DocuSense Help & Documentation

Quick Start

DocuSense checks your documents for grammar, spelling, style, custom guidelines, citation format, and much more — using up to three checking engines simultaneously.

1. Upload your document — drag and drop a .docx, .pptx, or .txt file, or paste text directly into the paste area.
2. Configure your check options — set document type, English variant, tone, and citation style in the left sidebar.
3. (Optional) Configure an AI mode — click the AI badge in the header to set up Ollama, OpenAI, Gemini, or Claude for semantic and style checks.
4. Click Run Check — results appear in the Comments panel. Each issue shows the text, a suggestion, the category, and its severity.
5. Accept or dismiss issues — hover over a comment card to accept the suggestion, dismiss, or add the word to your dictionary.

Uploading a Document

Supported Formats

• .docx / .dotx — Microsoft Word documents and templates
• .pptx — PowerPoint presentations (text is extracted from all slides)
• .txt — Plain text files

How to Upload

• Drag & drop — drag a file anywhere onto the document panel
• Browse — click the "Upload Document" button and select a file
• Paste text — click "Paste Text" to open a text area and paste directly

Running a Check

Click the Run Check button in the header. Results appear in tabs on the right panel: Grammar, Style, Spelling, Punctuation, Guidelines, Citations, Theory, and Template.

Check Options (left sidebar)

Script Engine

The script engine runs entirely in your browser with no internet required. It checks:

The script engine respects your Dictionary (words it won't flag) and Exception rules (wildcards like data*).

LanguageTool

LanguageTool is the default grammar checking engine. It adds over 1500 language rules covering grammar, style, typography, and regional variants.

Online (Default)

When LanguageTool is enabled in Setup, DocuSense connects to the LanguageTool public API. No login required. Limit: ~20,000 characters per check.

Local Server (Recommended for large documents)

1. Download the LanguageTool Desktop App from languagetool.org/download
2. Run the app — it starts a local server on port 8081
3. In DocuSense → Setup → Grammar Mode, switch to LanguageTool Local
4. Click Test Connection — the status turns green

What LanguageTool Adds (over the Script Engine)

• Context-aware grammar (e.g. "a apple" vs "an apple" in long sentences)
• Style suggestions for redundant phrases, colloquialisms
• Regional spelling variants (colour/color, organise/organize)
• Advanced punctuation rules

AI Modes

AI modes add a third layer of checking beyond the Script Engine and LanguageTool. They detect semantic issues, tone inconsistency, style deviations, and check custom guidelines that can't be expressed as simple rules.

What AI Adds

• Tone & voice consistency — detects tone shifts across paragraphs (formal → casual)
• Document-wide style profile — samples the full document, then flags paragraphs that deviate from the established style
• Semantic grammar — catches awkward phrasing, ambiguous pronoun references, unclear argumentation
• Custom guideline compliance — checks rules that require understanding of context (e.g. "Use active voice in instructions")
• Theory & framework validation — checks documents against Bloom's, Kirkpatrick, ADDIE, Gagne's, Kolb, Andragogy

AI Does Not Run in Live Check

Due to cost and latency, AI engines only run when you click Run Check or Check Selection. Live check uses Script + LanguageTool only.

Combined Mode

Combined Mode runs multiple AI engines together. Configure it in AI Mode Settings → Combined.

Strategies

Setting Up Combined Mode

1. Click the AI badge in the header → AI Mode Settings
2. Select the Combined mode card
3. Under "Combined Mode Configuration", choose Merge or Fallback Chain
4. Check the engines you want to include (only engines with valid credentials show as configured)
5. Click Save Settings

Custom Rules & Guidelines

You can upload or paste your organisation's style guide, branding guide, or any writing guidelines. DocuSense extracts rules automatically and uses them in all checking modes.

Adding Rules

• Upload a document — .docx, .txt, or .pdf guidelines file. Rules are auto-extracted (grammar, style, terminology, structure categories)
• Paste text — paste guidelines directly into the Guidelines panel
• Manual entry — use the Rules Manager to add individual rules with categories and severity levels

Rule Categories

Rules are classified as: Grammar, Spelling, Punctuation, Style, Formatting, Branding, Terminology, Structure, or Custom. Rules needing AI (⚡) are checked semantically when an AI mode is configured.

