synthon

Chemical / Bio-Based Materials / Polymer / Life Sciences Industry Multi-Dimensional Deep Research Claude Code Skill

化工 / 生物基材料 / 聚合物 / 生命科学产业的多维度深度研究

Mehrdimensionale Tiefenforschung für Chemie- / Biobasierte Materialien / Polymer- / Biowissenschaften

Built on academic-search for CDP infrastructure. Goes beyond literature — integrates market data, patents, standards, production capacity, supply chains, and AI-driven materials discovery. Now with biopharma / CGT / biologics dimension.

基于 academic-search CDP 基础设施。超越文献——整合市场数据、专利、标准、产能产量、产业链与 AI 驱动材料发现。新增生物制药/CGT/生物工艺维度。

Auf academic-search für CDP-Infrastruktur aufgebaut. Geht über Literatur hinaus — integriert Marktdaten, Patente, Normen, Produktionskapazitäten, Lieferketten und KI-gestützte Materialentdeckung. Jetzt mit Biopharma / CGT / Biologics-Dimension.

Preamble: Network Environment Detection / 网络环境检测 / Netzwerkumgebung

Users in China / 国内用户 / Nutzer in China

⚠️ CNKI must NOT use VPN / CNKI 禁止 VPN / CNKI darf KEIN VPN verwenden — VPN triggers HTTP 418 anti-bot.

International Users / 国际用户 / Internationale Nutzer

All sources directly accessible. / 所有源可直接访问。 / Alle Quellen direkt zugänglich.

Quick Routing / 快速路由 / Schnellnavigation

User Request / 用户需求 / Benutzeranfrage
├─ "latest papers" / "最新论文" → W1: OpenAlex
├─ "high-citation reviews" / "高引综述" → W2: GS CDP (VPN required / 需VPN)
├─ "CNKI" / "知网" / "学位论文" → W3: CNKI CDP (NO VPN! / 禁止VPN!)
├─ "annual report" / "年报" / "招股书" → W4: PyMuPDF + W5b: Batch Scraping
├─ "market" / "市场" / "产能" → W6: Sogou Enterprise Search + W1 OpenAlex
├─ "company news" / "企业新闻" / "合作" → W6: Sogou Enterprise Search
├─ "patents" / "专利" → W7: Patent Search
├─ "AI+materials" / "AI+材料" / "AI+医药" / "AI+化工" → W8: AI + All Industries
├─ "investment" / "IPO" / "M&A" / "投资" → W9: Investment Research
├─ "regulations" / "法规" / "政策" → W10: Regulatory Search
├─ "supply chain" / "产业链" → W11: Supply Chain Mapping
├─ "industrial park" / "产业园" / "化工区" / "高新区" / "产业聚集" → W12: Industrial Clusters & Parks
│
├─ 🏭 化工/材料 → W1→W2→W3 merge | W6 Sogou | W7 专利 | W12 化工园区
│   示例: BASF, Dow, DuPont, Evonik, Covestro, 万华化学, 恒力石化
├─ 🧬 生物制药+生物技术 → Upstream/Downstream/Analytics/CGT/Industrial biotech/CDMO
│   示例: Roche, Novartis, Pfizer, Genomatica, Ginkgo Bioworks, 恒瑞医药, 百济神州, 药明生物
├─ 💰 投资分析 → W4 招股书 + W6 新闻 + W9 财务提取 + W9d 竞对标
│   示例: 宁德时代(CATL), 比亚迪, 小米集团
├─ 🧪 精细化工 → W6 供应链 + W7 专利 + W11 产业链
│   示例: DSM, Symrise, Givaudan, IFF, Lonza, 新和成
├─ 🌾 农用化学品 → W1 文献 + W6 登记动态 + W10 各国法规
│   示例: Syngenta, Bayer CropScience, Corteva, UPL, 扬农化工
├─ 🔋 能源材料 → W1 文献 + W6 产能新闻 + W7 专利 + W12 产业集群
│   示例: 宁德时代, 比亚迪, LG Energy Solution, 天赐材料, 恩捷股份
├─ 🍽️ 食品科技 → W1 文献 + W9 投融资 + W7 专利
│   示例: Cargill, ADM, Nestlé, Impossible Foods, 伊利
├─ 💄 化妆品原料 → W1 文献 + W6 市场 + W7 专利 + W10 法规
│   示例: L'Oréal, Estée Lauder, 华熙生物, 珀莱雅
├─ 🏗️ 产业聚集区 → W12 集群分布 + W10 差异化政策 + W6 园区新闻
│   示例: 上海化工区, 南京江北新材料科技园, 泰州医药城, 张江药谷, BASF Ludwigshafen, Singapore Jurong Island, Houston Ship Channel, Frankfurt Höchst
├─ 🤖 AI+产业 → W8 技术文献 + W6 应用新闻 + W9 投融资 + 跨行业交叉
└─ "industry panorama" / "产业全景" → ALL → merge

Workflow W1: OpenAlex API Search

No CDP required. No API key. No rate limit. 无 CDP / 无 API Key / 无速率限制。 Kein CDP erforderlich. Kein API-Schlüssel. Keine Ratenbegrenzung.

curl -s "https://api.openalex.org/works?filter=title_and_abstract.search:KEYWORD,publication_year:2023-2026&sort=cited_by_count:desc&per_page=20&select=id,doi,title,publication_date,cited_by_count,authorships,open_access" \
  -H "User-Agent: synthon/0.3.0 (mailto:your@email.com)" \
  -o /tmp/oa_results.json

