Configure DeepSeek API for Zotero: Academic Reading, Translation & Research
Leverage DeepSeek-V3 and R1 language understanding inside Zotero to handle PDF line-break formatting, synthesize paper abstracts, interrogate methodologies, and identify research gaps.
On this page
- API Base URL:
https://api.deepseek.comorhttps://api.deepseek.com/v1 - Recommended model for translation:
deepseek-chat(High speed, cost-effective) - Recommended model for reasoning:
deepseek-reasoner(R1 deep thinking) - API key: Generated in your DeepSeek Platform console.
Install Zotero AI extension plugin
In Zotero 7 (or 6), plugins like PapersGPT, Zotero GPT, or Awesome GPT allow seamless LLM integration:
- Download the latest
.xpipackage from the plugin's GitHub Releases page. - Open Zotero and click Tools → Plugins (Add-ons) from the top menu.
- Click the gear icon in the top right, select Install Add-on From File..., choose the downloaded
.xpifile, and restart Zotero.
Configure DeepSeek API in the plugin
- In Zotero, navigate to Settings / Preferences → Zotero GPT / PapersGPT.
- Set API Provider to Custom OpenAI or OpenAI Compatible.
- Set API Base URL to
https://api.deepseek.com/v1(orhttps://api.deepseek.com). - Paste your DeepSeek API Key (
sk-...). - Set Model Name to
deepseek-chat(ordeepseek-reasoner) and save.
Academic prompt templates
Configure these tailored prompts in your plugin's custom action menu for rapid research workflows:
Two-column PDF cleanup and translation
Automatically reconciles PDF formatting breaks and hyphens:
You are an expert academic translator. Translate the following excerpt from a scientific paper into clear, natural English/Chinese.
Requirements:
1. Merge broken lines and remove trailing hyphens caused by two-column PDF layouts.
2. Preserve exact scientific terminology and formula symbols.
3. Return only the clean translation without introductory conversational text.
Source text:
{text}Methodology extraction and gap analysis (For R1)
Deconstructs research contributions and potential caveats:
Analyze the following paper excerpt and provide a structured synthesis:
1. Core Problem: What exact research bottleneck does this work address?
2. Novel Methodology: What is the core mechanism and how does it differ from prior work?
3. Limitations & Research Gaps: What boundary conditions exist, and what future work is implied?
Paper excerpt:
{text}Offline privacy mode for unpublished research (Ollama)
For confidential drafts and proprietary research data, run DeepSeek locally with Ollama:
# Run DeepSeek locally with Ollama
ollama run deepseek-r1:7bIn Zotero plugin preferences, set the Base URL to http://127.0.0.1:11434/v1 with API Key set to ollama. All document text stays strictly on your local machine.
Troubleshooting
401 Unauthorized / Authentication Error
Verify that your key begins with sk- with no surrounding whitespace, and verify that your account balance is active.
Translation Timeout
Selecting massive text blocks simultaneously can exceed request timeouts. Select text in smaller sections or increase plugin timeout to 60s.