# Manual E2E Test Plan: Inline AI Editing
> **Version**: 1.0.0
> **Feature**: Inline AI Text Editing
> **Date**: 2026-02-05
## Overview
This document outlines manual end-to-end testing procedures for the Inline AI Editing feature. These tests verify the complete user workflow from text selection through AI-powered transformation to final application.
---
## Prerequisites
1. **Obsidian** is running with the Claudesidian plugin installed
2. **LLM provider** is configured (API key set, model selected)
3. A **test note** is open with sample text for editing
---
## Test Environment Setup
Create a test note with the following content:
```markdown
# Test Document
This is a sample paragraph that we will use for testing the inline AI editing feature. It contains some text that could be improved.
## Code Example
```javascript
function oldStyleFunction(x) {
return x * 2;
}
```
## List Section
- First item needs improvement
- Second item is okay
- Third item could be more concise
> This blockquote has some **formatting** that should be preserved.
Special characters: "quotes" & 'apostrophes' and unicode: 中文 émoji 🚀
```
---
## Test Cases
### TC-001: Happy Path - Basic Text Transformation
**Priority**: P0
**Objective**: Verify basic select → instruct → generate → apply flow
**Steps**:
1. Open the test note
2. Select the sentence: "This is a sample paragraph..."
3. Trigger inline edit via:
- **Option A**: Right-click → "Edit with AI"
- **Option B**: Use hotkey (Cmd/Ctrl + Shift + E if configured)
4. In the modal, verify:
- Selected text is displayed in the preview area
- Instruction input is focused
- Model dropdown shows available models
5. Enter instruction: "Make this more concise"
6. Click "Generate"
7. Observe:
- Loading spinner appears
- Streaming preview shows text being generated
8. When complete, verify:
- Original text is shown (collapsed)
- Edited text is shown in editable textarea
- Retry, Cancel, Apply buttons are visible
9. Click "Apply"
10. Verify:
- Modal closes
- Selected text in editor is replaced with AI-generated text
- Notice shows "Changes applied"
**Expected Result**: Text successfully transformed and replaced in editor
---
### TC-002: Cancel During Loading
**Priority**: P0
**Objective**: Verify cancellation stops generation and returns to input state
**Steps**:
1. Select any text
2. Open inline edit modal
3. Enter instruction: "Translate to French" (something that takes time)
4. Click "Generate"
5. **Immediately** click "Cancel" while loading spinner is visible
6. Verify:
- Generation stops
- Modal returns to INPUT state (instruction input visible)
- No error message shown
- Can enter new instruction and try again
**Expected Result**: Clean cancellation without errors
---
### TC-003: Retry from Result State
**Priority**: P1
**Objective**: Verify retry returns to input state and allows regeneration
**Steps**:
1. Complete a successful generation (TC-001 steps 1-8)
2. In RESULT state, click "Retry"
3. Verify:
- Modal returns to INPUT state
- Original selected text is preserved
- Previous instruction may be preserved or cleared (document actual behavior)
4. Enter a different instruction
5. Click "Generate"
6. Verify new result is generated
**Expected Result**: Retry allows iterating on the transformation
---
### TC-004: Error - Empty Instruction
**Priority**: P0
**Objective**: Verify validation prevents empty instruction submission
**Steps**:
1. Select text and open inline edit modal
2. Leave instruction field empty
3. Click "Generate"
4. Verify:
- Notice appears: "Please enter an instruction"
- Instruction input is focused
- Modal stays in INPUT state
- No API call is made
**Expected Result**: Clear validation feedback, no wasted API calls
---
### TC-005: Context Menu Trigger
**Priority**: P1
**Objective**: Verify context menu entry appears only with selection
**Steps**:
1. Open test note
2. Right-click without selecting any text
3. Verify "Edit with AI" option is **NOT** in the menu
4. Select some text
5. Right-click on the selection
6. Verify "Edit with AI" option **IS** in the menu
7. Click "Edit with AI"
8. Verify modal opens with selected text
**Expected Result**: Context menu respects selection state
---
### TC-006: File Switch During Modal Open (HIGH RISK)
**Priority**: P0 (Flagged by Coder)
**Objective**: Verify error handling when active file changes during modal session
**Steps**:
1. Open test note A
2. Select text and open inline edit modal
3. Complete generation (get to RESULT state)
4. **Without closing the modal**, click on a different note (Note B) in the file explorer
5. Click "Apply" in the modal
6. Verify:
- Error notice appears (e.g., "The editor has changed. Please try again in the active editor.")
