github CYC2002tommy/Deep-Research-Agent

家里蹲国仙

We all know the drill. You ask an AI to help with a literature review, and it hands back fake DOIs, citations from predatory journals, and prose stuffed with words like "delve" and "tapestry."

It’s frustrating. Fact-checking the AI takes longer than writing the paper yourself.

That’s exactly why I spent the last few days building and open-sourcing the Deep Research Agent. Built on the Hermes/ECC framework, this isn't just a clever prompt. It’s an automated, multi-agent academic production line.

Here is how it actually works under the hood:

🔍 1. Parallel Sourcing (The 1,400+ Paper Funnel)
Feed it a topic. It dispatches two isolated subagents simultaneously—one hits the official Elsevier Scopus API, the other scans Open Science databases like OpenAlex. They pull over 1,400 papers combined, cross-reference them, and distill everything down to the 50 most relevant hits.

🛡️ 2. Academic Hygiene & Zero Hallucinations
Quality in, quality out. The system explicitly targets Q1-Q2 journals and outright bans MDPI and Q4 publications. Before writing a single word, a background script runs live curl tests on every generated DOI. Dead links get deleted instantly. Zero fake citations.

👨‍🏫 3. The "Remi" Peer Review
Named in tribute to my academic advisor, Remi Chauvy, I built an internal review loop. Once a draft generates, the "Remi" subagent acts as a ruthless peer reviewer. It strips out AI fluff, checks logic, and forces rewrites until the manuscript reads like an actual researcher wrote it.

⚙️ 4. End-to-End Output & Knowledge Syncing
It doesn't just dump a Markdown file into your chat. A Python script compiles a fully formatted .docx file, complete with APA 7th hanging indents and data charts. Finally, it logs the research summary directly to your Obsidian vault and syncs the abstracts to Google NotebookLM for future Q&A.

We just merged a major code quality update (v1.1.0) and the repo is fully public. If you are a researcher, grad student, or engineer tired of doing the heavy lifting for your AI, grab the code here:

🔗 GitHub: https://github.com/CYC2002tommy/Deep-Research-Agent

If this saves you a few dozen hours of literature hunting, a GitHub Star ⭐ would be amazing. Pull requests and feedback are always welcome!

#ArtificialIntelligence #LiteratureReview #AgenticAI #OpenSource #Scopus #NotebookLM #Python #Research

家里蹲国仙

1 yrs ago prompt

1. 
   
2. You are given various potential options or approaches for a project. Convert these into a  
   
3. well-structured research plan that:  
   
4. 
   
5. 1. Identifies Key Objectives  
   
6.    - Clarify what questions each option aims to answer  
   
7.    - Detail the data/info needed for evaluation  
   
8. 
   
9. 2. Describes Research Methods  
   
10.    - Outline how you’ll gather and analyze data  
    
11.    - Mention tools or methodologies for each approach  
    
12. 
    
13. 3. Provides Evaluation Criteria  
    
14.    - Metrics, benchmarks, or qualitative factors to compare options  
    
15.    - Criteria for success or viability  
    
16. 
    
17. 4. Specifies Expected Outcomes  
    
18.    - Possible findings or results  
    
19.    - Next steps or actions following the research  
    
20. 
    
21. Produce a methodical plan focusing on clear, practical steps.  
    
22. 
    
23. Your main subject is

复制代码