FeedForward Desktop
Private practice · no accounts · nothing uploaded

Practice your draft before you submit it.

Open the rubric your instructor shared, paste your draft, and get calm, rubric-aligned feedback from a model you control — watch your aim close on the bullseye with every revision.

Free & open source (MIT) · Windows, macOS & Linux · works fully offline with Ollama, or bring your own key

How it works

Three steps, as many times as you like.

01

Open your rubric

Your instructor shares a .ffrubric file — the same criteria your real submission is assessed against. Open it in the app.

02

Add a draft & a model

Paste or open your draft, then pick your engine: a free local model via Ollama (fully offline), your class server, or your own API key.

03

Aim, revise, repeat

Feedback lands per rubric criterion with strengths and next moves — and a dartboard that shows your aim closing on the bullseye across drafts.

Try it right here

See the feedback a draft gets.

Pick a sample draft and assess it. This demo plays canned feedback so it can run in your browser — the app runs a real model of your choosing, with the same rubric-aligned result.

Rubric: Essay 1 — Critical Analysis (Argument 40% · Evidence 60%), the same sample .ffrubric you can open in the app.

Feedback appears here — bullseyes, level words, strengths and next moves.
For the curious

Taste the real engine in 60 seconds.

The app’s engine is an ordinary Python package, feedforward-practice. If you have uv and Ollama, you can run a genuine assessment from your terminal before installing anything:

# grab the sample rubric, write a couple of paragraphs into draft.md, then:
curl -LO https://michael-borck.github.io/feed-forward-desktop/sample.ffrubric
uvx --from "feedforward-practice" feedforward-practice assess \
    --rubric sample.ffrubric --draft draft.md --model llama3.2:3b
# remote endpoint or your own key? add: --base-url … --api-key …
Download

Free, open source, three platforms.

First run sets up the feedback engine automatically (needs Python 3.11+). Fully offline once a local model is pulled.
All versions and release notes on the GitHub releases page.

For instructors

Give your students unlimited practice.

FeedForward Desktop is the standalone companion to FeedForward, the instructor-controlled feedback platform. Export any assignment’s rubric as a .ffrubric file and share it with your cohort — students rehearse privately against your actual criteria, then submit through your FeedForward site as usual. Practice feedback is formative guidance, never a grade, and no student work touches a server you have to look after.