hi-fangj/dsh-models-radar

dsh-models-radar adds a Model Radar page to DSH Settings that reads crowd-benchmarked capability data from deng.codexradar.com and shows the selected session model's live DeepSWE score beside the composer model selector: best-effort-per-base ranking with Harness attribution badges, 24h and 7d IQ trend windows, cost-by-IQ charts (composite, time and price) and community 7-day/24-hour experience ratings, all served through a same-origin host proxy with freshness windows so the browser never hits upstream directly and no credentials are requested.

dsh-models-radar adds a Model Radar page to DSH Settings that reads public crowd-benchmarked capability data from deng.codexradar.com and shows the selected session model's live DeepSWE score in the composer tool row, left of the model selector. The Settings page ranks base models best-effort-per-base on a fixed 0-110 scale with a 24h trend signal and a Harness attribution badge (Codex / Claude Code / DSH / ZCode / Grok / Kimi Code / Antigravity / CodeBuddy), expands to a base's full reasoning-effort ladder, and offers 24h/7d IQ trend tabs plus cost-by-IQ charts from three angles (composite, time and price, with color by base, shape by reasoning effort, and same-base tiers joined by ladder lines where upper-left is more efficient). A community ratings chart shows the codexradar.com community's 0-10 experience scores over rolling 7-day and 24-hour windows, and the live readout beside the model selector matches model@reasoningEffort exactly and opens a capability popover for cross-base comparison.

Models & Providers ★ 3
View on GitHub ↗

Install

dsh plugin --profile web add github:hi-fangj/dsh-models-radar

Copied verbatim from the README's Install block ('The simplest installation fetches the repository directly into the web profile'). The README also documents a git+https://github.com/hi-fangj/dsh-models-radar.git form.

Compatibility

Lightweight, credential-free and offline-tolerant: the browser never hits upstream directly (a same-origin host proxy does), freshness windows mean zero upstream requests inside a window, and the latest local snapshot serves as a fallback. Score attribution is best-effort - unmatchable base models get no Harness badge rather than a guess - and community ratings come from the codexradar.com site's own rolling windows. After a Git install, refresh http://127.0.0.1:3080 and restart DSH once if the process does not hot-load the new package.

Details

Alternatives

1624318455/dsh-plugin-adapter · aa2246740/dsh-antigravity-oauth · AmmarByFar/dsh-trustedrouter

More plugins in Models & Providers

Browse more in Models & Providers

Guides for Models & Providers plugins