Include extraction runtime modules required by webhook bot

Author: apartsin

telegram/src/intent-routing.js

@@ -0,0 +1,56 @@
+const { classifyMessageInput, inferLikelyWebUrlFromText } = require('./message-utils');
+
+const NON_TEXT_SOURCE_TYPES = new Set([
+  'html_url',
+  'pdf_url',
+  'pdf_file',
+  'image_url',
+  'image_file',
+  'audio_url',
+  'audio_file'
+]);
+
+function decideInputIntent({ incomingKind, text, sourceType, shortPromptMaxChars }) {
+  const normalizedText = String(text || '').trim();
+  const normalizedSource = String(sourceType || '').trim().toLowerCase();
+  const isNonTextSource = NON_TEXT_SOURCE_TYPES.has(normalizedSource);
+  const parsed = classifyMessageInput(normalizedText);
+  const isShortText = !isNonTextSource
+    && String(incomingKind || '').trim().toLowerCase() === 'text'
+    && normalizedText.length > 0
+    && normalizedText.length <= Math.max(1, Number(shortPromptMaxChars || 120));
+
+  let inferredUrl = '';
+  if (isShortText && parsed.kind !== 'url') {
+    inferredUrl = String(inferLikelyWebUrlFromText(normalizedText) || '').trim();
+  }
+
+  let route = 'text';
+  let reason = 'default_text';
+  if (isNonTextSource) {
+    route = 'source_extraction';
+    reason = `non_text_source:${normalizedSource}`;
+  } else if (parsed.kind === 'url' || inferredUrl) {
+    route = 'url';
+    reason = parsed.kind === 'url' ? 'input_contains_url' : 'short_text_url_inferred';
+  } else if (isShortText) {
+    route = 'invent';
+    reason = 'short_text_needs_story_expansion';
+  }
+
+  return {
+    route,
+    reason,
+    parsedKind: parsed.kind,
+    parsedValue: String(parsed.value || ''),
+    isShortText,
+    isNonTextSource,
+    inferredUrl
+  };
+}
+
+module.exports = {
+  NON_TEXT_SOURCE_TYPES,
+  decideInputIntent
+};
+

telegram/src/message-utils.js

@@ -56,7 +56,33 @@ function isLongStoryText(value) {
 function isLikelyMediaUrl(value) {
   const url = String(value || '').trim().toLowerCase();
   if (!url) return false;
-  return /\.(png|jpe?g|gif|webp|bmp|svg|mp4|mov|avi|mkv|webm|m4v|wmv|flv)(\?.*)?$/.test(url);
+  return /\.(png|jpe?g|gif|webp|bmp|svg|mp4|mov|avi|mkv|webm|m4v|wmv|flv|mp3|wav|ogg|m4a|aac|flac|opus)(\?.*)?$/.test(url);
+}
+
+function isLikelyImageUrl(value) {
+  const url = String(value || '').trim().toLowerCase();
+  if (!url) return false;
+  return /\.(png|jpe?g|gif|webp|bmp|svg)(\?.*)?$/.test(url);
+}
+
+function isLikelyAudioUrl(value) {
+  const url = String(value || '').trim().toLowerCase();
+  if (!url) return false;
+  return /\.(mp3|wav|ogg|m4a|aac|flac|opus)(\?.*)?$/.test(url);
+}
+
+function isLikelyPdfUrl(value) {
+  const candidate = normalizeUrlCandidate(value);
+  if (!looksLikeUrl(candidate)) return false;
+  try {
+    const parsed = new URL(candidate);
+    const joined = `${parsed.pathname || ''}${parsed.search || ''}`.toLowerCase();
+    if (joined.includes('.pdf')) return true;
+    const mimeHint = String(parsed.searchParams.get('format') || '').toLowerCase();
+    return mimeHint === 'pdf' || mimeHint === 'application/pdf';
+  } catch (_) {
+    return /\.pdf(\?.*)?$/i.test(candidate);
+  }
 }
 
 function isLikelyWebPageUrl(value) {
@@ -106,6 +132,60 @@ function extractTextFallbackFromUrlMessage(value) {
   return candidate;
 }
 
+function extractFirstPdfUrlLikeToken(value) {
+  const text = String(value || '').trim();
+  if (!text) return '';
+  const explicit = extractFirstUrl(text);
+  if (explicit && isLikelyPdfUrl(explicit)) return explicit;
+  const tokens = text.split(/\s+/).map((t) => normalizeUrlCandidate(t)).filter(Boolean);
+  for (const token of tokens) {
+    const inferred = inferLikelyWebUrlFromText(token);
+    if (inferred && isLikelyPdfUrl(inferred)) return inferred;
+  }
+  return '';
+}
+
+function extractFirstImageUrlLikeToken(value) {
+  const text = String(value || '').trim();
+  if (!text) return '';
+  const explicit = extractFirstUrl(text);
+  if (explicit && isLikelyImageUrl(explicit)) return explicit;
+  const tokens = text.split(/\s+/).map((t) => normalizeUrlCandidate(t)).filter(Boolean);
+  for (const token of tokens) {
+    const inferred = inferLikelyWebUrlFromText(token);
+    if (inferred && isLikelyImageUrl(inferred)) return inferred;
+  }
+  return '';
+}
+
+function extractFirstAudioUrlLikeToken(value) {
+  const text = String(value || '').trim();
+  if (!text) return '';
+  const explicit = extractFirstUrl(text);
+  if (explicit && isLikelyAudioUrl(explicit)) return explicit;
+  const tokens = text.split(/\s+/).map((t) => normalizeUrlCandidate(t)).filter(Boolean);
+  for (const token of tokens) {
+    const inferred = inferLikelyWebUrlFromText(token);
+    if (inferred && isLikelyAudioUrl(inferred)) return inferred;
+    if (!inferred) {
+      let candidate = token;
+      if (!/^https?:\/\//i.test(candidate)) {
+        if (/^www\./i.test(candidate) || /^[a-z0-9.-]+\.[a-z]{2,}(?:[/:?#].*)?$/i.test(candidate)) {
+          candidate = `https://${candidate}`;
+        } else {
+          continue;
+        }
+      }
+      try {
+        const parsed = new URL(candidate);
+        const normalized = parsed.toString();
+        if (isLikelyAudioUrl(normalized)) return normalized;
+      } catch (_) {}
+    }
+  }
+  return '';
+}
+
 function extractLinksFromEntities(baseText, entities) {
   const text = String(baseText || '');
   const list = Array.isArray(entities) ? entities : [];
@@ -189,9 +269,15 @@ module.exports = {
   extractFirstUrl,
   classifyMessageInput,
   isLikelyWebPageUrl,
+  isLikelyPdfUrl,
+  isLikelyImageUrl,
+  isLikelyAudioUrl,
   extractTextFallbackFromUrlMessage,
   inferLikelyWebUrlFromText,
   extractMessageInputText,
-  extractFirstUrlLikeToken
+  extractFirstUrlLikeToken,
+  extractFirstPdfUrlLikeToken,
+  extractFirstImageUrlLikeToken,
+  extractFirstAudioUrlLikeToken
 };
 const { LONG_STORY_TEXT_MIN_CHARS } = require('./data/thresholds');