WhiteHades/metldr

Private, on-device AI summaries in your browser.

4

stars

388

commits

TypeScript

primary language

Jan 27, 2026

updated

README

metldr

version users

metldr is a privacy first, local ai powered chrome extension designed to enhance productivity by integrating summarisation, smart replies, and dictionary lookups directly into your browser.

install

features

  • email intelligence: automatically summarises threads, extracts key facts (dates, amounts), and generates context aware reply suggestions in gmail.
  • page intelligence: detects article content, provides bulleted summaries, and allows you to chat with any webpage.
  • pdf intelligence: full featured pdf reader with summarisation, q&a (rag), and local file support.
  • instant definitions: three tier dictionary lookup (local > api > llm) for instant word definitions.
  • privacy first: runs entirely on device using chrome built in ai (gemini nano) or local ollama models.

usage

setting up ai providers

metldr uses chrome's built in gemini nano model by default.

  1. ensure you are on chrome version 138+
  2. enable "prompt api for gemini nano" in chrome://flags
  3. the extension will automatically detect availability

option 2: ollama (for power users)

run larger models (llama 3, qwen, etc.) locally on your machine.

  1. install ollama
  2. run a model: ollama run llama3.2
  3. go to metldr settings > ai provider > select "ollama"
  4. ensure ollama is running at http://127.0.0.1:11434

development

commands

commanddescription
npm run devstart development server (hmr for ui)
npm run buildbuild all targets (main, content, sandbox)
npm run type-checkrun typescript validation
npm run download:modelsdownload required ai models to public/

privacy

metldr is designed with a strict privacy first architecture:

  • local execution: all ai processing happens on your device (via chrome ai or localhost ollama).
  • no data collection: we do not collect, store, or transmit your emails, browsing history, or personal data.
  • offline capable: core features work without internet access.

built with ❤️ by efaz

Contributors

WhiteHades

388 commits

WhiteHades/metldr

Private, on-device AI summaries in your browser.

4

stars

388

commits

TypeScript

primary language

Jan 27, 2026

updated

README

metldr

version users

metldr is a privacy first, local ai powered chrome extension designed to enhance productivity by integrating summarisation, smart replies, and dictionary lookups directly into your browser.

install

features

  • email intelligence: automatically summarises threads, extracts key facts (dates, amounts), and generates context aware reply suggestions in gmail.
  • page intelligence: detects article content, provides bulleted summaries, and allows you to chat with any webpage.
  • pdf intelligence: full featured pdf reader with summarisation, q&a (rag), and local file support.
  • instant definitions: three tier dictionary lookup (local > api > llm) for instant word definitions.
  • privacy first: runs entirely on device using chrome built in ai (gemini nano) or local ollama models.

usage

setting up ai providers

metldr uses chrome's built in gemini nano model by default.

  1. ensure you are on chrome version 138+
  2. enable "prompt api for gemini nano" in chrome://flags
  3. the extension will automatically detect availability

option 2: ollama (for power users)

run larger models (llama 3, qwen, etc.) locally on your machine.

  1. install ollama
  2. run a model: ollama run llama3.2
  3. go to metldr settings > ai provider > select "ollama"
  4. ensure ollama is running at http://127.0.0.1:11434

development

commands

commanddescription
npm run devstart development server (hmr for ui)
npm run buildbuild all targets (main, content, sandbox)
npm run type-checkrun typescript validation
npm run download:modelsdownload required ai models to public/

privacy

metldr is designed with a strict privacy first architecture:

  • local execution: all ai processing happens on your device (via chrome ai or localhost ollama).
  • no data collection: we do not collect, store, or transmit your emails, browsing history, or personal data.
  • offline capable: core features work without internet access.

built with ❤️ by efaz

Contributors

WhiteHades

388 commits

Languages

TypeScript

69.1%

Vue

25.1%

JavaScript

4.7%