JevFilterForX

It is a browser extension for turning the X timeline into a higher-signal feed. It scores visible posts, explains each score with lightweight labels, and folds posts that match your filters while keeping them available to expand.

Jev rates information signal, actionability and originality from 0 to 4. The weighted result is (signal × 40% + actionability × 35% + originality × 25%) ÷ 4 × 100, banded as 0-29 Low, 30-69 Medium and 70-100 High.

Download the Chrome ZIP from the latest release, unzip it, open chrome://extensions, enable Developer mode and choose Load unpacked on the folder containing manifest.json. You can also download the CI artifact or build locally with Node.js 22+: npm install, then npm run build, and load dist/chrome-mv3.

For real scoring, yes. Open JevFilterForX settings and add a TypeSafe API key; without one the extension stays in local mock mode so the UI and filtering flow can still be explored. The key is stored on the current device and is not synced to a Google account.

The options page runs in its own tab and covers the API key, the Jev API address, language, topics, value preferences, category filters and score thresholds. Classification requests go to the Jev API address configured there, which defaults to TypeSafe's API.

The README argues timeline filtering is a high-frequency classification task, and Jev is designed for fast, low-cost requests, which makes continuous filtering practical compared with routing every post through a general-purpose LLM.

The README visible in the digest cuts off mid-sentence in the privacy and permissions section, the repository lists no license, and no accuracy or latency figures for the scoring model are published. Releases are built automatically by CI when a v* tag is pushed.