Openai Document
OpenAI-powered document and webpage generation — analyze and enhance .pptx/.docx, design PDFs from natural language, compose decks from outlines using your own theme template, and generate self-contained HTML/CSS pages from reference designs via vision. Vault-based API key, headless Chromium for rendering.
Docs: replace a real company name used as a fictional prospect in the design example with the Acme placeholder. No functional changes.
| Argument | Type | Description |
|---|---|---|
| filePath | string | Absolute path to the .pptx or .docx file |
| outputName | string |
| Argument | Type | Description |
|---|---|---|
| instructions | string | Design and content direction — describe the visual style, layout, colors, and what content to include. E.g., 'Modern proposal for Snap Inc, Kentucky blue and white color scheme, clean layout with geometric accents, two-column sections where appropriate. Cover page with large color block and company name.' |
| contentSource? | string | Optional: absolute path to a .docx file to extract content from. The design method will pull the text and restructure it into the new design. If omitted, GPT generates content based on instructions alone. |
| context? | string | Additional context — company info, client details, scope data, brand guidelines, etc. |
| outputName | string |
| Argument | Type | Description |
|---|---|---|
| filePath | string | Absolute path to the .pptx or .docx file |
| instructions | string | What to change in the text content |
| model | enum | |
| outputName | string |
| Argument | Type | Description |
|---|---|---|
| templatePath | string | Absolute path to the template .pptx or .docx file |
| instructions | string | What content to generate |
| context? | string | Additional context |
| model | enum | |
| outputName | string |
| Argument | Type | Description |
|---|---|---|
| outline | string | Outline, brief, or source notes that compose will expand into slide content. |
| templatePath | string | Absolute path to the .pptx template. Each slide in the template becomes one card; pre-trim or duplicate slides in the template to control the final card count. |
| tone? | string | Voice/tone (e.g., 'executive briefing', 'technical deep dive', 'sales pitch'). |
| audience? | string | Who's reading the deck (e.g., 'CIO and IT directors', 'AV trade press'). |
| additionalInstructions? | string | Extra direction — what to emphasize, what to avoid, brand voice notes. |
| includeSpeakerNotes | boolean | Generate speaker notes for each slide (written into ppt/notesSlidesN.xml when present). |
| outputName | string |
| Argument | Type | Description |
|---|---|---|
| brief | string | What the page is for and the content/copy it should contain. Include headings, body copy, and any required sections. |
| references | array | Reference design images — local file paths (.png/.jpg/.webp/.gif) or http(s) URLs. The model uses these for visual cues only; copy must come from the brief. |
| styleNotes? | string | Extra style direction — brand color overrides, fonts to prefer, accessibility requirements, etc. |
| fontFiles | array | Local font files embedded as @font-face data URIs AFTER generation. The model is only told the families exist and is forbidden from writing @font-face/@import/base64 itself — font bytes must never round-trip through the model. |
| inlineLocalAssets | boolean | After generation, inline every local absolute-path image reference (img src and CSS url()) as a data URI so the HTML file is fully self-contained and portable. |
| deckContract | boolean | Treat the output as a paginated slide deck: append hard sizing/fragmentation rules to the prompt, inject print-safety guard CSS after generation, validate referenced local assets exist, and check the rendered PDF's page count against the slide count. |
| pageWidth | number | Deck page width in px (used when deckContract is true). |
| pageHeight | number | Deck page height in px (used when deckContract is true). |
| responsive | boolean | Include mobile-first responsive layout with media queries. |
| renderPreview | boolean | If true, render the HTML to a PNG screenshot (and PDF) via headless Chromium. |
| outputName | string |
Resources
webpageDesign: fontFiles embedding, inlineLocalAssets, deckContract print guards + preflight (slide/page count, missing assets). Deterministic post-gen pipeline replaces model-side base64.
Fix the LICENSE link target in the README.
Added 1, removed 1 models
Security: genericized hardcoded host paths (now via SWAMP_OUTPUT_COPY_DIR env), README example paths, and standardized LICENSE. No functional change.
Replace jszip@3.10.1 with fflate@0.8.3 for .docx/.pptx read/modify/rezip. Clears the non-allowlisted dual-license and stale-dependency warnings; fflate is MIT and actively maintained. Behavior-preserving (OOXML round-trip validated on real files); zip backend moved to a sibling _ooxml.ts adapter.
Align model version fields with manifest (2026.05.26.1); republish under swamp 20260526 conventions
Add JSDoc to model entrypoints for rubric v2 symbols-docs
Fix JSDoc so symbols-docs factor passes — wrap type name in backticks so it isn't parsed as a JSDoc tag. Also bumps openai-document to v2026.05.13.1 (was staged at 2026.04.29.1, never published).
Modified 1 models. updated labels
Add curated README and LICENSE to tarball (additionalFiles), bringing extension's Swamp Club quality grade from F/B to A-/B.
Initial release. Four methods: analyze (parse .pptx/.docx structure), enhance (rewrite text preserving formatting), generate (populate templates with GPT content), design (generate pixel-perfect PDFs from natural-language instructions via headless Chromium).
- Has README or module doc2/2earned
- README has a code example1/1earned
- README is substantive1/1earned
- Most symbols documented1/1earned
- No slow types (deprecated)1/1earned
- Dependencies pass trust audit2/2earned
- Has description1/1earned
- Platform support declared (or universal)2/2earned
- License declared1/1earned
- Verified public repository2/2earned