PPTX Editor Bot

\n \n \n

Upload PPTX File

\n \n

Click to upload or drag and drop your PPTX file here

Only .pptx files are supported

\n\n \n

Chat with PPTX Bot

\n \n

\n \n \n

\n Bot: Hello! Upload a PPTX file and I'll help you edit it. I can add text or replace existing content using natural language commands like \"Add lorem ipsum\" or \"Replace 'Hello' with 'Hi'\".\n

\n\n \n

\n \n \n

\n\n \n\n",[],[240,247],[240],{"_102":258,"_95":259,"_153":260,"_155":-5,"_156":136,"_157":261,"_159":160,"_104":262,"_99":263,"_162":163,"_164":-5,"_165":22,"_166":-5,"_167":264,"_98":326},"63dd925e-c1f6-11f0-a945-42dde27851f2","pptx_examples","5da65bb4-c1f6-11f0-bac6-42dde27851f2",170,["D",1763192442978],2,[265,273],{"_102":266,"_95":267,"_153":268,"_155":260,"_156":136,"_157":269,"_159":160,"_104":270,"_99":158,"_162":163,"_164":-5,"_165":22,"_166":-5,"_167":271,"_98":272},"8b8c5326-c1f6-11f0-b14a-42dde27851f2","Introduction_to_Swayable_pptx_unzipped","8148b59e-c1f6-11f0-87db-42dde27851f2",174,["D",1763192509574],[],[259,267],{"_102":274,"_95":275,"_153":276,"_155":260,"_156":136,"_157":277,"_159":160,"_104":278,"_99":158,"_162":163,"_164":-5,"_165":22,"_166":-5,"_167":279,"_98":325},"7e84c690-c1f6-11f0-b8b3-42dde27851f2","Mini_template_01_pptx_unzipped","70ae81a0-c1f6-11f0-844b-42dde27851f2",172,["D",1763192487715],[280],{"_102":281,"_95":282,"_153":283,"_155":276,"_156":136,"_157":284,"_159":160,"_104":285,"_99":158,"_162":163,"_164":-5,"_165":22,"_166":-5,"_167":286,"_98":324},"37d7dd8a-c1f7-11f0-92b8-42dde27851f2","slides","35b6d92a-c1f7-11f0-a2f8-42dde27851f2",182,["D",1763192798639],[287,296,306,315],{"_102":288,"_95":289,"_153":290,"_155":283,"_156":136,"_157":291,"_159":160,"_104":292,"_99":242,"_162":252,"_164":293,"_165":22,"_166":-5,"_167":294,"_98":295},"6a2e700a-c1f7-11f0-a8c5-42dde27851f2","side2.xml","59307618-c1f7-11f0-91d1-42dde27851f2",193,["D",1763192883080],"\nExecutive SummaryClient X tested two ads for Brand Y on Swayable in 48 hours from August 13-14, analyzing over 3,300 responses from verified Brazilian consumers each exposed to one creative assetUso Combinado advertised Power Plus alongside Brand Y’s hero product, whereas Video 2 focused on Fresh Power 6Both ads featured a couple challenging gender roles as the man learns to clean the bathroom using Brand X productsProduct Y had the greatest lift in purchasing intent and brand lift overall as well as among Young, Established Families and Empty NestersHowever, Video A and Video B were equally effective among Young Singles, signaling that Video B may be more effective with the audience with lowest exposure to cleaning supply purchasesOverall, women reacted positively to seeing a man cleaning the house and had an overwhelmingly positive emotional response towards the concept",[],[259,275,282,289],{"_102":297,"_95":298,"_153":299,"_155":283,"_156":136,"_157":300,"_159":160,"_104":301,"_99":302,"_162":252,"_164":303,"_165":22,"_166":-5,"_167":304,"_98":305},"4fd4cea2-c1f7-11f0-80c1-42dde27851f2","slide1.xml","3a387120-c1f7-11f0-940a-42dde27851f2",188,["D",1763192838883],5,"\nFront cover templateTest descriptionDate, clientSwayable",[],[259,275,282,298],{"_102":307,"_95":308,"_153":309,"_155":283,"_156":136,"_157":310,"_159":160,"_104":311,"_99":302,"_162":252,"_164":312,"_165":22,"_166":-5,"_167":313,"_98":314},"81e8d42e-c1f7-11f0-b83d-42dde27851f2","slide3.xml","6be4ff9a-c1f7-11f0-8c2e-42dde27851f2",199,["D",1763192922901],"\nSlide headingParagraph with no bullet 1Paragraph with no bullet 2Bullet 1Bullet 2",[],[259,275,282,308],{"_102":316,"_95":317,"_153":318,"_155":283,"_156":136,"_157":319,"_159":160,"_104":320,"_99":242,"_162":252,"_164":321,"_165":22,"_166":-5,"_167":322,"_98":323},"a7386cd0-c1f7-11f0-8a02-42dde27851f2","slide4.xml","84f73188-c1f7-11f0-940a-42dde27851f2",204,["D",1763192985498],"\nInterstitial slide title",[],[259,275,282,317],[259,275,282],[259,275],[259],{"_102":328,"_95":329,"_153":330,"_155":-5,"_156":136,"_157":331,"_159":160,"_104":332,"_99":333,"_162":252,"_164":334,"_165":22,"_166":-5,"_167":335,"_98":336},"943cf2f2-c34d-11f0-9a88-42dde27851f2","README-content-editing.md","a9cea4fc-c349-11f0-9836-42dde27851f2",333,["D",1763339841532],3,"# Content Editing System\n\n## How It Works\n\nWhen a user says \"change the executive summary,\" the system:\n\n1. **AI identifies the target**: Determines if user wants entire text box or single element\n2. **AI generates markdown**: Creates clean bullet points like `• First point\\n• Second point`\n3. **System parses markdown**: Identifies bullets, sub-bullets, plain text\n4. **System generates XML**: Converts to proper PowerPoint structure with bullet formatting\n5. **System replaces content**: Swaps entire text box content while preserving structure\n\n## AI Decision Logic\n\n```\nUser: \"Change the executive summary\"\n→ AI: Text box replacement (entire content section)\n→ Response: