SKILL: Bohrium Paper & Patent Search

Overview

Search academic papers and patents using the Bohrium RAG search engine. Combines keyword matching with semantic understanding, supporting natural language queries, date range filtering, JCR zone/database filtering, and AI reranking.

Supported Search Types:

Use cases: Literature review, technical survey, method comparison, trend analysis.

No CLI support — all operations use the HTTP API.

Authentication

ACCESS_KEY is read from the OpenClaw config ~/.openclaw/openclaw.json:

"bohrium-paper-search": {
  "enabled": true,
  "apiKey": "YOUR_ACCESS_KEY",
  "env": {
    "ACCESS_KEY": "YOUR_ACCESS_KEY"
  }
}

OpenClaw automatically injects env.ACCESS_KEY into the runtime.

Common Code Template

import os, requests

AK = os.environ.get("ACCESS_KEY", "")
BASE = "https://open.bohrium.com/openapi/v1/paper"
HEADERS_JSON = {"accessKey": AK, "Content-Type": "application/json"}

English Paper Search

Basic Search

r = requests.post(f"{BASE}/rag/pass/keyword", headers=HEADERS_JSON, json={
    "words": ["deep learning", "molecular dynamics"],
    "question": "How to use deep learning for molecular dynamics simulation?",
    "startTime": "",
    "endTime": "",
    "pageSize": 10
})
data = r.json()
print(f"Found {len(data['data'])} papers")
for p in data["data"]:
    print(f"  [{p['doi']}] {p['enName']}")
    print(f"    Journal: {p.get('publicationEnName', '')}, IF: {p.get('impactFactor', 0)}")
    print(f"    Date: {p['coverDateStart']}, Citations: {p['citationNums']}")

Advanced Search (Date Range + JCR Zone + Database + Type)

r = requests.post(f"{BASE}/rag/pass/keyword", headers=HEADERS_JSON, json={
    "words": ["deep learning", "protein structure"],
    "question": "deep learning protein structure prediction",
    "type": 5,                          # Search version (see below)
    "startTime": "2024-01-01",          # Start date YYYY-MM-DD
    "endTime": "2025-01-01",            # End date
    "jcrZones": ["Q1", "Q2"],           # JCR zone filter
    "includeDbs": ["SCI"],              # Database filter
    "areaIds": [],                      # Area IDs (optional)
    "publicationIds": [],               # Publication IDs (optional)
    "pageSize": 20
})

Request Parameters

Response Fields

Patent Search

r = requests.post(f"{BASE}/rag/pass/patent", headers=HEADERS_JSON, json={
    "keyword": "neural network training optimization",
    "page": 1,
    "pageSize": 10
})
data = r.json()
for p in data:
    print(f"  Patent: {p}")

Note: Patent search API uses simple parameters. Advanced parameters like rerank, type, words, question are not supported (will cause backend errors).

Patent Request Parameters

Patent Response Fields

Returns array format with patent information objects.

Search Tips

Keyword Selection

# GOOD: 3-8 professional terms
words = ["molecular dynamics", "force field", "deep potential", "neural network"]

# BAD: Too generic
words = ["science", "research"]

Combine question for Better Relevance

words is for exact keyword matching, question is for semantic understanding. Best results come from combining both (for paper search only):

{
    "words": ["GNN", "molecular property", "prediction"],
    "question": "How do graph neural networks predict molecular properties?",
    "pageSize": 20
}

Filter for High-Quality Journals

{
    "words": ["..."],
    "question": "...",
    "jcrZones": ["Q1"],          # Q1 journals only
    "includeDbs": ["SCI"],       # SCI-indexed only
    "startTime": "2023-01-01",   # Recent 2 years
    "endTime": "2025-12-31"
}

curl Examples

AK="YOUR_ACCESS_KEY"
BASE="https://open.bohrium.com/openapi/v1/paper"

# English paper search
curl -s -X POST "$BASE/rag/pass/keyword" \
  -H "Content-Type: application/json" \
  -H "accessKey: $AK" \
  -d '{"words":["deep learning","protein"],"question":"deep learning protein structure prediction","type":5,"startTime":"2024-01-01","endTime":"2025-01-01","jcrZones":["Q1"],"pageSize":5}'

# Patent search
curl -s -X POST "$BASE/rag/pass/patent" \
  -H "Content-Type: application/json" \
  -H "accessKey: $AK" \
  -d '{"words":["neural network"],"question":"neural network training","type":3,"rerank":1,"pageSize":5}'

Troubleshooting