Tell Pepper what you care about, in plain words. She researches it around the clock, writes the bulletin with an on-device model, and presents it from her own studio โ with her own voice. No cloud, no accounts, no telemetry.
Open source, AGPL-3.0 ยท Benchmarked, not vibes ยท Full on-device brain on Apple Silicon
Install โ thirty seconds to air
curl -fsSL pepper.software | shpepper startPrefer npm? npm i -g @pepperchan/pepper ยท Node 18.17+ ยท everything lives in ~/.pepper ยท her voice tier: pepper voice install
"Watch open-source robotics and GPU prices for me." Plain words in her studio โ she works out the rest and starts sweeping the wire every fifteen minutes.
An on-device model turns the wire into a sourced broadcast โ vendor claims labeled, rumors kept off the air, nothing invented. Your reading interests never leave the machine.
A 3D newsroom in your browser: chyron, ticker, her designed voice, deep-dive reports on demand โ and a one-command export that puts your station on the public web.
Benchmarked, not vibes
MoltBench scores wire-to-report work: grounding, attribution, and adjudication of deliberately contaminated bundles, under blind judges. Pepper's first model failed it โ that failure became the release gate. Her current desk model cleared it: +23 grounding points over its own untuned base, 10/12 blind wins against Qwen3-4B-2507, 11/12 against Qwen2.5-7B. Protocol and bundles โ
| Model | Grounding | Blind vs pepper-desk |
|---|---|---|
| pepper-desk-e2b (2B active) | 88.6% | โ |
| Qwen3-4B-Instruct-2507 | 88.3% | 2 / 12 |
| Qwen2.5-7B-Instruct | 77.5% | 1 / 12 |
| gemma-4-e2b (her untuned base) | 65.2% | โ |
Run a station of your own
The reference station at pepper.watch โ served static from Cloudflare Pages, her voice included.
Run Pepper locally, then pepper export โ npx wrangler pages deploy pepper-site. Your beats, your bulletins, your voice files โ a complete station from your own desk.