- Text in Note B is **NOT** modified
- Text in Note A is **NOT** modified (since it's no longer active)
- Modal remains open OR closes gracefully
**Expected Result**: Safe failure - no unintended edits to wrong file
---
### TC-007: Rapid Generate/Cancel Sequences (HIGH RISK)
**Priority**: P0 (Flagged by Coder)
**Objective**: Verify stability with rapid user interactions
**Steps**:
1. Select text and open modal
2. Enter instruction
3. Perform rapid sequence:
- Click "Generate"
- Immediately click "Cancel"
- Immediately click "Generate" again
- Wait for result
4. Verify:
- No errors or crashes
- Final state is consistent (either INPUT or RESULT)
- Modal is responsive
**Variations to test**:
- Generate → Cancel → Generate → Cancel → Generate
- Generate → (wait 1 second) → Cancel → Generate
**Expected Result**: System handles rapid interactions gracefully
---
### TC-008: Model Dropdown Edge Case (MEDIUM RISK)
**Priority**: P1 (Flagged by Coder)
**Objective**: Verify behavior when no models are available
**Steps**:
1. Temporarily disable all LLM providers in settings (remove API keys)
2. Restart Obsidian or reload plugin
3. Select text and attempt to open inline edit modal
4. Verify:
- Modal opens (or shows helpful error)
- If modal opens, dropdown shows meaningful state (empty or "No models available")
- Generate button is disabled OR clicking it shows helpful error
**Expected Result**: Graceful handling of no-models scenario
---
### TC-009: Markdown Preservation
**Priority**: P1
**Objective**: Verify markdown formatting is preserved through transformation
**Steps**:
1. Select the blockquote text: `> This blockquote has some **formatting**...`
2. Open inline edit modal
3. Enter instruction: "Make this more enthusiastic"
4. Generate and Apply
5. Verify:
- Blockquote syntax (`>`) is preserved
- Bold formatting (`**...**`) is preserved
- Result renders correctly in preview mode
**Expected Result**: Markdown formatting survives round-trip
---
### TC-010: Special Characters Preservation
**Priority**: P1
**Objective**: Verify special characters are not corrupted
**Steps**:
1. Select the special characters line: `Special characters: "quotes" & 'apostrophes'...`
2. Open inline edit modal
3. Enter instruction: "Keep this exactly the same but add a period at the end"
4. Generate and Apply
5. Verify:
- Quotes, ampersand, apostrophes are preserved
- Unicode Chinese characters (中文) are preserved
- Emoji (🚀) is preserved
**Expected Result**: No character corruption or encoding issues
---
### TC-011: Undo After Apply (Ctrl/Cmd + Z)
**Priority**: P1
**Objective**: Verify editor undo works after applying AI changes
**Steps**:
1. Note the exact original text
2. Complete a full transformation (TC-001)
3. With cursor in the editor, press Ctrl+Z (Windows/Linux) or Cmd+Z (Mac)
4. Verify:
- Original text is restored
- No additional undo steps required
**Expected Result**: Single undo restores original text
**Implementation Note**: The `editor.replaceRange()` Obsidian API is used to apply changes. This API automatically integrates with Obsidian's native undo stack. Undo functionality is expected to work without additional implementation because `replaceRange` is the standard Obsidian editor API that preserves undo history.
**Verification Criteria**:
- After clicking "Apply" and confirming the change is visible in the editor
- Pressing Ctrl/Cmd+Z once should restore the exact original text
- The undo operation should be atomic (single step, not character-by-character)
---
### TC-012: Code Block Transformation
**Priority**: P1
**Objective**: Verify code blocks can be transformed
**Steps**:
1. Select the code block content (inside the fences)
2. Open inline edit modal
3. Enter instruction: "Convert to arrow function syntax"
4. Generate and Apply
5. Verify:
- Code is transformed appropriately
- Syntax highlighting still works
- No extra escaping introduced
**Expected Result**: Code transformation works correctly
---
### TC-013: Long Text Selection
**Priority**: P2
**Objective**: Verify handling of large text selections
**Steps**:
1. Create or find a note with a very long section (1000+ words)
2. Select the entire long section
3. Open inline edit modal
4. Verify:
- Selected text preview is truncated appropriately (with "...")
- Full text is sent to LLM
5. Enter instruction: "Summarize this"
6. Generate and Apply
7. Verify transformation completes without timeout
**Expected Result**: Long text handled within reasonable time
---
### TC-014: Mobile Testing (Hotkey Only)
**Priority**: P2
**Objective**: Verify hotkey access works on mobile (no context menu)
**Steps** (on mobile device or tablet):
1. Open Obsidian mobile app
2. Open a note and select text
3. Use command palette to find "Edit selection with AI"
4. Trigger the command
5. Verify modal opens and workflow functions
**Note**: Context menu is not available on mobile - this is expected behavior per design decision.
**Expected Result**: Feature accessible via command palette on mobile
---
## Test Results Template
| Test ID | Date | Tester | Result | Notes |
|---------|------|--------|--------|-------|
| TC-001 | | | PASS/FAIL | |
| TC-002 | | | PASS/FAIL | |
| TC-003 | | | PASS/FAIL | |
| TC-004 | | | PASS/FAIL | |
| TC-005 | | | PASS/FAIL | |
| TC-006 | | | PASS/FAIL | |
| TC-007 | | | PASS/FAIL | |
| TC-008 | | | PASS/FAIL | |
| TC-009 | | | PASS/FAIL | |
| TC-010 | | | PASS/FAIL | |
| TC-011 | | | PASS/FAIL | |
| TC-012 | | | PASS/FAIL | |
| TC-013 | | | PASS/FAIL | |
| TC-014 | | | PASS/FAIL | |
---
## Known Issues / Limitations
Document any discovered issues here during testing:
1. _[Issue description]_
2. _[Issue description]_
---
## Revision History
| Version | Date | Author | Changes |
|---------|------|--------|---------|
| 1.0.0 | 2026-02-05 | PACT Test Engineer | Initial test plan |