textBoxId: \"TB1\", slideNumber: 2\n\nUser: \"Change the executive summary title\" \n→ AI: Single element replacement\n→ Response: elementId: \"2A\"\n\nUser: \"Use this template to make a deck about tortoises\"\n→ AI: Template detected (placeholder content like \"Front cover template\")\n→ Response: action: \"multi-slide\" (replace all template content)\n```\n\n## Template Recognition\n\nThe AI automatically detects template files by looking for placeholder content:\n- \"Front cover template\"\n- \"Test description\" \n- \"Executive Summary\" with generic content\n- Other template-like placeholder text\n\nWhen templates are detected, the AI defaults to **multi-slide operations** that replace all placeholder content with coordinated, themed content. This matches user intent - templates are meant to be completely transformed, not just have content added.\n\n## Markdown Format\n\nAI generates content in simple markdown:\n```\n• Main bullet point\n• Another main point\n • Sub-bullet (indented)\n• Final main point\n```\n\n## XML Generation\n\nSystem converts markdown to PowerPoint XML by generating only the inner content:\n```xml\n\n\n \n\n\n\n \n \n \n Main bullet point\n\n```\n\nThe system then replaces the content between existing `` tags to avoid XML structure corruption.\n\n## Text Box vs Element Targeting\n\n**Text Box Replacement** (entire content):\n- User says: \"change executive summary\", \"replace all bullets\"\n- AI returns: `textBoxId` + `slideNumber`\n- System replaces: Entire `` content\n\n**Element Replacement** (single item):\n- User says: \"change title\", \"replace first bullet\" \n- AI returns: `elementId`\n- System replaces: Single `` tag\n\n## Current Limitations\n\n- Only handles body text boxes (bullets) for full replacement\n- Sub-bullet indentation is simplified\n- No support for complex formatting (bold, italic, etc.)\n- Text box identification relies on placeholder type matching\n\n## Troubleshooting\n\n**Content duplication**: Check if AI returned `elementIds` array instead of `textBoxId`\n**Missing bullets**: Verify markdown parsing found bullet markers (`•`, `-`, `*`)\n**Wrong text box**: Check if text box regex pattern matches the slide structure\n**XML corruption**: Ensure generated content doesn't include `` wrapper tags\n**PowerPoint repair needed**: Usually indicates nested XML tags or missing required attributes",[],[329],{"_102":338,"_95":339,"_153":340,"_155":-5,"_156":136,"_157":341,"_159":160,"_104":342,"_99":32,"_162":252,"_164":343,"_165":22,"_166":-5,"_167":344,"_98":345},"7239c0a0-c34c-11f0-b7c8-42dde27851f2","README-development.md","7239c172-c34c-11f0-b7c8-42dde27851f2",327,["D",1763339354975],"# Development Roadmap\n\n## Next Major Feature: Slide Structural Operations\n\n### Current Limitation\nThe system can only edit content within existing slides. It cannot:\n- Duplicate slides\n- Delete slides \n- Reorder slides\n- Create new slides from templates\n\n### Implementation Plan: Slide Duplication & Deletion\n\n#### Phase 1: PPTX Structure Understanding\n**Goal**: Map out all the files and relationships that need updating\n\n**Key Files in PPTX**:\n```\n├── [Content_Types].xml # Declares all file types\n├── _rels/.rels # Root relationships\n├── ppt/\n│ ├── presentation.xml # Slide order and references\n│ ├── _rels/presentation.xml.rels # Slide file relationships\n│ └── slides/\n│ ├── slide1.xml # Individual slide content\n│ ├── slide2.xml\n│ └── slide3.xml\n```\n\n**Research Tasks**:\n1. Document exact XML structure for slide references in `presentation.xml`\n2. Map relationship IDs in `presentation.xml.rels`\n3. Understand slide numbering and ID assignment patterns\n4. Test what happens when files are missing/extra\n\n#### Phase 2: Safe Slide Duplication\n**Goal**: Copy existing slide with all proper relationships\n\n**Implementation Strategy**:\n```typescript\ninterface SlideOperation {\n type: 'duplicate' | 'delete' | 'reorder';\n sourceSlide?: number; // For duplication\n targetSlide?: number; // For deletion \n insertAfter?: number; // Where to place duplicated slide\n modifications?: { // Content changes to make after duplication\n textBoxId: string;\n content: string;\n };\n}\n```\n\n**Steps for Slide Duplication**:\n1. **Copy slide XML file**: `slide2.xml` → `slide4.xml`\n2. **Update presentation.xml**: Add new `` entry with unique ID\n3. **Update presentation.xml.rels**: Add relationship to new slide file\n4. **Update [Content_Types].xml**: Add entry for new slide file\n5. **Renumber subsequent slides**: If inserting in middle, update file names\n6. **Apply content modifications**: Use existing text box replacement logic\n\n**Error Handling**:\n- Validate all relationship IDs are unique\n- Ensure slide numbering is sequential\n- Verify all referenced files exist\n- Test with PowerPoint to ensure no corruption\n\n#### Phase 3: Safe Slide Deletion\n**Goal**: Remove slide and clean up all references\n\n**Steps for Slide Deletion**:\n1. **Remove slide XML file**: Delete `slide3.xml`\n2. **Update presentation.xml**: Remove corresponding `` entry\n3. **Update presentation.xml.rels**: Remove relationship entry\n4. **Update [Content_Types].xml**: Remove slide file entry\n5. **Renumber remaining slides**: `slide4.xml` → `slide3.xml`, etc.