shengshifantang/dsh-evidence-arena

README excerpt

dsh-evidence-arena English 中文 Evidence Arena is a multi-model coding comparison workbench independent of chat Sessions . Two or three Builders solve the same task in isolated Git worktrees and independent DSH child runtimes. Reproducible gates, independent Reviewers, and per-file diffs make the results inspectable. Only an explicitly confirmed candidate is written back to the original workspace. Delivery status Question Current answer Does it require patches to official DSH Host, command, Session, or chat UI source? No. It uses only the plugin manifest, Host RPC, Workspace Registry, sidebar.footer.action , and shell.overlay . Where is it opened after installation? The A/B action at the bottom of the DSH sidebar. Arena no longer registers an /arena slash command. Does a new conversation automatically show preflight warnings? No. Setup and preflight are loaded only after the user opens Arena and selects that tab. Can reviewers inspect actual candidate code? Yes. Selecting a contender and file lazily renders the sealed line-numbered unified diff, not merely a changed-line count. Can it compare two models directly? Yes. Each contender may use an independent provider, model, credential …

Category