jev-conductor

Featured

Curate a live Jev improvisation in OpenHarness's viewer, where Jev composes chords, bass and melody bar by bar in real time.

AI & Automation 957 stars 79 forks Updated today MIT

Install

View on GitHub

Quality Score: 93/100

Stars 20%
99
Recency 20%
100
Frontmatter 20%
70
Documentation 15%
100
Issue Health 10%
50
License 10%
100
Description 5%
100

Skill Content

# Jev Conductor composition Jev Conductor shows a live scrolling score: Jev (TypeSafe's System One model) picks the next bar's chord, bass note, lead phrase, mood and energy on a bar clock, and the pane performs it with Web Audio. The agent curates `piece.json` — the musical constraints Jev improvises within. ## The loop 1. Update `piece.json`. The viewer watches it and reshapes Jev's playing live — no restart, no second server. 2. `node "$JEV_DSH/toolchain/check.mjs"` verifies the workspace's `piece.json` is valid and range-checked, and reports the current verdict. Run it before you call a piece done. 3. Listen (or reason about the design): a narrow consonant scale + small harmonic palette yields coherent improv; a wide or chromatic scale yields chaos. Name the moods you want to hear. ## Reading Jev The viewer colors each bar by mood and shows Jev's choices (chord, bass, lead, mood, energy). Because Jev's *choice* questions return a probability distribution, `jev.mjs`'s mock and the real model both pick a `choice` — but what matters musically is the *constraint design*: a strong tonic, a small consonant scale, vivid mood labels. Design those well and any path Jev walks sounds like music. ## Verifying a piece `node "$JEV_DSH/toolchain/check.mjs"` returns non-zero when `piece.json` is invalid (bad notes, out of range, empty palettes). It does not replace judging the *music*. The real test is coherence: with a tonic-centered chord list and a consonant scale, Jev...

Details

Author
autonomous-ai
Repository
autonomous-ai/openharness
Created
1 months ago
Last Updated
today
Language
C
License
MIT

Similar Skills

Semantically similar based on skill content — not just same category