\n6. **Update all relationship references**: Fix any broken IDs\n\n#### Phase 4: Enhanced AI Integration\n**Goal**: AI can plan and execute structural changes\n\n**Enhanced Command Structure**:\n```json\n{\n \"action\": \"structural-multi-slide\",\n \"operations\": [\n {\n \"type\": \"duplicate\",\n \"sourceSlide\": 2,\n \"insertAfter\": 2,\n \"modifications\": {\n \"textBoxId\": \"TB0\",\n \"content\": \"Advanced Tortoise Care\"\n }\n },\n {\n \"type\": \"edit\", \n \"slideNumber\": 3,\n \"textBoxId\": \"TB1\",\n \"content\": \"• Advanced feeding tips\\n• Habitat optimization\"\n },\n {\n \"type\": \"delete\",\n \"slideNumber\": 5\n }\n ]\n}\n```\n\n**AI Prompt Enhancement**:\n- Recognize when user needs more slides than available\n- Suggest slide duplication with content variations\n- Plan slide deletion when content is redundant\n- Coordinate structural changes with content changes\n\n### Risk Mitigation\n\n#### High-Risk Areas\n1. **Relationship ID conflicts**: Duplicate IDs break PowerPoint\n2. **File numbering gaps**: Missing slide3.xml with slide4.xml present\n3. **Orphaned references**: presentation.xml referencing deleted slides\n4. **XML namespace issues**: Malformed relationship XML\n\n#### Testing Strategy\n1. **Unit tests**: Test each XML modification function\n2. **Integration tests**: Full PPTX generation and PowerPoint validation\n3. **Edge cases**: Empty slides, complex layouts, embedded media\n4. **Rollback capability**: Ability to revert to original file on failure\n\n#### Implementation Phases\n- **Phase 1**: Research and documentation (1-2 sessions)\n- **Phase 2**: Slide duplication only (2-3 sessions) \n- **Phase 3**: Slide deletion (1-2 sessions)\n- **Phase 4**: AI integration (1 session)\n\n### Success Criteria\n- [ ] Can duplicate any slide without corruption\n- [ ] Can delete any slide without breaking references\n- [ ] PowerPoint opens files without repair prompts\n- [ ] AI can plan multi-slide operations with structural changes\n- [ ] Comprehensive error handling and rollback capability\n\n### Notes\n- Start with duplication (safer than deletion)\n- Test extensively with Mini_template_01.pptx\n- Keep detailed logs of all XML modifications\n- Consider creating backup before structural operations",[],[339],{"_102":347,"_95":348,"_153":349,"_155":-5,"_156":136,"_157":350,"_159":160,"_104":351,"_99":333,"_162":252,"_164":352,"_165":22,"_166":-5,"_167":353,"_98":354},"3565625a-c31b-11f0-a71a-42dde27851f2","README-how_pptx-works.md","14158e04-c31b-11f0-8d02-42dde27851f2",270,["D",1763318207577],"# PPTX Open XML Cheat Sheet for Editing Agents\n\nThis document explains how a `.pptx` file works internally so you can parse and edit PowerPoint presentations via XML: change text, swap images, tweak layouts, and safely add or remove slides.\n\n---\n\n## 1. What a PPTX File Actually Is\n\nA `.pptx` file is a ZIP archive following the Open Packaging Convention (OPC) and the Office Open XML (OOXML) PresentationML standard.\n\nInside the ZIP you’ll find:\n\n- XML “parts” (content)\n- Binary assets (images, media, embedded files)\n- Relationship files (`*.rels`)\n- A content-type manifest (`[Content_Types].xml`)\n\nYou must:\n\n1. Treat the PPTX as a ZIP. \n2. Use the relationships to navigate, not just filenames. \n3. Preserve content types and relationship integrity when editing.\n\n---\n\n## 2. High-Level Directory Layout\n\nTypical PPTX structure (paths are inside the ZIP):\n\n- `/_rels/.rels` \n - Root relationships (e.g., to `ppt/presentation.xml`).\n- `/[Content_Types].xml` \n - Declares MIME types for each part type.\n- `/ppt/presentation.xml` \n - Main presentation part (list of slide references, masters, etc.).\n- `/ppt/_rels/presentation.xml.rels` \n - Relationships from the presentation to slide parts, slide masters, theme, etc.\n- `/ppt/slides/slide1.xml`, `slide2.xml`, ... \n - Individual slide parts.\n- `/ppt/slides/_rels/slide1.xml.rels`, ... \n - Relationships from slides to layouts, images, charts, etc.\n- `/ppt/slideMasters/slideMaster1.xml`, ... \n - Slide master parts.\n- `/ppt/slideLayouts/slideLayout1.xml`, ... \n - Slide layout parts.\n- `/ppt/notesSlides/notesSlide1.xml`, ... \n - Notes for slides.\n- `/ppt/theme/theme1.xml`, ... \n - Theme (colors, fonts).\n- `/ppt/media/image1.png`, `image2.jpeg`, ... \n - Images and other media.\n\nOther optional parts: charts, embedded objects, custom XML, etc.