1. Getting Started

1.1 System Requirements

• OS: Windows 10/11 (64-bit), macOS (Apple Silicon & Intel), Linux (Ubuntu/Debian/Fedora)
• RAM: 8 GB minimum, 16 GB recommended for local LLM (7B models)
• Storage: ~500 MB app + additional for index, cache, and GGUF models
• Optional: Python 3.11+ for local AI mode

1.2 Quick Install

Download the latest release from GitHub Releases. The application bundles the Python backend for a seamless experience.

1.3 First Launch & Setup Wizard

On first run, ResearchMind will:

• Initialize the local SQLite database (researchmind.db)
• Create the ChromaDB vector index
• Launch the backend FastAPI server (auto‑starts)
• Prompt you to choose your AI provider (Cloud Free, API key, or Local)

1.4 Troubleshooting Installation

• Backend fails to start: Ensure Python 3.11+ is installed and ports 8000–8005 are free.
• Index rebuild: If search fails, go to Settings → System → Rebuild FTS Index.
• Logs: Check ~/.researchmind/logs/ for detailed error logs.

2. Import & Library

2.1 Supported Formats

• PDF (with OCR for scanned documents via RapidOCR)
• DOCX, EPUB, TXT, MD, HTML
• BibTeX (.bib) and RIS (.ris)
• Zotero (direct sync from SQLite)

2.2 Import Methods

• Drag & Drop: Drop files/folders directly into the Library view.
• Scan Folder: Import all supported files from a directory.
• Zotero Sync: Connect your Zotero account to sync metadata and PDFs.
• Import Metadata Only: Add papers by DOI, arXiv ID, or manual entry.

2.3 Import Queue & Status Tracking

Each file goes through a queue with real‑time progress:

• queued → parsing → embedding → indexing → OCR (if needed) → summarizing → ready
• Status is displayed per file; you can retry individual failed jobs.

2.4 Retry Import & Manual OCR

• Click Retry on any failed import to restart processing.
• Manual OCR: Trigger OCR again for scanned PDFs (useful if the first attempt failed).

2.5 BibTeX & RIS Import

ResearchMind parses standard BibTeX and RIS files to extract metadata (title, authors, year, journal, DOI). You can import these files along with or without associated PDFs.

2.6 Paper Management

• View / Edit: Click any paper to see details, edit metadata, add notes.
• Delete: Remove a paper from your library (also removes its chunks and vector entries).
• Star / Read Status: Mark papers as starred or read to track progress.
• Bulk Operations: Select multiple papers to add to collections, delete, or change status.

3. Collections & Organization

3.1 Creating & Managing Collections

• Create: Name your collection and optionally add a description.
• Edit / Delete: Update collection name/description or delete it (papers remain in library).
• Add/Remove Papers: Drag papers into a collection or use the bulk action menu.

3.2 Saved Searches

Save any search query with filters (collection, tag, year, type). Saved searches appear in the sidebar for quick access.

3.3 Tags & Filters

• Assign custom tags to papers.
• Filter by collection, tag, publication year, document type, read status, and starred.

3.4 Bulk Operations

Select multiple papers to add to a collection, delete, change read status, or apply tags.

4. Hybrid Search

ResearchMind combines three retrieval strategies for maximum relevance:

4.1 BM25 Full‑Text Search

Fast keyword matching using SQLite FTS5. Supports stemming, stop‑word removal, and prefix searches.

4.2 Vector Semantic Search

Dense vector retrieval using bge‑m3 embeddings stored in ChromaDB. Finds conceptually similar content even when keywords don't match.

4.3 Cross‑Encoder Reranking

An optional cross‑encoder model re‑ranks the top results from BM25 and vector search to improve precision.

4.4 Filters

• Collection membership
• Tags
• Publication year range
• Document type (PDF, DOCX, etc.)
• Read status / starred

4.5 Saved Searches

Save any search with its filters for one‑click re‑use.

5. AI Chat

5.1 Chat Modes

• Free Mode: General AI chat with optional paper context.
• Evidence‑Only Mode: Responses must be cited from your library.
• Critique Mode: AI critically analyses papers' methodologies and limitations.
• Debate Mode: Multiple AI personas debate a research question.

5.2 Multi‑Paper Conversations

Select multiple papers to include as context. The AI can compare and synthesize across them.

5.3 Citation Panel & Trust Panel

Each response includes source chips with page numbers. Click to expand the exact quote. The trust panel shows confidence scores for each claim.

5.4 Suggest Questions

AI generates recommended questions based on the paper(s) you're currently viewing.

5.5 Chat History & Usage

• View all past conversations.
• Delete individual chat sessions.
• Usage stats: number of messages, tokens consumed.

5.6 Reasoning Mode

• Fast Mode: Quick, concise answers.
• Deep (Reasoning) Mode: More thorough, step‑by‑step reasoning (higher latency, more tokens).

5.7 Multi‑Language Support

Chat supports English, Vietnamese, Japanese, and more (depends on provider). You can switch language in Settings.

6. Literature Review Builder

6.1 Generate Outline

Select a set of papers and ask the AI to generate a structured outline for your literature review.

6.2 Generate Draft (7 Sections)

Create a full draft with the standard sections: Background, Methods, Results, Discussion, Gap Analysis, Future Directions, and Conclusion.

6.3 Streaming Draft Generation

Watch the draft being written in real‑time, section by section.

6.4 Individual Section Generation & Regeneration

Regenerate any section independently with new instructions or after editing other sections.

