Righting Sentences User Guide - 4.0

Righting Sentences — User Guide (v4.0)

Last updated: February 2026

Production URL: https://righting-sentences-573737403792.us-east1.run.app/

1. Purpose and Scope

Righting Sentences is a web-based chat application that generates AI text that reads more like standard written English. This application is a data science project that includes a unique two-pass editorial system, the Morey Language Model which automatically detects and removes AI-typical language patterns. Usage improves the Morey Language Model, an editorial system that eventually will include a fine-tuned foundational LLM.

2. Access and Authentication

2.1 Google account required

To use the application, you must sign in with a Google account.

2.2 Invitation-only access

Access is provided by invitation. If you cannot sign in or do not see the application after signing in, contact the administrator (Saxton Publishing).

2.3 Sign in

1. Navigate to the production URL
2. Select Sign in with Google
3. Complete Google authentication
4. After sign-in, you will see the main chat interface

3. User Interface

3.1 Layout overview

The interface has two main areas:

• Left sidebar - Chat history, model selector, and menu
• Main chat window - Conversation view and message input

3.2 Left Sidebar

Chat history list

Each row shows:

• The conversation title
• The model used for that conversation

To open a conversation, select it from the list.

Model selector

Choose which AI model to use for new conversations:

• GPT-4o - OpenAI's capable general-purpose model
• GPT-5.2 Pro - OpenAI's advanced reasoning model
• Claude Sonnet 4.5 - Anthropic's balanced model
• Gemini-3 Flash - Google's fast, efficient model

Important: Models cannot be changed mid-conversation. Start a new chat to use a different model.

Menu items

• Account - View your profile and sign out
• Settings - Theme, colors, and custom instructions
• Insights - Editorial analytics dashboard
• Help - Link to documentation
• Sign Out - End your session

3.3 Main Chat Window

Header

Shows the conversation title and current model badge.

Messages

• Your messages appear on the right
• Assistant messages appear on the left
• Hover over assistant messages to reveal the copy button

Message input

Type your message and press Enter or click Send.

4. Sending Messages

4.1 Basic messaging

1. Type your message in the input field
2. Press Enter or click Send
3. Wait for the assistant to respond

4.2 Multi-line messages

Press Shift+Enter to create a new line without sending.

4.3 Stop generation

While the assistant is responding, click Stop to halt generation early.

5. Options Menu

The Options button inside the message input provides access to:

• File uploads
• Prompt templates
• Quick actions

5.1 Opening Options

Click the Options button (three dots icon) to reveal the dropdown menu.

6. File Uploads

6.1 Upload a file

1. Click Options
2. Select Upload files
3. Choose one or more files

6.2 Supported file types

• Text: .txt, .md, .csv, .tsv, .json
• Documents: .pdf, .doc, .docx
• Presentations: .ppt, .pptx
• Spreadsheets: .xls, .xlsx
• Images: .png, .jpg, .gif, .webp, .bmp, .tiff

6.3 File status

Uploaded files appear as chips above the input:

• Uploaded - File received
• Parsed - Text extracted and available to the model

6.4 Remove a file

Click the × on the file chip to remove it from the conversation.

6.5 File size limit

Maximum upload size is 25 MB per file.

7. Prompt Templates

Templates provide structured prompts for common content types.

7.1 Using a template

1. Click Options
2. Select a template from the list
3. The template badge appears above the input
4. Type your specific content or context
5. Send your message

7.2 Available templates

Landing Pages

Template	Description
Locations	Global offices and delivery centers page
Services	Service overview with capabilities, proof, and outcomes
Industries	Industry overview or specialization page
Leadership	Leadership team introduction and credibility
Clients	Client portfolio and proof of impact

Case Studies

Template	Description
Case Study (PPT)	PowerPoint presentation structure (7 slides)
Case Study (Website)	Web-optimized case study with metrics

7.3 Template behavior

• Templates automatically use Google Gemini for optimal results
• The template badge shows which template is active
• Click × on the badge to clear the template

8. Quick Actions

Quick actions transform the previous AI response with one click.