Exporting & Re-using Rules

Use Rules Manager → Export to save your rule set as a JSON file for re-use across documents or team members.

Citation Styles

Select a citation style in Check Options. The Script Engine checks format patterns; AI mode checks semantic compliance.

Dictionary & Exceptions

User Dictionary

Words in your dictionary are never flagged by any engine (Script, LanguageTool, or AI). Common uses: organisation names, product names, technical acronyms.

• Add a word: type in the "Add word" box in the Dictionary & Exceptions panel and click Add, or right-click a flagged word → "Add to Dictionary"
• Remove a word: click the × next to the word in the Dictionary panel
• Import: import a .txt file with one word per line
• Export: download your dictionary as a .txt file to share with colleagues

Exception Rules

Exception rules let you suppress issues matching a pattern. These are typically set by uploading a guidelines document that includes lines like:

• Exception: data — exact word match (will NOT suppress "metadata")
• Exception: data* — wildcard prefix — suppresses "data", "database", "datacenter"

Live Check

Live Check highlights issues as you type, without clicking Run Check.

How to Enable

Click the Live button in the header bar. The button turns blue when active. Click again to disable. Preference is saved across sessions.

Engine Behaviour in Live Mode

Check Selection

Check just the text you've selected rather than the entire document.

1. Click and drag to select text in the document viewer (on mobile: long-press to start selection)
2. A Selection button appears in the header bar
3. Click Selection — only the selected paragraphs are checked
4. Results replace existing issues for those paragraphs; issues in non-selected paragraphs are preserved

Template Check

Compare your document against a template to find formatting and structural deviations.

Running a Template Check

1. Switch to the Template tab at the top of the document area
2. Upload your template file (.docx or .dotx)
3. Upload the document to compare
4. Click Run Template Check

What It Checks

• Font family, size, and weight per heading/paragraph style
• Line spacing and paragraph spacing
• Heading hierarchy and order
• Required sections present
• Colour usage (text colour, highlight colour)

AI Enhancement

When an AI mode is configured, each template issue gets an impact explanation — why the deviation matters — and a specific fix suggestion. AI also checks section order and content expectations.

Theory Checks

Theory checks validate your document against educational or instructional design frameworks using AI. DocuSense supports:

How to Run a Theory Check

1. Load your document first
2. Configure an AI mode (Theory check requires AI)
3. In the left sidebar, find the Theory Checks section
4. Upload a theory file or select the framework from the list
5. Click AI Check

What's Active Panel

The What's Active panel in the left sidebar shows you exactly which engines are running and what they're checking:

• ● Green dot — engine is active and connected
• ● Amber dot — engine available but not fully configured
• ● Grey dot — engine disabled or not connected

Click Configure next to any engine to go directly to its settings.

Browser AI (Transformers.js)

Browser AI runs a small language model directly in your browser using Transformers.js. No API key required. No data leaves your device.

Setup

1. In AI Mode Settings, select Browser AI
2. Choose a model from the list (flan-t5-base is recommended for speed)
3. Click Download Model — the model downloads once and is cached locally
4. Once downloaded, run checks offline indefinitely

Limitations

• Smaller model = less accurate than GPT-4o or Claude
• First download: 300MB – 1GB depending on model
• Not recommended on mobile (memory constraints)
• Initial download requires internet; subsequent use is fully offline

Ollama (Local AI)

Ollama runs open-source AI models locally on your computer. Excellent accuracy with full privacy.