6.5 Quality Check

AI evaluates the draft on coherence, completeness, citation coverage, and academic tone.

6.6 Save, Version, Restore

• Save multiple drafts per project.
• View version history and restore any previous version.

6.7 Export

• DOCX: For Microsoft Word.
• Markdown: For online publishing.
• LaTeX: For academic journals.

7. Evidence Matrix

7.1 Custom Criteria Definition

Define your own evaluation criteria (e.g., sample size, blinding, statistical power).

7.2 AI Claim Extraction

AI extracts relevant claims from each paper and maps them to your criteria, with source quotes.

7.3 Visual Comparison Table

Rows = criteria, columns = papers. Cells contain extracted claims with clickable quotes.

7.4 Save & Version Matrix

Save multiple matrices and track changes over time.

7.5 Export to CSV

Export the matrix for further analysis in Excel or other tools.

8. Insights & Analysis

8.1 Research Gap Analysis

Automatically detect gaps in your literature—topics that are underrepresented or conflicting.

8.2 Conflict Detection

Identify papers that contradict each other on methodology, findings, or conclusions.

8.3 Topic Suggestion

AI suggests new research topics based on your library's current coverage.

8.4 Research Evolution Map

Timeline view showing how a field has evolved over time (based on publication dates and citation networks).

8.5 Paper Comparison

Side‑by‑side comparison of two or more papers on multiple dimensions (methodology, results, limitations).

9. Academic Verification

9.1 DOI Lookup

Query OpenAlex, Crossref, and Semantic Scholar to fetch metadata for a DOI.

9.2 Claim Verification

Given a claim, the AI searches external sources to verify or refute it.

9.3 Paper Discovery

Find new papers related to your library via academic APIs.

9.4 PDF Proxy

Download publicly available PDFs through our proxy (for open‑access papers).

10. Deep Research

10.1 Deep Research Mode

AI performs multi‑step research: decomposes your question, searches, reads papers, synthesizes, and returns a comprehensive report.

10.2 Query Decomposition

Complex questions are broken into sub‑queries, each answered separately, then aggregated into a final answer.

11. Graph & Personal Brain

11.1 Paper Relationship Graph

Visual network of papers connected by embedding similarity, citations, co‑authors, and shared topics.

11.2 GraphRAG

• Global Search: Search across the entire knowledge graph.
• Local Search: Search within the neighbourhood of a specific paper.
• Drift Search: Detect emerging topics or shifts in research direction.

11.3 Cluster Detection

Automatically group papers into thematic clusters (e.g., "Transformer Models", "RCT Methods").

11.4 Personal Brain

• Reading Stats: Time spent, pages read, papers completed.
• Suggestions: Papers recommended based on your reading history.
• Tag Cloud: Visual representation of your most used tags.

12. Workspace & Projects

12.1 Workspaces & Members

Create separate workspaces for different research areas. Invite team members (Pro/Lab) and manage permissions.

12.2 Projects

Organise papers, reviews, evidence matrices, and chats into projects within a workspace.

12.3 Screening Board & PRISMA

• Systematic review screening with Include / Exclude / Maybe statuses.
• Auto‑generate a PRISMA flow diagram based on your screening decisions.

12.4 Annotations & Highlights

Highlight text and add margin notes on PDFs. All annotations are searchable and linked to the paper.

12.5 Reading Progress

Track percentage of pages read for each paper.

12.6 Backup & Restore

• Backup: Export your entire database (papers, chunks, vectors, settings) to a ZIP file.
• Restore: Import a previous backup.
• Privacy Data Export: Export all personal data in a machine‑readable format (JSON).

13. AI Providers Configuration

13.1 Cloud Free Mode

Default mode, no setup required. Limited daily quota (fair use).

13.2 Gemini, DeepSeek, Claude, Groq

Enter your API key in Settings → AI Providers. Test the connection with the "Validate" button.

13.3 Local AI (llama-server, GGUF)

Download a GGUF model (e.g., Llama 3 7B) and run llama-server. Point ResearchMind to the local endpoint (default http://localhost:8080).

13.4 Provider Fallback Chain

Define an ordered list of providers. If one fails, the next in chain is used automatically.

13.5 API Key Validation

Test your API key before saving to ensure it's valid.

13.6 Test Embedding Connection

Verify that the embedding model (bge‑m3) is reachable and working.

14. Settings & System

14.1 General Settings

• Interface language
• Default AI provider
• Theme (Light/Dark) — currently Light

14.2 Cache Management

• View cache size (embeddings, search results, LLM responses).
• Clear cache to free up disk space.

14.3 System Diagnostics

• Health metrics: backend status, database integrity.
• Reliability score: success rate of AI calls.
• Detect system specs: RAM, GPU, CPU.

14.4 Rebuild Index

• Rebuild FTS Index: Re‑create the full‑text search index.
• Rebuild Vector Index: Re‑embed all papers (time‑consuming).

15. Licensing & Account

15.1 Free vs Pro vs Lab

• Free: 50 papers, basic search, 14‑day Pro trial.
• Pro ($4/mo): Unlimited papers, all AI features, priority support.
• Lab: Team features, custom deployment, dedicated support.

Compare full details and features on our Pricing Page.

15.2 License Activation

Enter your license key (sent via email after purchase) to activate Pro or Lab features. Learn more about subscription terms in our Terms of Service.

15.3 Google Sign‑in

Optional: Sign in with your Google account to sync settings across devices (no library data is synced).

16. FAQ

Privacy & Data

Offline Usage

Troubleshooting

Support