Traps / 陷阱 / Fallen: OA-01 (only academic journals / 学术期刊限定 / nur akademische Zeitschriften), OA-02 (pre-2022 sparse / 2022前覆盖稀疏 / vor 2022 lückenhaft), OA-03 (100k/day limit / 100k/天限制 / 100k/Tag-Limit)

Workflow W2: Google Scholar CDP Search

Requires CDP + VPN (China users). See [[scholar-chinese-citations]] for Chinese citation regex. 需 CDP + VPN（国内用户）。中文引用提取见 [[scholar-chinese-citations]]。 Erfordert CDP + VPN (China-Nutzer). Siehe [[scholar-chinese-citations]] für chinesische Zitations-Regex.

PROXY="http://127.0.0.1:3456"
T=$(curl -s "$PROXY/new?url=https://scholar.google.com/scholar?q=KEYWORD&as_ylo=2023&hl=en" \
  | node -p "JSON.parse(require('fs').readFileSync(0,'utf8')).targetId")
sleep 3

curl -s -X POST "$PROXY/eval?target=$T" -d '
JSON.stringify(Array.from(document.querySelectorAll(".gs_r.gs_or.gs_scl")).map(el => ({
  allText: el.textContent.slice(0, 600)
})))' -o /tmp/gs_results.json

# Python parsing — matches both "Cited by X" and "被引用次数：X"
# Python解析 — 同时匹配 "Cited by X" 和 "被引用次数：X"
# Python-Parsing — erkennt sowohl "Cited by X" als auch "被引用次数：X"

curl -s "$PROXY/close?target=$T"

Traps: GS-01~04 (full-width colon / 全角冒号 / vollbreiter Doppelpunkt, <a> tag citations / <a> 标签引用 / <a>-Tag-Zitate, JS filter failure / JS filter 失效 / JS-Filter-Versagen, VPN IP blocked / VPN IP 被屏蔽 / VPN-IP blockiert)

Workflow W3: CNKI CDP Search

⚠️ NO VPN — direct connection only. VPN triggers HTTP 418. ⚠️ 禁止 VPN — 仅直连。 VPN 触发 HTTP 418。 ⚠️ KEIN VPN — nur Direktverbindung. VPN löst HTTP 418 aus.

PROXY="http://127.0.0.1:3456"

# 1. HTTP homepage (avoid SSL CNKI-01) / HTTP 主页 (避免 SSL CNKI-01) / HTTP-Startseite (SSL CNKI-01 vermeiden)
CNKI=$(curl -s "$PROXY/new?url=http://www.cnki.net" \
  | node -p "JSON.parse(require('fs').readFileSync(0,'utf8')).targetId")
sleep 5

# 2. location.href jump to KNS8 (don't use /navigate — CNKI-02) / location.href 跳转 KNS8 (不用 /navigate — CNKI-02)
curl -s -X POST "$PROXY/eval?target=$CNKI" -d \
  'location.href = "https://kns.cnki.net/kns8s/defaultresult/index?korder=SU&kw=URL_ENCODED_KEYWORD"' > /dev/null
sleep 8

# 3. Extract (KNS8 selectors, validated 2026-05) / 提取 (KNS8 选择器, 2026-05 验证) / Extrahieren (KNS8-Selektoren, validiert 2026-05)
curl -s -X POST "$PROXY/eval?target=$CNKI" -d '
JSON.stringify({
  totalCount: document.querySelector("#countPageDiv .countText")?.textContent?.trim(),
  papers: Array.from(document.querySelectorAll(".result-table-list tbody tr")).map(tr => ({
    title:   tr.querySelector("td.name a")?.textContent?.trim(),
    authors: tr.querySelector("td.author")?.textContent?.trim(),
    source:  tr.querySelector("td.source a")?.textContent?.trim(),
    date:    tr.querySelector("td.date")?.textContent?.trim(),
    cites:   tr.querySelector("td.quote a")?.textContent?.trim()
  }))
})' -o /tmp/cnki_results.json

curl -s "$PROXY/close?target=$CNKI"

⚠️ Pagination trap / 翻页陷阱 / Paginierungsfalle: .page-next may return page 1 results (KNS8 bug, 2026-05).

Traps: CNKI-01~08 (SSL, /navigate, CAPTCHA, URL encoding, 10-page limit, KNS8 search box, pagination failure, VPN triggers 418)

Reference / 参考 / Referenz: [[cnki-kns8-selectors]]

Workflow W4: PyMuPDF Industrial PDF Extraction

Annual report / patent full-text extraction. 年报 / 专利全文提取。 Geschäftsbericht / Patent-Volltextextraktion.

import fitz, tempfile, os

doc = fitz.open("report.pdf")
out = os.path.join(tempfile.gettempdir(), 'pdf_output.txt')

with open(out, 'w', encoding='utf-8') as f:
    for i in range(len(doc)):
        f.write(f'\n===== PAGE {i+1} =====\n')
        f.write(doc[i].get_text())
        f.flush()  # Memory safety for large PDFs / 大PDF内存安全 / Speichersicherheit

doc.close()
# NEVER print() Chinese — always write to UTF-8 file (PDF-01)
# 绝不 print() 中文 — 始终写 UTF-8 文件 (PDF-01)
# NIEMALS print() für Chinesisch — imm