跳到主要内容
知仓学习社ZHICANG

onecite-reference-guide

AI toolkit to parse, complete, and format academic references

不碰外部(只输出文字)无严重或高危命中brycewang-stanford/Auto-Empirical-Research-Skills

它会碰到什么

扫了多少1 个文本文件,4 KB
它会碰到什么不碰外部(只输出文字)
命中总数2 处
命中统计严重 0 · 高 0 · 中 0 · 低 0

这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。

技能内容

OneCite Reference Guide

Overview

OneCite is an AI-powered toolkit for parsing, completing, and formatting academic references. Given incomplete or messy citation strings, it extracts structured metadata, fills in missing fields via API lookups (CrossRef, OpenAlex), and outputs clean formatted references in any style (APA, MLA, BibTeX, Chicago). Available as a Python library and MCP server for agent integration.

Installation

# Python package
pip install onecite

# MCP server
npx @onecite/mcp-server

Reference Parsing

from onecite import parse_reference

# Parse messy reference string
ref = parse_reference(
    "Vaswani et al. Attention Is All You Need. "
    "NeurIPS 2017. arXiv:1706.03762"
)

print(ref.title)      # "Attention Is All You Need"
print(ref.authors)    # ["Vaswani, A.", "Shazeer, N.", ...]
print(ref.year)       # 2017
print(ref.venue)      # "NeurIPS"
print(ref.arxiv_id)   # "1706.03762"
print(ref.doi)        # "10.48550/arXiv.1706.03762"

Metadata Completion

from onecite import complete_reference

# Fill in missing metadata from APIs
ref = complete_reference(
    title="Attention Is All You Need",
    # Automatically looks up: DOI, authors, venue,
    # abstract, citation count, pages, volume
)

print(f"DOI: {ref.doi}")
print(f"Authors: {', '.join(ref.authors)}")
print(f"Pages: {ref.pages}")
print(f"Volume: {ref.volume}")
print(f"Citations: {ref.citation_count}")

Format Conversion

from onecite import format_reference, parse_reference

ref = parse_reference(
    "B. Kerbl et al., '3D Gaussian Splatting for Real-Time "
    "Radiance Field Rendering,' SIGGRAPH 2023"
)

# Output in different styles
print(format_reference(ref, style="apa"))
# Kerbl, B., Kopanas, G., Leimkühler, T., & Drettakis, G.
# (2023). 3D Gaussian Splatting for Real-Time Radiance Field
# Rendering. ACM SIGGRAPH 2023.

print(format_reference(ref, style="bibtex"))
# @inproceedings{kerbl2023,
#   title = {3D Gaussian Splatting...},
#   author = {Kerbl, Bernhard and ...},
#   booktitle = {ACM SIGGRAPH 2023},
#   year = {2023}
# }

print(format_reference(ref, style="mla"))
print(format_reference(ref, style="chicago"))
print(format_reference(ref, style="ieee"))

Batch Processing

from onecite import process_references

# Process a list of raw reference strings
raw_refs = [
    "Vaswani et al. Attention Is All You Need. NeurIPS 2017",
    "Devlin et al. BERT. NAACL 2019",
    "Brown et al. Language Models are Few-Shot Learners. 2020",
]

results = process_references(
    raw_refs,
    complete=True,     # Fill missing metadata
    format="bibtex",   # Output format
    deduplicate=True,  # Remove duplicates
)

# Save as .bib file
with open("references.bib", "w") as f:
    for ref in results:
        f.write(ref.formatted + "\n\n")

MCP Server Usage

{
  "mcpServers": {
    "onecite": {
      "command": "npx",
      "args": ["@onecite/mcp-server"],
      "tools": [
        "parse_reference",
        "complete_reference",
        "format_reference",
        "search_paper",
        "get_bibtex"
      ]
    }
  }
}

Reference Validation

from onecite import validate_references

# Check a .bib file for issues
issues = validate_references("references.bib")

for issue in issues:
    print(f"[{issue.severity}] {issue.entry}: {issue.message}")
    # [WARNING] smith2023: Missing DOI
    # [ERROR] jones2024: Author field empty
    # [INFO] chen2022: Title case inconsistency

Use Cases

  1. Reference cleanup: Fix messy bibliographies
  2. Format conversion: Convert between citation styles
  3. Metadata completion: Fill in missing DOIs, pages, venues
  4. Agent integration: MCP server for AI citation workflows
  5. Validation: Check references for completeness and errors

References

想直接用这个技能?

本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。