();
+}
+class ArxivEngine {
+  name = "arxiv";
+  async search(query, options = {}) {
+    const timeoutMs = options.timeoutMs ?? 1e4;
+    const maxResults = options.maxResults ?? 10;
+    const params = new URLSearchParams({
+      search_query: `all:${query}`,
+      max_results: String(maxResults),
+      sortBy: "relevance"
+    });
+    const url = `http://export.arxiv.org/api/query?${params}`;
+    log.debug("arxiv search", { query });
+    const response = await fetch(url, { signal: AbortSignal.timeout(timeoutMs) });
+    if (!response.ok) throw new Error(`arXiv returned ${response.status}`);
+    const xml = await response.text();
+    const mapped = this.parseAtom(xml);
+    return applyDateFilter(mapped, options);
+  }
+  parseAtom(xml) {
+    const { document } = parseHTML(xml);
+    const entries = Array.from(document.querySelectorAll("entry"));
+    const results = [];
+    const total = entries.length;
+    for (let i = 0; i < total; i++) {
+      const entry = entries[i];
+      const id = textContent(entry.querySelector("id"));
+      const title = textContent(entry.querySelector("title"));
+      const summary = textContent(entry.querySelector("summary"));
+      const published = textContent(entry.querySelector("published"));
+      if (!id || !title) continue;
+      let published_date;
+      if (published) {
+        const d = new Date(published);
+        if (!isNaN(d.getTime())) published_date = d.toISOString();
+      }
+      results.push({
+        title,
+        url: id,
+        snippet: summary.slice(0, SNIPPET_LIMIT),
+        relevance_score: 1 - i / Math.max(total, 1),
+        engine: "arxiv",
+        ...published_date ? { published_date } : {}
+      });
+    }
+    return results;
+  }
+}
+function applyDateFilter(results, options) {
+  if (!options.fromDate && !options.toDate) return results;
+  const from = options.fromDate ? new Date(options.fromDate).getTime() : -Infinity;
+  const to = options.toDate ? new Date(options.toDate).getTime() : Infinity;
+  if (isNaN(from) || isNaN(to)) return results;
+  return results.filter((r) => {
+    if (!r.published_date) return false;
+    const t = new Date(r.published_date).getTime();
+    if (isNaN(t)) return false;
+    return t >= from && t <= to;
+  });
+}
+export {
+  ArxivEngine
+};
+//# sourceMappingURL=arxiv.js.map

