feat(chat): add support for inline magazine sections and enhance content extraction
- Updated ChatMessage component to display a button for viewing brief magazine sections when available. - Enhanced useContentPanel to extract and manage magazine sections, improving content organization. - Refactored extractMagazineSections function to comprehensively handle various content formats, including headings and bullet points. - Improved formatting of content to preserve bold text and newlines as paragraphs. Made-with: Cursor
This commit is contained in:
Dorian
@@ -95,22 +95,106 @@ function extractFirstImageFromText(text: string): string | undefined {
|
||||
return ext ? ext[1] : undefined
|
||||
}
|
||||
|
||||
/** Extract magazine-style bullets: - **Title**: Content or - **Title** — Content */
|
||||
const MAGAZINE_CONTENT_MAX = 2000
|
||||
|
||||
function addSection(
|
||||
sections: MagazineSection[],
|
||||
title: string,
|
||||
content: string,
|
||||
seen: Set<string>,
|
||||
): void {
|
||||
const t = title.trim().replace(/\s+/g, ' ').slice(0, 150)
|
||||
const c = content.trim().replace(/\n{2,}/g, '\n\n').slice(0, MAGAZINE_CONTENT_MAX)
|
||||
if (t.length < 2 || c.length < 15) return
|
||||
const key = `${t.slice(0, 50)}`
|
||||
if (seen.has(key)) return
|
||||
seen.add(key)
|
||||
sections.push({
|
||||
title: t,
|
||||
content: c,
|
||||
url: extractUrlFromText(content),
|
||||
author: extractAuthorFromText(content),
|
||||
imageUrl: /!\[[^\]]*\]\((https?:\/\/[^)]+)\)/.exec(content)?.[1],
|
||||
})
|
||||
}
|
||||
|
||||
/** Extract magazine sections comprehensively: ## headings, **camp** blocks, bullets, intro */
|
||||
function extractMagazineSections(text: string): MagazineSection[] {
|
||||
const sections: MagazineSection[] = []
|
||||
const re = /^\s*[-•]\s*\*\*([^*]+)\*\*\s*[:\u2014\u2013–]\s*(.+?)(?=\n\s*[-•]\s*\*\*|\n\s*[-•]\s*[^*]|\n\n##\s|$)/gms
|
||||
let match: RegExpExecArray | null
|
||||
while ((match = re.exec(text)) !== null) {
|
||||
const title = match[1].trim().slice(0, 120)
|
||||
const raw = match[2].trim().replace(/\n+/g, ' ')
|
||||
const content = raw.slice(0, 800)
|
||||
if (title.length > 1 && content.length > 10) {
|
||||
const url = extractUrlFromText(raw)
|
||||
const author = extractAuthorFromText(raw)
|
||||
const imageUrl = /!\[[^\]]*\]\((https?:\/\/[^)]+)\)/.exec(raw)?.[1]
|
||||
sections.push({ title, content, url, author, imageUrl })
|
||||
const seen = new Set<string>()
|
||||
const blockContents = new Set<string>() // avoid duplicate bullets already in ## blocks
|
||||
|
||||
// 1. ## Heading blocks: full content until next ## or **Section**
|
||||
const headingRe = /^##\s*[^\n]*?(?:⚡|🔥|📌|✨)?\s*(.+?)\n\n([\s\S]+?)(?=\n##\s|\n\*\*[^*]+\*\*[🟠🔴🟢]?|\n\nFor deeper|\n\nThis is being|\z)/gim
|
||||
let m: RegExpExecArray | null
|
||||
while ((m = headingRe.exec(text)) !== null) {
|
||||
const title = m[1].trim().replace(/^[⚡🔥📌✨]\s*/, '').split(':')[0].trim()
|
||||
const content = m[2].trim()
|
||||
if (content.length > 20) {
|
||||
blockContents.add(content)
|
||||
addSection(sections, title, content, seen)
|
||||
}
|
||||
}
|
||||
|
||||
// 2. **Pro/ Anti camp** blocks with their bullets
|
||||
const campRe = /\*\*([^*]+(?:camp| side| view)[^*]*)\*\*[🟠🔴🟢]?\s*\n([\s\S]+?)(?=\n\*\*[^*]+(?:camp| side)[^*]*\*\*|\n##\s|\n\nThis is|\n\nFor deeper|\z)/gim
|
||||
while ((m = campRe.exec(text)) !== null) {
|
||||
const title = m[1].trim()
|
||||
const content = m[2].trim()
|
||||
if (content.length > 15) addSection(sections, title, content, seen)
|
||||
}
|
||||
|
||||
// 3. Bullets: - **Title**: Content (skip if already in a ## block)
|
||||
const bulletRe = /^\s*[-•]\s*\*\*([^*]+)\*\*\s*[:\u2014\u2013–]\s*(.+?)(?=\n\s*[-•]\s*\*\*|\n\s*[-•]\s+\w|\n\n##\s|\n\*\*[^*]+\*\*[🟠🔴]?|\z)/gms
|
||||
while ((m = bulletRe.exec(text)) !== null) {
|
||||
