ManuView 🔬📄

Open-Source AI Pre-Submission Manuscript Review & Scientific Diagnostic Suite
Because scientific feedback and education should be free, accessible, and transparent for every researcher worldwide.

---

🌟 Mission

Commercial pre-submission review and academic consulting platforms charge researchers $39 to $1,800+ per paper simply to catch the obvious structural, methodological, and citation issues that trigger journal desk rejections.

ManuView is an open-source, community-driven platform built to democratize pre-submission peer-review diagnostics. It provides the same reviewer-calibrated rigor for free, with zero paywalls, complete transparency, and local-first privacy.

---

🎯 What ManuView Does (And What It Doesn't)

Feature	Sentence-Level Grammar Tools (Grammarly, Paperpal)	Generic LLMs (ChatGPT, Claude)	ManuView (Open Source)
Diagnostic Focus	Commas, typos, passive voice	Conversational text summaries	Structural, scientific, & methodological soundness
Reviewer Posture	Mechanical spelling fixes	Agreeable / flattering bias	Calibrated editorial & peer-reviewer scrutiny
Failure Detection	Punctuation	Surface-level prose commentary	Causal overclaims, missing controls, statistical power gaps, desk-reject risks
Citations	Style format check only	Frequently hallucinates papers	Real-time verification against Crossref, OpenAlex, & Retraction Watch
Simulated Personas	None	Single-prompt chat	Multi-stage 5-Persona Reviewer Simulation
Data Privacy	Cloud servers	Cloud / training retention	Local-first (run via Ollama/vLLM) or private API keys

---

🚀 Key Modules & Capabilities

1. The Pre-Submission Diagnostic Engine

• The 6 Scoring Dimensions (1–5 scale):
  1. Originality: Checks stated contributions against current literature; flags incremental-only or derivative claims.
  2. Importance & Broad Interest: Evaluates whether the narrative connects to field-level questions rather than niche technicalities.
  3. Strength of Claims vs. Evidence: Detects causal overclaims (e.g. correlational observations phrased as causal mechanisms; claiming "demonstrates" without requisite controls).
  4. Methodological Soundness: Verifies sample size power, blinding, randomization, negative/positive controls, and figure-data consistency.
  5. Clarity & Presentation: Evaluates abstract structure (Problem → Gap → Approach → Finding → Impact) and logical flow.
  6. Prior Work & Reference Integrity: Flags outdated citations, high self-citation ratios (>25%), and missing landmark papers.
• Prioritized Action Plan:
  • 🚨 Priority A (Must-Fix): Flaws that will trigger immediate editorial desk rejections.
  • ⚠️ Priority B (Should-Fix): Major technical challenges peer reviewers will raise.
  • 💡 Priority C (Worth-Improving): Presentation and contextual enhancements.

2. The 5-Persona Reviewer Simulator

• Methods Reviewer: Scrutinizes protocols, sample sizes, experimental controls, reagents, and code/data reproducibility.
• Domain Expert: Assesses novelty, relevance to the field, benchmark comparisons, and biological/theoretical significance.
• Journal Editor: Evaluates scope alignment, target readership interest, and desk-rejection hazards.
• Statistician: Audits distribution assumptions, multiplicity corrections, p-hacking risks, and error bar definitions.
• Devil's Advocate: Attacks rival hypotheses, unruled-out confounders, and overclaimed causal mechanisms.

3. Citation & Reference Integrity Scanner

• Hallucinated Reference Detection: Resolves DOIs in real-time against Crossref and OpenAlex to catch hallucinated AI citations.
• Retraction Watch Integration: Flags any references that have been retracted, expressed with concern, or corrected.
• Citation Claim Checker: Compares an in-text assertion against the cited paper's abstract to confirm if the claim is Supported, Partially Supported, or Unsupported.

4. Modular Researcher Submission Tools

1. Journal Fit Predictor: Analyzes title + abstract against an open corpus of 1,300+ journals to rank reach, realistic, and fallback venues.
2. PRISMA 2020 Flow Diagram Generator: Interactive, client-side tool for systematic review counts that reconciles arithmetic and exports editable SVGs.
3. Journal Cover Letter Generator: Generates concise, editor-calibrated cover letters highlighting key discoveries and fit.
4. Graphical Abstract Checker: In-browser checker validating image dimensions, DPI, color space, and readability against target journal guidelines.
5. Submission Declaration Builder: Step-by-step generator for Data Availability, Competing Interests, CRediT Author Contributions, and Ethics statements.
6. Response to Reviewers Workspace: Converts decision letters into structured, point-by-point rebuttal matrices and revision checklists.

---

🔒 Privacy & Data Safety

Unpublished research is sensitive intellectual property. ManuView is designed with strict privacy principles:

• Local-First Support: Run the entire review pipeline on your local machine using open-source models (Llama 3, DeepSeek, Mistral) via Ollama or vLLM.
• Zero Data Training: When using remote API providers (Gemini, Anthropic, Groq, OpenAI), zero data is ever retained or used for model training.
• Ephemeral Processing: Ingested manuscripts are held only in memory during analysis and never persisted to external databases.

---

🛠️ Architecture & Tech Stack

graph TD
    User["Researcher / Student"] --> WebUI["Web UI (Next.js + Tailwind + shadcn/ui)"]
    WebUI --> API["API Layer (Next.js Server Actions / FastAPI)"]
    
    subgraph Core Engines
        Parser["Document Parser (PDF / DOCX Extraction)"]
        VerifyEngine["Reference Verification Engine (Crossref + OpenAlex + Retraction Watch)"]
        DiagEngine["Multi-Stage Diagnostic Engine (Reviewer Rubrics)"]
        PersonaSim["5-Persona Reviewer Simulator"]
    end
    
    API --> Parser
    API --> VerifyEngine
    API --> DiagEngine
    API --> PersonaSim
    
    subgraph External Open Data APIs
        Crossref["Crossref REST API (Free)"]
        OpenAlex["OpenAlex API (Free)"]
        RetractWatch["Retraction Watch Open DB"]
    end
    
    VerifyEngine --> Crossref
    VerifyEngine --> OpenAlex
    VerifyEngine --> RetractWatch
    
    subgraph Model Agnostic AI Layer
        LLMRouter["Model Router (Local Ollama / Gemini / Groq / OpenAI / Anthropic)"]
    end
    
    DiagEngine --> LLMRouter
    PersonaSim --> LLMRouter

Loading

• Frontend: Next.js 14/15 (App Router), React, TypeScript, Tailwind CSS, Lucide Icons.
• Document Ingestion: pdfplumber, PyMuPDF, mammoth.js, pdf-parse.
• Bibliographic Verification: Crossref REST API, OpenAlex REST API, Retraction Watch database.
• LLM Engine: Multi-provider support (Ollama for 100% offline local privacy, or API keys for cloud providers).

---

📋 Roadmap

• Initial repository structure & architecture blueprint
• Manuscript Ingestion Engine (PDF & DOCX section extractor)
• Reference Integrity & Retraction Verification Engine
• 6-Dimension Diagnostic Scoring Engine
• 5-Persona Reviewer Simulator
• Standalone Submission Tools (Journal Fit, PRISMA, Cover Letter, Rebuttal Builder)
• Report Export (.docx, .pdf, and interactive web report)
• One-click Docker setup for 100% offline local deployment

---

🤝 Contributing

Contributions from researchers, software engineers, editors, and peer reviewers are warmly welcomed! Please read our Contributing Guide to get started.

---

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.