\n\n---\n\n## 3. Core OPC Concepts: Parts & Relationships\n\n### 3.1 Parts\n\nA part is a file inside the package, such as:\n\n- `ppt/slides/slide3.xml` (XML) \n- `ppt/media/image5.png` (PNG)\n\nEach part has:\n\n- A path inside the ZIP \n- A content type (from `[Content_Types].xml`) \n- Zero or more relationships to other parts or external URIs \n\n### 3.2 Relationships (`*.rels`)\n\nRelationships are stored in `.rels` XML files next to their “source” part.\n\nEach `` has:\n\n- `Id` – local identifier like `rId1` \n- `Type` – URI describing the relationship type \n- `Target` – relative path to target part \n- `TargetMode` – `Internal` (default) or `External`\n\nImportant:\n\n- `rId` values are only unique within their source part.\n- You must resolve targets via the appropriate `.rels` file, not by guessing filenames.\n\n---\n\n## 4. Presentation Structure: How Slides Are Wired\n\nThe entry point is `ppt/presentation.xml`.\n\n### 4.1 Main Presentation XML\n\n`ppt/presentation.xml` root is typically ``.\n\nKey elements:\n\n- `` – ordered list of slide instances \n - Child `` elements:\n - `id`: unique numeric ID within this list\n - `r:id`: references a relationship in `presentation.xml.rels`\n- `` – slide masters \n- `` – notes master \n- `` – handout master \n- `` – slide size \n- `` – default text styles \n\nSlide order is determined by the order of `` elements in ``, not by the numeric suffix in `slideN.xml`.\n\n### 4.2 How to Find All Slides\n\nAlgorithm:\n\n1. Open `ppt/presentation.xml`. \n2. Read all `` elements in `` in order. \n3. For each ``, get its `r:id`. \n4. In `ppt/_rels/presentation.xml.rels`, find the `` with `Id=\"\"`. \n5. Its `Target` is a slide part path, e.g. `slides/slide3.xml`.\n\nUse this relationship-based mapping instead of assuming `slideN.xml` is slide number N.\n\n---\n\n## 5. Slide XML (PresentationML Basics)\n\nA slide part (e.g. `ppt/slides/slide1.xml`) typically looks like:\n\n~~~xml\n\n \n \n \n \n \n ...\n\n~~~\n\nInside `` you’ll see:\n\n- `` – shapes (rectangles, text boxes, titles, etc.) \n- `` – images \n- `` – group shapes \n- `` – connectors \n- `` – charts, tables, SmartArt, etc.\n\nA typical `` (shape) has:\n\n- `` – non-visual properties (ID, name, placeholder info) \n- `` – shape properties (geometry, line, fill, transform) \n- `` – text content (if the shape holds text)\n\n---\n\n## 6. Text Model: Finding and Editing Text\n\nText is stored using DrawingML (`a:` namespace) inside ``.\n\n### 6.1 Text Body Structure\n\nTypical text-bearing shape:\n\n~~~xml\n\n ...\n ...\n \n \n \n \n ... \n \n \n Some text \n \n \n \n \n More text\n \n \n ... \n \n\n~~~\n\nElements:\n\n- `` – paragraph \n- `` – paragraph properties (bullets, level, alignment) \n- `` – run (a sequence of uniformly formatted text) \n- `` – run properties (size, font, color, bold, etc.) \n- `` – literal text \n- `` – manual line break \n- `` – field (date, slide number, etc.), also containing ``\n\n### 6.2 Practical Rules for Text Editing\n\n- To collect text from a slide:\n - Traverse `p:sld → p:cSld → p:spTree → p:sp` with a `p:txBody`.\n - For each `a:p`:\n - Concatenate all `a:t` text nodes in order.\n - Insert line breaks when encountering `a:br/`.\n- For search/replace:\n - Operate at the `a:t` level where possible.\n - Preserve the structure of `a:r` and `a:rPr` to keep formatting.\n - Handle text that may be split across multiple runs, even mid-word.\n- Placeholders (title, body, footer, etc.) are identified by `` in the shape’s non-visual properties, not by the text itself.\n\n---\n\n## 7. Layouts, Masters, and Placeholders\n\n### 7.1 Slide Masters and Layouts\n\nHierarchy:\n\n- Presentation → Slide Masters (`/ppt/slideMasters/slideMasterN.xml`) \n- Each Slide Master → Slide Layouts (`/ppt/slideLayouts/slideLayoutN.xml`) via relationships and `` \n- Each Slide → a Slide Layout via its slide `.rels` file\n\nSlide master/layout parts look similar to slides:\n\n- `` / `` root \n- `/` – shapes and placeholders \n- `` – default text styles \n- References to theme (`/ppt/theme/themeN.xml`)\n\nSlides inherit positioning and styles from their layout and master.\n\n### 7.2 Placeholders (``)\n\nPlaceholders designate semantic roles:\n\n~~~xml\n\n \n \n \n \n \n \n \n ...\n\n~~~\n\nCommon `type` values:\n\n- `title`, `ctrTitle` – title, centered title \n- `subTitle` – subtitle \n- `body` – main content \n- `pic` – picture placeholder \n- `dt` – date \n- `sldNum` – slide number \n- `ftr` – footer \n\nSlides override placeholder text by providing shapes with matching placeholder metadata (`type`, `idx`).\n\n### 7.3 Changing Layout of a Slide\n\nConceptually:\n\n1. In the slide’s `.rels` file, change the relationship of type `slideLayout` to target a different `slideLayoutX.xml`. \n2. Ensure the new layout’s placeholders are compatible (same logical roles), or be prepared to reposition or re-create shapes.\n\n---\n\n## 8. Images, Graphics, and Other Media\n\n### 8.1 Images (`` and ``)\n\nTypical picture shape:\n\n~~~xml\n\n ...