const raw = m[2].trim().replace(/\n+/g, ' ')
|
||||
if (raw.length < 15) continue
|
||||
const inBlock = [...blockContents].some((b) => b.includes(raw.slice(0, 100)))
|
||||
if (!inBlock) addSection(sections, m[1].trim(), raw, seen)
|
||||
}
|
||||
|
||||
// 4. - **Name** (Role) description — attributed quotes
|
||||
const attrRe = /^\s*[-•]\s*\*\*([^*]+)\*\*\s*\(([^)]+)\)\s+([^-\n].+?)(?=\n\s*[-•]|\n\*\*|\n##\s|\z)/gms
|
||||
while ((m = attrRe.exec(text)) !== null) {
|
||||
const content = m[3].trim().replace(/\n+/g, ' ')
|
||||
if (content.length > 20) addSection(sections, `${m[1].trim()} (${m[2].trim()})`, content, seen)
|
||||
}
|
||||
|
||||
// 5. Intro paragraph before first ## or ---
|
||||
const introMatch = /^([\s\S]+?)(?=\n##\s|\n---\s*\n|\n-\s+\*\*[^*]+\*\*\s*[:\u2014])/m.exec(text)
|
||||
if (introMatch) {
|
||||
const intro = introMatch[1].trim().replace(/^[#*_\s-]+/gm, '').trim()
|
||||
if (intro.length > 60 && !seen.has('Summary')) {
|
||||
addSection(sections, 'Summary', intro, seen)
|
||||
}
|
||||
}
|
||||
|
||||
// 6. "This is being called..." / closing paragraph
|
||||
const closingMatch = /(This is being called[^.]+\.[^"]*"[^"]+"[^.]*\.)/i.exec(text)
|
||||
if (closingMatch && !seen.has('Key')) {
|
||||
addSection(sections, 'Key takeaway', closingMatch[1].trim(), seen)
|
||||
}
|
||||
|
||||
// 7. "For deeper analysis" / podcast or further reading block
|
||||
const deeperMatch = /(?:For deeper analysis[^:]*:[\s\S]+?)(?=\n\z)/im.exec(text)
|
||||
if (deeperMatch) {
|
||||
const block = deeperMatch[1].trim()
|
||||
if (block.length > 30 && !seen.has('Further')) {
|
||||
addSection(sections, 'Further reading', block, seen)
|
||||
}
|
||||
}
|
||||
|
||||
// Preserve document order: Summary first, then ## blocks order, then camps, then key/further
|
||||
const order = ['Summary', 'Key takeaway', 'Further reading']
|
||||
sections.sort((a, b) => {
|
||||
const ai = order.indexOf(a.title)
|
||||
const bi = order.indexOf(b.title)
|
||||
if (ai >= 0 && bi >= 0) return ai - bi
|
||||
if (ai >= 0) return -1
|
||||
if (bi >= 0) return 1
|
||||
return 0
|
||||
})
|
||||
|
||||
return sections
|
||||
}
|
||||
|
||||
@@ -529,7 +613,7 @@ export function useContentPanel() {
|
||||
|
||||
// Magazine = bullet-style sections (- **Title**: Content)
|
||||
const magazineSections = extractMagazineSections(text)
|
||||
const hasMagazine = magazineSections.length >= 2 && (isNewsQuery(userQuery) || isNewsLikeResponse(text) || /sentiment|bearish|bull case|macro|%|BTC|bitcoin|BIP|protocol|debate|what'?s happening/i.test(text))
|
||||
const hasMagazine = magazineSections.length >= 1 && (isNewsQuery(userQuery) || isNewsLikeResponse(text) || /sentiment|bearish|bull case|macro|%|BTC|bitcoin|BIP|protocol|debate|what'?s happening/i.test(text))
|
||||
|
||||
// News = actual articles (web search + RSS from website domains). Plain links → Websites.
|
||||
const newsContext = isNewsQuery(userQuery) || isNewsLikeResponse(text)
|
||||
@@ -620,7 +704,7 @@ export function useContentPanel() {
|
||||
const songs = extractAllSongs(text)
|
||||
const podcasts = extractAllPodcasts(text)
|
||||
const magazineSections = extractMagazineSections(text)
|
||||
const hasMagazine = magazineSections.length >= 2 && (isNewsQuery(userQuery) || isNewsLikeResponse(text) || /sentiment|bearish|bull case|macro|%|BTC|bitcoin|BIP|protocol|debate|what'?s happening/i.test(text))
|
||||
const hasMagazine = magazineSections.length >= 1 && (isNewsQuery(userQuery) || isNewsLikeResponse(text) || /sentiment|bearish|bull case|macro|%|BTC|bitcoin|BIP|protocol|debate|what'?s happening/i.test(text))
|
||||
const fromMarkdown = extractMarkdownLinks(text)
|
||||
const boldDomains = extractBoldDomainLinks(text)
|
||||
const hasNews = webResults.length > 0 && (isNewsQuery(userQuery) || isNewsLikeResponse(text))
|
||||
|
||||
Reference in New Issue
Block a user