Mdfy

Mdfy turns text, screenshots, slide photos, documents, and public articles into clean Markdown without leaving Obsidian. It uses your own OpenAI-compatible endpoint and can follow the structure and terminology of the note you are editing.

Features

• Format pasted text or existing Markdown.
• Start from text selected in the current note; Mdfy fills the Text source automatically.
• Process up to 10 ordered JPEG, PNG, WebP, or HEIC images with a vision-capable model. HEIC is converted locally to JPEG.
• Read one UTF-8 TXT or MD file through Chat Completions, or send one PDF, DOCX, or PPTX file to a provider that supports the Responses API.
• Extract the readable content of public, static HTML articles.
• Use the current note as optional style and deduplication context.
• Add a one-off instruction such as “summarize this” or “make a table”.
• Review and edit the result before inserting, replacing a selection, or copying it.
• Connect directly to OpenAI-compatible Chat Completions APIs, including local endpoints that do not require a key.

Requirements

• Obsidian 1.11.4 or newer on desktop.
• An OpenAI-compatible /chat/completions endpoint.
• A vision-capable model for image input.
• A provider and model with Responses API file-input support only if you want to use PDF, DOCX, or PPTX. Support varies among OpenAI-compatible endpoints.

Install

Mdfy is not yet listed in Obsidian's Community plugins directory, so install this release manually:

1. Open the latest release and download main.js, manifest.json, and styles.css from Assets. Do not download the source code archive instead.
2. Open your Obsidian vault folder. Inside it, create .obsidian/plugins/mdfy/ if it does not already exist. The .obsidian folder may be hidden in your file manager.
3. Put all three downloaded files directly in mdfy/, so the paths end with mdfy/main.js, mdfy/manifest.json, and mdfy/styles.css.
4. Restart Obsidian. Open Settings → Community plugins, turn on community plugins if needed, and enable Mdfy under installed plugins.
5. Open Settings → Community plugins → Mdfy to enter your API base URL and model, then select or create an API key in Obsidian SecretStorage. Use Test connection to check the settings.

To update a manual installation, replace the same three files with the files from a newer release and restart Obsidian.

Build from source

npm install
npm run build

Copy manifest.json, main.js, and styles.css into <vault>/.obsidian/plugins/mdfy/, then enable Mdfy in Settings → Community plugins. During development, npm run dev watches and rebuilds main.js.

Configure

Open Settings → Community plugins → Mdfy and set:

1. The API base URL, usually ending in /v1.
2. The exact model name exposed by the endpoint.
3. An API key selected through Obsidian SecretStorage, unless the endpoint does not require one.

Use Test connection to verify the configuration. The test sends a minimal Chat Completions request and may incur a small provider charge.

Use

Open a Markdown note and choose Mdfy: Open Mdfy from the command palette, assign a hotkey, or use the ribbon icon. Choose Text, URL, Images, or Files, optionally add an instruction, and generate Markdown. If you select text before opening Mdfy, it fills the Text source, and the selected passage is excluded from the note context so the model can rewrite it. Images can be selected, pasted, or dropped into the image area. Files accepts one TXT, MD, PDF, DOCX, or PPTX file up to 20 MB. Nothing is written to the note until you choose an insertion action.

TXT and MD are decoded locally as UTF-8, subject to the configured text-character limit, and sent through /chat/completions. HEIC is converted locally to JPEG before sending; only the first frame of a multi-image HEIC is used. PDF, DOCX, and PPTX are read in memory and sent in full to the configured provider through /responses; Mdfy does not upload them to a separate file storage service. OpenAI's Responses API sends PDF text and page images to vision-capable models, but extracts only text from DOCX and PPTX. Embedded images and charts in those formats may be omitted. Other providers may behave differently or reject file inputs. The Test connection button checks Chat Completions only and does not verify Responses file support.

URL article extraction supports public HTML returned by the server. It does not run page JavaScript, bypass authentication or paywalls, or process PDF links; choose a local PDF in Files instead.

Privacy

Mdfy has no backend, accounts, analytics, or telemetry. Requests go directly from Obsidian to the endpoint you configure. Depending on your choices, a request can contain:

• the source text, images, complete selected document, or extracted article;
• your additional instruction;
• the complete active note when Use current note as context is enabled.

API keys are referenced through Obsidian SecretStorage and are not stored in Mdfy's data.json. Your provider's privacy, retention, and pricing policies apply to every request.

Quality checks

npm run typecheck
npm run lint
npm test
npm run build

Release

Run npm version patch, npm version minor, or npm version major to keep package.json, manifest.json, and versions.json synchronized. Push the generated version tag without a v prefix. The release workflow verifies the tag, rebuilds the plugin, and publishes manifest.json, main.js, and styles.css as GitHub Release assets.

License

MIT. The bundled HEIC decoder uses LGPL-licensed components; see third-party notices, the LGPL license text, and the GPL license text. These notices and license texts are embedded in the installed main.js as well as attached to each release.

Only import HEIC files you trust. The bundled heic-to 1.5.2 currently uses libheif 1.22.2, which has published security advisories. Mdfy processes HEIC locally, but this bundled decoder has not yet been updated to a patched libheif version.