\n \n \n ...\n \n ... \n\n~~~\n\n- `r:embed=\"rId5\"` refers to a relationship in `ppt/slides/_rels/slideX.xml.rels`. \n- That relationship’s `Target` is something like `../media/image3.png`. \n- The actual image is at `ppt/media/image3.png`.\n\nTo swap an image while preserving position and size:\n\n1. Find the `` with `r:embed=\"rIdX\"`. \n2. Resolve `rIdX` in the slide’s `.rels` file. \n3. Overwrite the binary file at the `Target` path with the new image bytes (preferably same format). \n4. Do not change XML unless pointing to a new media part.\n\n### 8.2 Shapes and Coordinates\n\nShapes use EMUs (English Metric Units):\n\n- 1 inch = 914400 EMUs \n- 1 cm ≈ 360000 EMUs \n\nPositions and sizes are defined in ``:\n\n~~~xml\n\n \n \n\n~~~\n\nAdjust these attributes to move/resize shapes.\n\n### 8.3 Charts, Tables, SmartArt\n\nThese reside in `` with `` inside. Charts often reference:\n\n- `/ppt/charts/chartN.xml` \n- Possibly embedded Excel parts under `/xl/`\n\nFor a generic editing agent:\n\n- Avoid restructuring chart/table XML unless you fully know the schema. \n- For simple text edits (titles, labels), locate `` text nodes within chart/table parts using similar traversal.\n\n---\n\n## 9. Themes, Colors, and Fonts\n\nThemes live under `/ppt/theme/themeN.xml`.\n\nThey define:\n\n- Color schemes \n- Font schemes \n- Effects and other defaults \n\nSlide masters reference themes, and slides/masts can override color mapping using `` and ``.\n\nFor text-only edits, you normally do not need to touch theme parts. Just keep them intact.\n\n---\n\n## 10. `[Content_Types].xml`: Do Not Break It\n\n`/[Content_Types].xml` declares MIME types for parts.\n\nExamples:\n\n~~~xml\n\n \n \n \n \n \n\n~~~\n\nRules:\n\n- When adding a new slide `slideN.xml`, if an `` for `/ppt/slides/slideN.xml` or a generic slide override is already present, you typically don’t need to change anything. \n- When introducing a totally new extension (e.g., `.foo`), add a `` or `` entry. \n- Do not remove or corrupt existing entries.\n\nBreaking this file can make the PPTX unreadable.\n\n---\n\n## 11. Common Editing Tasks: Recommended Algorithms\n\n### 11.1 Get All Text in All Slides\n\n1. From `ppt/presentation.xml`, list all `` in order. \n2. For each `r:id`, resolve to `slides/slideX.xml` via `presentation.xml.rels`. \n3. For each slide:\n - Parse `ppt/slides/slideX.xml`. \n - Traverse `p:cSld → p:spTree → p:sp`. \n - For each `p:sp` that has `p:txBody`:\n - For each `a:p`:\n - Collect `a:t` values, respecting `a:br/` as line breaks.\n\nOptional: use `` info to categorize text (title, body, footer, etc.).\n\n### 11.2 Replace Text with New Text (Within a Shape)\n\nScenario: Replace entire text of a title or content placeholder.\n\nSimplest robust approach:\n\n1. Identify the target shape by placeholder (`p:ph type=\"title\"`, `body`, etc.) or by existing text. \n2. Inside its `p:txBody`, remove existing `a:p` children. \n3. Insert new structure:\n\n~~~xml\n\n \n \n New text here\n \n\n~~~\n\nFor multiple paragraphs, add multiple `a:p` elements.\n\nFormatting-preserving approach:\n\n- Keep `a:pPr` and `a:rPr` nodes. \n- Only modify the content of `a:t`, preserving run structure.\n\n### 11.3 Replace an Image\n\n1. Identify which `p:pic` to edit (by placeholder type `pic`, by name, or by its current image). \n2. Find `` inside its `p:blipFill`. \n3. In `ppt/slides/_rels/slideX.xml.rels`, find `Relationship Id=\"rIdX\"`. \n4. Resolve its `Target`, e.g., `../media/image5.png` → `ppt/media/image5.png`. \n5. Overwrite that file with your new image bytes (matching format, e.g., PNG → PNG). \n\nNo changes needed in XML or `[Content_Types].xml` if you reuse the same part.\n\n### 11.4 Add a New Slide Based on an Existing Layout\n\n1. Choose a layout:\n - From slide masters and their ``, or \n - From an existing slide’s `slideLayout` relationship.\n\n2. Create a new slide part:\n - Option 1: clone an existing slide and clear the content you want to reset. \n - Option 2: construct a minimal valid `p:sld` referencing the chosen layout.\n\n3. Save it as `ppt/slides/slideN.xml` (unique filename).\n\n4. In `ppt/_rels/presentation.xml.rels`, add a relationship:\n\n~~~xml\n\n~~~\n\n5. In `ppt/presentation.xml`, add a new `` under ``:\n\n~~~xml\n\n~~~\n\n- `id` must be a unique integer within ``.\n\n6. In `ppt/slides/_rels/slideN.xml.rels`, add a `slideLayout` relationship:\n\n~~~xml\n\n~~~\n\n7. Ensure `[Content_Types].xml` has an appropriate `` for slide parts (usually already present).\n\n### 11.5 Reorder Slides\n\n1. In `ppt/presentation.xml`, locate ``. \n2. Reorder the `` elements themselves. \n\nDon’t change their `id` or `r:id`; slide order is determined solely by element order.\n\n---\n\n## 12. Robustness Guidelines for an Editing Agent\n\nTo make safe, reliable changes:\n\n1. Always use relationships:\n - Never infer master/layout/media links purely from filenames or numeric IDs.