Setup

1. Download and install Ollama from ollama.com
2. Open a terminal and run: ollama pull phi3:mini (or another model)
3. Ollama starts automatically at http://localhost:11434
4. Reload DocuSense — Ollama is auto-detected on startup
5. In AI Mode Settings, select Local AI (Ollama) and choose your model

Recommended Models

OpenAI (GPT)

Setup

1. Sign up at platform.openai.com and create an API key
2. In AI Mode Settings, select OpenAI GPT
3. Paste your API key (stored locally in your browser only — never sent to DocuSense servers)
4. Choose model: GPT-4o (recommended) for best accuracy, or GPT-4o-mini for lower cost

Google Gemini

Setup

1. Get an API key from Google AI Studio
2. In AI Mode Settings, select Gemini AI
3. Paste your Gemini API key
4. Choose model: Gemini 1.5 Pro (recommended) or Gemini 1.5 Flash (faster, lower cost)

Anthropic Claude

Setup

1. Get an API key from console.anthropic.com
2. In AI Mode Settings, select Claude AI
3. Paste your API key (starts with sk-ant-)
4. Choose model: Claude Sonnet 4 (recommended), Claude Opus 4 (best accuracy), or Haiku (fastest)

Mobile vs Desktop

DocuSense is fully usable on mobile devices. On screens under 768px, a bottom tab bar replaces the desktop panel layout.

Mobile Navigation

Use the bottom tab bar:

• Guidelines — access the left panel (rules, check options, dictionary)
• Document — view and edit your document
• Run Check (FAB) — the centre floating button runs the check
• Comments — view check results
• Settings — access check options modal

Keyboard Shortcuts

Troubleshooting

LanguageTool not connecting

• Check your internet connection (for online mode)
• For local mode: ensure the LanguageTool Desktop App is running
• Check the URL in Setup matches http://localhost:8081
• Try clicking Test Connection in Setup

Ollama not detected

• Open a terminal and run ollama serve to ensure it's running
• Check the URL in AI Settings matches http://localhost:11434
• Reload the page — Ollama is auto-detected on startup
• On some systems, CORS may block requests — try running Ollama with OLLAMA_ORIGINS="*" ollama serve

AI check returns no results

• Verify your API key is correct and has sufficient credit
• Check the browser console for error messages (F12 → Console)
• For OpenAI: ensure your key has access to the selected model
• For Gemini: check your API key is enabled for the Gemini API (not just AI Studio preview)
• Try a shorter document — some providers have token limits per request

Dictionary words still being flagged

• Words are matched with word-boundary detection — "data" in your dictionary will NOT suppress "metadata"
• Check the word is in the Dictionary panel (left sidebar)
• After adding a word, re-run the check — results from the previous run do not update automatically

Browser AI (Transformers.js) fails to download

• Ensure you have 500MB–1GB of free disk space
• Try a different model (flan-t5-base is the smallest)
• Disable any browser extensions that may block CDN downloads
• Use a desktop browser (Chrome or Edge recommended)

Template check shows no differences

• Ensure both files are in .docx format (not .doc or .odt)
• Template must have named paragraph styles — generic formatting won't be compared

Frequently Asked Questions

Is my document sent to any server?

Only when you use an AI mode (OpenAI, Gemini, or Claude). Text is sent to the respective AI provider. The Script Engine, LanguageTool Local, Ollama, and Browser AI all run entirely on your device. DocuSense itself never receives your document content.

Are API keys stored securely?

API keys are stored in your browser's localStorage — they never leave your browser and are never sent to DocuSense servers. They are sent directly to the AI provider's API from your browser.

Does DocuSense work offline?

Yes — the Script Engine works fully offline. LanguageTool works offline with the local server installed. Ollama and Browser AI also work offline. Only OpenAI, Gemini, Claude, and LanguageTool Online require internet.

How does DocuSense compare to Grammarly?

What file types are supported?

.docx, .dotx (Word), .pptx (PowerPoint slides as text), .txt (plain text).

How large a document can I check?

The Script Engine has no practical limit. LanguageTool Online is limited to ~20,000 characters — use the local server for longer documents. AI providers process the document in chunks and can handle any length, though very long documents take more time.

Can I share my settings with my team?

Yes — export your rule sets via Rules Manager → Export. Import them on a colleague's machine. Dictionary export is available in the Dictionary & Exceptions panel.