Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
48 changes: 45 additions & 3 deletions USER_GUIDE.md
Original file line number Diff line number Diff line change
Expand Up @@ -803,8 +803,9 @@ twice.
The pre-push hook picks a reviewer in this order:
1. `git config prism.reviewer <kiro|claude>`: a per-clone override for one developer. It is never
committed. With `--global --reviewer`, the same key is set in your global git config.
1. `git config prism.reviewer <kiro|claude>`: git reads it from the clone first (`--local`, an
override for this repo, never committed), then from your global git config (`--global`, your
default for every repo). Either one wins over the team file.
2. `eval.reviewer` in `.prism/config.json`: the team default, set by `enable codewatch`.
3. Whichever of kiro-cli and claude is on PATH.
Expand Down Expand Up @@ -857,7 +858,8 @@ The pre-push review picks the model for the tool that actually runs, in this ord
1. `eval.model` in `.prism/config.json`, when it was set for this tool (or `eval.reviewer` is
unset): the team's pin. `auto` or `default` here also stops the search, with the reviewer's
default.
2. `git config prism.<tool>Model`: your own default (usually `--global`) for repos that pin none.
2. `git config prism.<tool>Model`: this clone's value (`--local`), else your default (`--global`),
for repos that pin none.
3. The reviewer's default: `auto` for kiro-cli, Claude Code's own default for claude.
An invalid value at either level is ignored, with a note in the hook output. The hook prints the
Expand All @@ -870,6 +872,46 @@ then `auto`. CI always runs kiro-cli and reads `eval.model` whichever tool it wa
whose `eval.reviewer` is claude should set `PRISM_REVIEW_MODEL` for CI. It never reads
`prism.<tool>Model`. See [Kiro-CLI Eval Gate](#kiro-cli-eval-gate).
### Changing the Reviewer and Model
```bash
prism-cli codewatch reviewer # what a push here uses, and where each value comes from
prism-cli codewatch reviewer --reviewer claude --model sonnet # override for this repo (git config --local)
prism-cli codewatch reviewer --scope global --reviewer claude --model sonnet # your default for every repo (git config --global)
prism-cli codewatch reviewer --scope team --model auto # the team file (.prism/config.json eval); commit it
prism-cli codewatch reviewer --reset # this repo follows your default again
```
`--reset` removes the settings at its scope (`--scope global --reset` removes your default).
`--clear-overrides <dir...>` removes each listed repo's own `prism.reviewer` and `prism.<tool>Model`
so they follow your default; it never touches `.prism/config.json`. The same rules apply as above:
the reviewer comes from this repo's override, then your default, then the team file, then the
installed tool; the model comes from the team's `eval.model` (when it was set for the reviewer that
runs), then this repo's override, then your default, then the tool's default. kiro-cli's
`chat --list-models` answers from your account, so neither the CLI nor the dashboard offers it
offline; the dashboard suggests Claude Code's documented aliases.
In the dashboard, a repo's page has a **Reviewer and model** block in its Repo settings card. It
shows the reviewer and model a push there uses, where each comes from (**overridden for this
repo**, **your default**, **team file** or **tool default**), whether that tool is installed, and,
when the chosen tool is missing, the one that runs instead. **Change** offers **Override for this
repo** (the default choice), **Use my default** (removes the override) and **Team default**. The
team default warns that `.prism/config.json` is a repo file to commit, which the CI eval gate reads
too, and shows the diff before saving and again in the confirm dialog. Picking a personal model
while the team pins one for the same reviewer says that the pin wins and offers the team default.
**Which setting wins** spells out the order. The model field takes an id or `auto`, and a value
that breaks the model id rule is refused on the page and again on the server.
![The Reviewer and model block of the demo repo with the team default chosen: the repo-file warning and the .prism/config.json diff before saving](assets/images/prism-cli/dashboard-codewatch-reviewer-dark.png)
**Settings › Repos** starts with **Codewatch default reviewer and model**: your default
(`git config --global prism.reviewer` and `prism.<tool>Model`) with **Change my default** and
**Remove my default**, then the repos that override it, each with a link and **Reset to default**.
**Apply my default to all repos** removes every override, after a confirm dialog that lists the
repos it changes; team files are never touched. Each row in the repos table links to that repo's
page to change it. Every change runs `prism-cli codewatch reviewer` as a fixed command, with the
reviewer and model checked by the server.
### What the Hooks Do
**pre-commit** runs `gitleaks protect --staged` on the staged changes, using `.gitleaks.toml` and
Expand Down
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
4 changes: 2 additions & 2 deletions cli/package-lock.json

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion cli/package.json
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
{
"name": "@prism-d1/cli",
"version": "1.19.2",
"version": "1.20.0",
"description": "PRISM D1 Velocity CLI — AI development lifecycle assessment tool",
"author": "andklee",
"engines": {
Expand Down
Loading
Loading