package/dist/search/engines/arxiv.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/search/engines/arxiv.ts"],"sourcesContent":["import { parseHTML } from 'linkedom';\nimport type { SearchEngine, SearchEngineOptions, RawSearchResult } from '../../types.js';\nimport { createLogger } from '../../logger.js';\n\nconst log = createLogger('search');\n\nconst SNIPPET_LIMIT = 200;\n\nfunction textContent(el: Element | null): string {\n  return (el?.textContent ?? '').replace(/\\s+/g, ' ').trim();\n}\n\nexport class ArxivEngine implements SearchEngine {\n  name = 'arxiv';\n\n  async search(query: string, options: SearchEngineOptions = {}): Promise<RawSearchResult[]> {\n    const timeoutMs = options.timeoutMs ?? 10000;\n    const maxResults = options.maxResults ?? 10;\n\n    const params = new URLSearchParams({\n      search_query: `all:${query}`,\n      max_results: String(maxResults),\n      sortBy: 'relevance',\n    });\n    const url = `http://export.arxiv.org/api/query?${params}`;\n    log.debug('arxiv search', { query });\n\n    const response = await fetch(url, { signal: AbortSignal.timeout(timeoutMs) });\n    if (!response.ok) throw new Error(`arXiv returned ${response.status}`);\n\n    const xml = await response.text();\n    const mapped = this.parseAtom(xml);\n    return applyDateFilter(mapped, options);\n  }\n\n  private parseAtom(xml: string): RawSearchResult[] {\n    const { document } = parseHTML(xml);\n    const entries = Array.from(document.querySelectorAll('entry'));\n    const results: RawSearchResult[] = [];\n    const total = entries.length;\n\n    for (let i = 0; i < total; i++) {\n      const entry = entries[i];\n      const id = textContent(entry.querySelector('id'));\n      const title = textContent(entry.querySelector('title'));\n      const summary = textContent(entry.querySelector('summary'));\n      const published = textContent(entry.querySelector('published'));\n      if (!id || !title) continue;\n\n      let published_date: string | undefined;\n      if (published) {\n        const d = new Date(published);\n        if (!isNaN(d.getTime())) published_date = d.toISOString();\n      }\n\n      results.push({\n        title,\n        url: id,\n        snippet: summary.slice(0, SNIPPET_LIMIT),\n        relevance_score: 1 - i / Math.max(total, 1),\n        engine: 'arxiv',\n        ...(published_date ? { published_date } : {}),\n      });\n    }\n\n    return results;\n  }\n}\n\nfunction applyDateFilter(results: RawSearchResult[], options: SearchEngineOptions): RawSearchResult[] {\n  if (!options.fromDate && !options.toDate) return results;\n  const from = options.fromDate ? new Date(options.fromDate).getTime() : -Infinity;\n  const to = options.toDate ? new Date(options.toDate).getTime() : Infinity;\n  if (isNaN(from) || isNaN(to)) return results;\n  return results.filter((r) => {\n    if (!r.published_date) return false;\n    const t = new Date(r.published_date).getTime();\n    if (isNaN(t)) return false;\n    return t >= from && t <= to;\n  });\n}\n"],"mappings":"AAAA,SAAS,iBAAiB;AAE1B,SAAS,oBAAoB;AAE7B,MAAM,MAAM,aAAa,QAAQ;AAEjC,MAAM,gBAAgB;AAEtB,SAAS,YAAY,IAA4B;AAC/C,UAAQ,IAAI,eAAe,IAAI,QAAQ,QAAQ,GAAG,EAAE,KAAK;AAC3D;AAEO,MAAM,YAAoC;AAAA,EAC/C,OAAO;AAAA,EAEP,MAAM,OAAO,OAAe,UAA+B,CAAC,GAA+B;AACzF,UAAM,YAAY,QAAQ,aAAa;AACvC,UAAM,aAAa,QAAQ,cAAc;AAEzC,UAAM,SAAS,IAAI,gBAAgB;AAAA,MACjC,cAAc,OAAO,KAAK;AAAA,MAC1B,aAAa,OAAO,UAAU;AAAA,MAC9B,QAAQ;AAAA,IACV,CAAC;AACD,UAAM,MAAM,qCAAqC,MAAM;AACvD,QAAI,MAAM,gBAAgB,EAAE,MAAM,CAAC;AAEnC,UAAM,WAAW,MAAM,MAAM,KAAK,EAAE,QAAQ,YAAY,QAAQ,SAAS,EAAE,CAAC;AAC5E,QAAI,CAAC,SAAS,GAAI,OAAM,IAAI,MAAM,kBAAkB,SAAS,MAAM,EAAE;AAErE,UAAM,MAAM,MAAM,SAAS,KAAK;AAChC,UAAM,SAAS,KAAK,UAAU,GAAG;AACjC,WAAO,gBAAgB,QAAQ,OAAO;AAAA,EACxC;AAAA,EAEQ,UAAU,KAAgC;AAChD,UAAM,EAAE,SAAS,IAAI,UAAU,GAAG;AAClC,UAAM,UAAU,MAAM,KAAK,SAAS,iBAAiB,OAAO,CAAC;AAC7D,UAAM,UAA6B,CAAC;AACpC,UAAM,QAAQ,QAAQ;AAEtB,aAAS,IAAI,GAAG,IAAI,OAAO,KAAK;AAC9B,YAAM,QAAQ,QAAQ,CAAC;AACvB,YAAM,KAAK,YAAY,MAAM,cAAc,IAAI,CAAC;AAChD,YAAM,QAAQ,YAAY,MAAM,cAAc,OAAO,CAAC;AACtD,YAAM,UAAU,YAAY,MAAM,cAAc,SAAS,CAAC;AAC1D,YAAM,YAAY,YAAY,MAAM,cAAc,WAAW,CAAC;AAC9D,UAAI,CAAC,MAAM,CAAC,MAAO;AAEnB,UAAI;AACJ,UAAI,WAAW;AACb,cAAM,IAAI,IAAI,KAAK,SAAS;AAC5B,YAAI,CAAC,MAAM,EAAE,QAAQ,CAAC,EAAG,kBAAiB,EAAE,YAAY;AAAA,MAC1D;AAEA,cAAQ,KAAK;AAAA,QACX;AAAA,QACA,KAAK;AAAA,QACL,SAAS,QAAQ,MAAM,GAAG,aAAa;AAAA,QACvC,iBAAiB,IAAI,IAAI,KAAK,IAAI,OAAO,CAAC;AAAA,QAC1C,QAAQ;AAAA,QACR,GAAI,iBAAiB,EAAE,eAAe,IAAI,CAAC;AAAA,MAC7C,CAAC;AAAA,IACH;AAEA,WAAO;AAAA,EACT;AACF;AAEA,SAAS,gBAAgB,SAA4B,SAAiD;AACpG,MAAI,CAAC,QAAQ,YAAY,CAAC,QAAQ,OAAQ,QAAO;AACjD,QAAM,OAAO,QAAQ,WAAW,IAAI,KAAK,QAAQ,QAAQ,EAAE,QAAQ,IAAI;AACvE,QAAM,KAAK,QAAQ,SAAS,IAAI,KAAK,QAAQ,MAAM,EAAE,QAAQ,IAAI;AACjE,MAAI,MAAM,IAAI,KAAK,MAAM,EAAE,EAAG,QAAO;AACrC,SAAO,QAAQ,OAAO,CAAC,MAAM;AAC3B,QAAI,CAAC,EAAE,eAAgB,QAAO;AAC9B,UAAM,IAAI,IAAI,KAAK,EAAE,cAAc,EAAE,QAAQ;AAC7C,QAAI,MAAM,CAAC,EAAG,QAAO;AACrB,WAAO,KAAK,QAAQ,KAAK;AAAA,EAC3B,CAAC;AACH;","names":[]}

package/dist/search/engines/bing-news.d.ts

@@ -0,0 +1,7 @@
+import type { SearchEngine, SearchEngineOptions, RawSearchResult } from '../../types.js';
+export declare class BingNewsEngine implements SearchEngine {
+    name: string;
+    search(query: string, options?: SearchEngineOptions): Promise<RawSearchResult[]>;
+    parseResults(html: string, maxResults: number): RawSearchResult[];
+}
+//# sourceMappingURL=bing-news.d.ts.map

package/dist/search/engines/bing-news.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"bing-news.d.ts","sourceRoot":"","sources":["../../../src/search/engines/bing-news.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,YAAY,EAAE,mBAAmB,EAAE,eAAe,EAAE,MAAM,gBAAgB,CAAC;AAsDzF,qBAAa,cAAe,YAAW,YAAY;IACjD,IAAI,SAAe;IAEb,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,OAAO,GAAE,mBAAwB,GAAG,OAAO,CAAC,eAAe,EAAE,CAAC;IA2B1F,YAAY,CAAC,IAAI,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,GAAG,eAAe,EAAE;CAwClE"}