8.1 Available quick actions

Action	Effect
Make it shorter	Condense to ~50% of original length
Make it more formal	Increase professional tone
Simplify this	Make clearer and more accessible
Expand on this	Add detail and elaboration

8.2 Using quick actions

1. Ensure there is a previous assistant message in the conversation
2. Click Options
3. Select a quick action
4. The action executes immediately (no additional prompt needed)

8.3 Quick action behavior

• Quick actions automatically reference the previous assistant response
• They use Google Gemini regardless of conversation model
• The transformed response appears as a new message

9. Editorial Insights

The Insights dashboard shows your personal editorial analytics.

9.1 Accessing Insights

1. Open the sidebar
2. Click Insights in the menu

9.2 Metrics displayed

• Total Messages - Number of messages processed
• Passed - Messages without AI pattern violations
• Violations - Messages where patterns were detected and corrected
• Pass Rate - Percentage of clean messages

9.3 Top Violations

Shows the most common AI-typical words detected in your conversations, such as:

• "utilize" (instead of "use")
• "landscape" (overused in AI writing)
• "delve" (AI favorite)

9.4 Recent Corrections

Examples of recent messages where the editorial system improved the output.

10. Conversation Export

Export your conversations for documentation or backup.

10.1 Export formats

• Markdown - Formatted text file
• JSON - Structured data

10.2 How to export

1. Open the conversation you want to export
2. Click the Export button in the header
3. Select Markdown or JSON
4. The file downloads automatically

11. Settings

11.1 Theme

• Light - Light background
• Dark - Dark background
• System - Match your operating system preference

11.2 Primary color

Choose an accent color for buttons and highlights.

11.3 Custom instructions

Provide information about your preferences that the assistant should remember.

11.4 LLM instructions

Add rules for how the assistant should respond in all conversations.

12. The Two-Pass Editorial System

Righting Sentences uses a unique two-pass system to improve AI output.

12.1 How it works

1. Pass 1: The AI generates a natural response
2. Detection: The system identifies AI-typical patterns:
   • Banned words (e.g., "utilize", "landscape", "delve")
   • Three-part lists (triads)
   • Rhetorical patterns (e.g., "It's not X, it's Y")
3. Pass 2: The AI regenerates with stricter guidelines
4. Post-processing: Any remaining issues are cleaned programmatically

12.2 Why it matters

AI models tend to overuse certain words and patterns that make text sound artificial. The editorial system produces writing that sounds more natural and human-like.

12.3 Banned words

The system currently flags: utilize, landscape, unlock, critical, guardrails, tracks, shape, shift, teams, delve, tapestry.

12.4 Triad exceptions

Some three-part phrases are culturally significant and exempt from editing:

• "executive, legislative, and judicial"
• "life, liberty, and the pursuit of happiness"
• "red, white, and blue"

13. Troubleshooting

13.1 I cannot sign in

• Confirm you are using a Google account
• Confirm your account has been invited
• Contact Saxton Publishing if the problem persists

13.2 I cannot change the model

Models cannot be changed mid-conversation. Start a new chat to use a different model.

13.3 Quick actions are not available

Quick actions require at least one assistant message in the conversation. Send a message first, then use quick actions.

13.4 Template is not working

• Ensure the template badge is visible above the input
• Templates require you to provide context in your message
• If issues persist, clear the template and try again

13.5 File upload fails

• Check that the file is under 25 MB
• Verify the file type is supported
• Contact the administrator if errors persist

14. Data and Privacy

14.1 What is stored

• Conversations and messages
• Uploaded files
• Editorial logs (for system improvement)

14.2 Where data is stored

• Google Cloud Firestore (conversations, metadata)
• Google Cloud Storage (uploaded files)

14.3 Third-party providers

Messages are sent to the selected model provider (OpenAI, Anthropic, or Google) to generate responses.

---

Righting Sentences is developed by Saxton Publishing.