\n\n2. Preserve unknown XML:\n - If you don’t need to modify it, copy elements and attributes through unchanged.\n\n3. Preserve namespaces:\n - Keep existing `xmlns:p`, `xmlns:a`, `xmlns:r` declarations and prefixes as-is.\n\n4. Make minimal edits:\n - Focus on local text/shape/media changes instead of global restructures.\n\n5. Keep `[Content_Types].xml` consistent:\n - Only add entries when introducing new part types or extensions.\n - Avoid changing or removing existing entries.\n\n6. Maintain well-formed XML:\n - Ensure all tags are properly closed and nesting is valid.\n\n---\n\n## 13. Minimal Parsing Strategy Summary\n\nFor most real-world editing tasks, the following subset is sufficient:\n\n- Understand the ZIP and locate:\n - `ppt/presentation.xml`\n - `ppt/_rels/presentation.xml.rels`\n - `ppt/slides/slide*.xml` and their `.rels`\n - `ppt/media/*`\n\n- Use:\n - `` + `presentation.xml.rels` to enumerate slides in order.\n - `p:cSld → p:spTree → p:sp → p:txBody → a:p → a:r → a:t` to read/write text.\n - `p:pic` + `a:blip r:embed=\"...\"` + slide `.rels` to handle images.\n - `p:ph` metadata to recognize placeholders (title, body, footer, etc.).\n - `a:xfrm` (`a:off`, `a:ext`) for shape coordinates (optional for text-only edits).\n\nStaying within this structure lets you:\n\n- Extract and modify text (titles, bullet lists, body text). \n- Replace images without disturbing layouts. \n- Add, remove, and reorder slides. \n- Respect existing themes, masters, and layouts without needing to deeply modify them.",[],[348],{"_102":356,"_95":357,"_153":358,"_155":-5,"_156":136,"_157":359,"_159":160,"_104":360,"_99":361,"_162":252,"_164":362,"_165":22,"_166":-5,"_167":363,"_98":364},"80f5e182-c34c-11f0-a254-42dde27851f2","README.md","019a84ad-a102-73ff-a194-7c1b3c0d93d0",329,["D",1763339379692],37,"# PPTX XML Editing Bot - Content-Aware AI System\n\n## 🎯 **CURRENT STATUS: Content-Aware AI Editing Restored! 🎉**\n\nThe sophisticated content-aware AI system has been fully restored after fixing syntax errors.\n\n### **✅ Current Features**\n- **Multi-Slide Content-Aware AI**: AI sees ALL slides and their text elements with context\n- **Multi-Slide Operations**: AI can edit multiple slides in one request (\"make whole deck about cats\")\n- **Text Box Relationship Understanding**: AI knows which elements belong to the same text box\n- **Smart Slide Targeting**: AI automatically finds the right slide based on content\n- **Text Box Replacement**: AI can replace entire text boxes (all bullets) or individual elements\n- **Markdown to PowerPoint XML**: AI generates clean markdown, system converts to proper XML\n- **Precise Element Selection**: AI chooses the right text element using slide-aware IDs (1A, 2B, 3C, etc.)\n- **Natural Language Commands**: \"Change the executive summary to something about cats\"\n- **Surgical XML Editing**: Makes precise edits to specific elements without corruption\n- **File Upload/Download**: Drag & drop PPTX processing\n- **Thread-Based Chat**: Multiple conversation threads with persistent history\n\n### **🔧 Recently Fixed**\n- **Syntax Errors**: Resolved all missing braces and orphaned code blocks\n- **Content-Aware System**: Fully restored the sophisticated AI processing\n- **Element Mapping**: Restored precise element targeting and surgical edits\n\n### **Example Commands That Work**\n```\n\"Change the main heading on the first slide to 5 words about tortoises\"\n\"Replace the title with something about giraffes\"\n\"Update the large text to say hello world\"\n\"Add some lorem ipsum to the slide\"\n```\n\n## 🚀 **MILESTONE ACHIEVED: Content-Aware AI Editing**\n\n### **How It Works**\n1. **Content Extraction**: System reads slide XML and extracts all text elements with context\n2. **AI Analysis**: OpenAI sees actual slide content: `[A] \"Kwaku Slide One\" (18pt, top-center)`\n3. **Smart Selection**: AI chooses the right element based on user's natural language\n4. **Precise Edit**: System makes surgical XML changes to exactly the right element\n\n### **Technical Architecture**\n\n#### **Content-Aware AI Flow**\n```typescript\ninterface SlideText {\n text: string; // \"Welcome to Company\"\n fontSize: number; // 44\n position: {...}; // top-center\n elementId: string; // \"A\"\n xmlPath: string; // For precise editing\n}\n```\n\n#### **AI Decision Making**\n```\nPPTX Content:\n\nSlide 1 text elements:\nText Box TB0 (title):\n [1A] \"Welcome to Company\" (44pt, top-center)\nText Box TB1 (body):\n [1B] \"Quarterly Review\" (18pt, middle-left)\n\nSlide 2 text elements:\nText Box TB0 (title):\n [2A] \"Executive Summary\" (36pt, top-center)\nText Box TB1 (body):\n [2B] \"Key findings and recommendations\" (16pt, middle-left)\n [2C] \"First bullet point\" (16pt, middle-left)\n [2D] \"Second bullet point\" (16pt, middle-left)\n\nUser: \"Change the executive summary content to 3 bullet points about cats\"\nAI Decision: \"User wants to replace entire body content under Executive Summary\"\nAI Response: {\n \"textBoxId\": \"TB1\", \n \"slideNumber\": 2,\n \"content\": \"• Cats are independent animals\\n• Cats love to sleep in sunny spots\\n• Cats communicate through purring\"\n}\nSystem: Parses markdown → Generates proper PowerPoint XML → Replaces entire text box\n```\n\n#### **PPTX XML Processing**\nThe system uses deep knowledge of PPTX structure to make precise edits. See [`README-how_pptx-works.md`](./README-how_pptx-works.md) for complete technical details on:\n- PPTX ZIP structure and relationships\n- XML parsing and text extraction\n- Safe editing patterns that preserve file integrity\n\n## 🏗️ **Important: Structure of Slide Elements**\n\nUnderstanding how PowerPoint organizes content is crucial for the AI system to make intelligent editing decisions.\n\n### **Text Boxes vs. Text Elements**\n- **Text Box** = `` element (PowerPoint shape container)\n- **Text Elements** = `` tags within the text box (individual text runs)\n- **One text box can contain multiple paragraphs** (``) with multiple text runs (``)\n\n### **How Bullets Are Actually Structured**\n```xml\n\n\n\n\n Executive Summary\n\n\n\n\n First bullet point\n Second bullet point\n Third bullet point\n\n```\n\n### **Why Element-by-Element Parsing Is Safer**\n- **Preserves XML Structure**: Each `` replacement maintains bullet formatting\n- **Avoids Corruption**: No need to reconstruct complex XML with indentation, bullet chars, etc.\n- **Granular Control**: Can replace individual bullets or groups of bullets\n- **User Flexibility**: \"Change first bullet\" vs \"change all executive summary content\"\n\n### **PowerPoint Placeholder Types**\n- `` - Main slide titles\n- `` - Content areas (usually containing bullets)\n- `` - Centered titles\n- `` - Subtitles\n\n### **AI Strategy: Text Box Awareness Without Grouping**\nThe AI sees individual elements but understands their relationships:\n```\nText Box 1 (title): [2A] \"Executive Summary\"\nText Box 2 (body): [2B] \"First bullet\" [2C] \"Second bullet\" [2D] \"Third bullet\"\n```\n\nWhen user says \"change executive summary content\" → AI targets the entire text box\nWhen user says \"change executive summary title\" → AI targets element 2A\n\n### **Content Generation: Markdown to PowerPoint XML**\nThe system uses a clean separation between content generation and technical implementation:\n\n1. **AI generates markdown-like content**:\n ```\n • First bullet point about cats\n • Second bullet point about dogs \n • Third bullet point about birds\n ```\n\n2. **System parses and converts to PowerPoint XML**:\n ```xml\n \n First bullet point about cats\n Second bullet point about dogs\n Third bullet point about birds\n \n ```\n\n**Benefits:**\n- ✅ **AI focuses on content**: No need to understand PowerPoint XML complexity\n- ✅ **System handles formatting**: Proper bullet structure, indentation, XML escaping\n- ✅ **Clean separation**: Content generation vs. technical implementation\n- ✅ **Flexible**: Supports bullets, sub-bullets, plain text, mixed content\n\n### **Success Criteria Met ✅**\n- [x] AI sees actual slide content (not blind heuristics)\n- [x] Smart element selection based on context\n- [x] Natural language commands work reliably\n- [x] Precise XML editing without file corruption\n- [x] Clean, maintainable codebase under 500 lines per file\n\n## 🏗️ **Clean Project Structure**\n\n```\n├── backend/\n│ ├── index.ts # Routes & database (~200 lines)\n│ ├── pptx.ts # Content-aware PPTX processing (~400 lines) \n│ └── ai.ts # AI command parsing (~150 lines)\n├── frontend/\n│ └── index.html # Simple web interface\n└── README.md # This file\n```\n\n### **Architecture Benefits**\n- **Content-Aware**: AI sees actual slide content before making decisions\n- **Precise Editing**: Surgical XML changes to specific elements\n- **Clean Code**: All files under 500 lines, clear separation of concerns\n- **No Legacy Cruft**: Removed all broken fallback code\n- **Maintainable**: Easy to find, understand, and modify functionality\n\n### **API Endpoints**\n- `GET /` - Main chat interface\n- `GET /health` - Health check\n- `GET /api/chat` - Get chat history \n- `POST /api/chat` - Send message (with content-aware AI)\n- `POST /api/upload` - Upload PPTX file\n- `GET /api/download/:fileId` - Download processed file\n\n## 🔧 **Development Guidelines**\n\n### **Code Organization**\n- **File Size Limit**: Keep files under **500 lines** for maintainability\n- **Single Responsibility**: Each file has one clear purpose\n- **No Legacy Cruft**: Remove broken code instead of keeping \"fallbacks\"\n- **Content-Aware First**: AI should see actual content, not use blind heuristics\n\n### **Documentation Requirements**\n- **Keep README Updated**: When you make code changes, update this README to reflect the current state\n- **Technical Reference**: See [`README-how_pptx-works.md`](./README-how_pptx-works.md) for detailed PPTX XML structure documentation\n- **Document Milestones**: Update status sections when major features are completed\n- **Remove Outdated Content**: Delete old implementation details that no longer apply\n\n### **Key Principles**\n1. **Intelligence Over Heuristics**: Let AI see content and make smart decisions\n2. **Precise Over Broad**: Make surgical edits to specific elements\n3. **Clean Over Complex**: Remove unnecessary complexity and legacy code\n4. **Working Over Perfect**: Focus on functionality that actually works\n5. **Document Over Assume**: Keep documentation current with implementation\n\n## 📚 **Additional Documentation**\n\n- **[`README-content-editing.md`](./README-content-editing.md)** - How the AI analyzes content and converts markdown to PowerPoint XML\n- **[`README-development.md`](./README-development.md)** - Development roadmap and plans for slide structural operations\n- **[`README-how_pptx-works.md`](./README-how_pptx-works.md)** - Comprehensive guide to PPTX XML structure, relationships, and editing patterns\n\n---\n\n**Monitoring**: GET requests to `/health` return \"PPTX Bot is running\" for health checks.",[],[357],"files",[150,169,180,191,200,210,216,224,230,238,245,257,265,273,280,287,296,306,315,327,337,346,355],"counts",{"_369":32,"_370":158,"_371":32,"_372":373},"pulls","branches","forks","envVars",{"_374":375,"_376":377},"count",7,"emptyCount",6,"contents",[380,383,386,389,392,395,398,401,404],{"_102":151,"_95":152,"_153":154,"_155":-5,"_156":136,"_157":158,"_159":160,"_104":381,"_99":32,"_162":163,"_164":-5,"_165":22,"_166":-5,"_98":382},["D",1763162562720],[152],{"_102":211,"_95":212,"_153":213,"_155":-5,"_156":136,"_157":158,"_159":160,"_104":384,"_99":32,"_162":163,"_164":-5,"_165":22,"_166":-5,"_98":385},["D",1763162562720],[212],{"_102":225,"_95":226,"_153":227,"_155":-5,"_156":136,"_157":158,"_159":160,"_104":387,"_99":32,"_162":163,"_164":-5,"_165":22,"_166":-5,"_98":388},["D",1763162562720],[226],{"_102":239,"_95":240,"_153":241,"_155":-5,"_156":136,"_157":242,"_159":160,"_104":390,"_99":32,"_162":163,"_164":-5,"_165":22,"_166":-5,"_98":391},["D",1763163421719],[240],{"_102":258,"_95":259,"_153":260,"_155":-5,"_156":136,"_157":261,"_159":160,"_104":393,"_99":263,"_162":163,"_164":-5,"_165":22,"_166":-5,"_98":394},["D",1763192442978],[259],{"_102":328,"_95":329,"_153":330,"_155":-5,"_156":136,"_157":331,"_159":160,"_104":396,"_99":333,"_162":252,"_164":334,"_165":22,"_166":-5,"_98":397},["D",1763339841532],[329],{"_102":338,"_95":339,"_153":340,"_155":-5,"_156":136,"_157":341,"_159":160,"_104":399,"_99":32,"_162":252,"_164":343,"_165":22,"_166":-5,"_98":400},["D",1763339354975],[339],{"_102":347,"_95":348,"_153":349,"_155":-5,"_156":136,"_157":350,"_159":160,"_104":402,"_99":333,"_162":252,"_164":352,"_165":22,"_166":-5,"_98":403},["D",1763318207577],[348],{"_102":356,"_95":357,"_153":358,"_155":-5,"_156":136,"_157":359,"_159":160,"_104":405,"_99":361,"_162":252,"_164":362,"_165":22,"_166":-5,"_98":406},["D",1763339379692],[357],"upstreamProject","upstreamBranch","emptyEnvVars","isNew","flags",[],"routes/_app.x.$handle.$name.environment-variables",{"_93":94,"_95":96,"_100":415,"_372":420},{"_102":103,"_95":96,"_104":416,"_106":-5,"_107":22,"_108":109,"_110":-5,"_111":-5,"_14":112,"_113":-5,"_114":22,"_115":-5,"_116":32,"_16":117,"_118":119,"_120":119,"_121":122,"_123":122,"_124":122,"_125":119,"_126":119,"_23":417,"_130":418,"_133":419,"_147":22},["D",1763162562720],{"_93":94,"_128":129},{"_93":94,"_132":129,"_14":112},{"_99":135,"_95":43,"_102":136,"_137":103,"_138":-5,"_139":-5,"_140":141,"_142":143,"_144":117,"_145":96,"_146":94},[421,428,433,437,440,444,448],{"_422":423,"_104":424,"_425":-5,"_110":426,"_427":160},"key","MONGODB_CONNECTION_URL",["D",1763162562720],"updatedAt","james","isEmpty",{"_422":429,"_104":430,"_425":431,"_110":432,"_427":22},"OPENAI_API_KEY",["D",1763162562720],["D",1763162903451],"openAI key for swayable bot",{"_422":434,"_104":435,"_425":-5,"_110":436,"_427":160},"OPENAI_ASSISTANT_ID",["D",1763162562720],"V2 assistant ID from openAI",{"_422":438,"_104":439,"_425":-5,"_110":72,"_427":160},"SLACK_BOT_USER_ID",["D",1763162562720],{"_422":441,"_104":442,"_425":-5,"_110":443,"_427":160},"slackToken",["D",1763162562720],"OAuth token?",{"_422":445,"_104":446,"_425":-5,"_110":447,"_427":160},"slackVerificationToken",["D",1763162562720],"Slack token",{"_422":449,"_104":450,"_425":-5,"_110":451,"_427":160},"TOOL_API_KEY",["D",1763162562720],"from js-bot-API","actionData","errors"]

Key	Description
MONGODB_CONNECTION_URL	james
OPENAI_API_KEY	openAI key for swayable bot
OPENAI_ASSISTANT_ID	V2 assistant ID from openAI
SLACK_BOT_USER_ID
slackToken	OAuth token?
slackVerificationToken	Slack token
TOOL_API_KEY	from js-bot-API

jslez

slidebot

Environment variables