From 319e9de958784e979df2b98c0d171b656ca1bb0e Mon Sep 17 00:00:00 2001
From: "Sharad." <sharadvc@users.noreply.github.com>
Date: Sat, 12 Sep 2026 14:39:37 +0000
Subject: [PATCH] docs(cli): document --concurrency, --junit, and --degrade in
 help

Fixes #4

Co-authored-by: Sharad. <sharadvc@users.noreply.github.com>
---
 src/cli/help.ts        | 32 ++++++++++++++++++++++++++++++++
 src/cli/index.ts       | 35 +++--------------------------------
 tests/cli-help.test.ts | 10 ++++++++++
 3 files changed, 45 insertions(+), 32 deletions(-)
 create mode 100644 src/cli/help.ts
 create mode 100644 tests/cli-help.test.ts

diff --git a/src/cli/help.ts b/src/cli/help.ts
new file mode 100644
index 0000000..d8e6a61
--- /dev/null
+++ b/src/cli/help.ts
@@ -0,0 +1,32 @@
+export const CLI_HELP = `evalgate - the build fails when your prompt gets dumber.
+
+Usage:
+  evalgate run <suite>            Run a suite and print a report.
+  evalgate baseline <suite>       Run a suite and save it as a baseline.
+  evalgate compare <suite>        Run a suite and compare it to a baseline.
+  evalgate compare               Compare two existing result files.
+  evalgate init [file]            Write a starter suite you can edit.
+  evalgate list                   List the available scorers and providers.
+
+Common flags:
+  --provider <name>     Override the provider (default: mock).
+  --model <name>        Override the model.
+  --tags <a,b>          Only run cases with one of these tags.
+  --concurrency <n>     Run cases in parallel (default: 1).
+  --out <file>          Write the JSON result artifact here.
+  --md <file>           Write a Markdown report here.
+  --junit <file>        Write a JUnit XML report here.
+  --json                Print the JSON result to stdout.
+  --no-fail             Do not exit non-zero on failure/regression.
+  --degrade             Use degraded mock outputs (for regression demos).
+
+compare flags:
+  --base <file>         Baseline result JSON (required for compare).
+  --head <file>         Candidate result JSON (skips running the suite).
+  --tolerance <n>       Allowed score drop before it counts as a regression.
+  --comment             Upsert a PR comment via the GitHub API (needs token).
+
+Other:
+  --version, -v         Print version.
+  --help, -h            Print this help.
+`;
diff --git a/src/cli/index.ts b/src/cli/index.ts
index 5e679ac..75dc060 100644
--- a/src/cli/index.ts
+++ b/src/cli/index.ts
@@ -15,6 +15,7 @@ import { defaultScorerRegistry } from "../scorers/registry.js";
 import { MockProvider } from "../providers/mock.js";
 import { contextFromEnv, upsertComment } from "../github.js";
 import type { RunResult } from "../types.js";
+import { CLI_HELP } from "./help.js";
 
 const require = createRequire(import.meta.url);
 
@@ -26,36 +27,6 @@ function version(): string {
   }
 }
 
-const HELP = `evalgate - the build fails when your prompt gets dumber.
-
-Usage:
-  evalgate run <suite>            Run a suite and print a report.
-  evalgate baseline <suite>       Run a suite and save it as a baseline.
-  evalgate compare <suite>        Run a suite and compare it to a baseline.
-  evalgate compare               Compare two existing result files.
-  evalgate init [file]            Write a starter suite you can edit.
-  evalgate list                   List the available scorers and providers.
-
-Common flags:
-  --provider <name>     Override the provider (default: mock).
-  --model <name>        Override the model.
-  --tags <a,b>          Only run cases with one of these tags.
-  --out <file>          Write the JSON result artifact here.
-  --md <file>           Write a Markdown report here.
-  --json                Print the JSON result to stdout.
-  --no-fail             Do not exit non-zero on failure/regression.
-
-compare flags:
-  --base <file>         Baseline result JSON (required for compare).
-  --head <file>         Candidate result JSON (skips running the suite).
-  --tolerance <n>       Allowed score drop before it counts as a regression.
-  --comment             Upsert a PR comment via the GitHub API (needs token).
-
-Other:
-  --version, -v         Print version.
-  --help, -h            Print this help.
-`;
-
 /** Build a provider registry, allowing a --degrade toggle for the mock. */
 function registryFor(args: ParsedArgs) {
   const degrade = boolFlag(args, "degrade");
@@ -215,7 +186,7 @@ async function main(): Promise<number> {
   }
   const command = args._.shift();
   if (!command || boolFlag(args, ["help", "h"]) || command === "help") {
-    console.log(HELP);
+    console.log(CLI_HELP);
     return 0;
   }
 
@@ -232,7 +203,7 @@ async function main(): Promise<number> {
       return cmdInit(args);
     default:
       console.error(`[evalgate] unknown command "${command}"\n`);
-      console.log(HELP);
+      console.log(CLI_HELP);
       return 2;
   }
 }
diff --git a/tests/cli-help.test.ts b/tests/cli-help.test.ts
new file mode 100644
index 0000000..40eb074
--- /dev/null
+++ b/tests/cli-help.test.ts
@@ -0,0 +1,10 @@
+import { describe, it, expect } from "vitest";
+import { CLI_HELP } from "../src/cli/help.js";
+
+describe("CLI help text", () => {
+  it("documents --concurrency, --junit, and --degrade", () => {
+    expect(CLI_HELP).toContain("--concurrency");
+    expect(CLI_HELP).toContain("--junit");
+    expect(CLI_HELP).toContain("--degrade");
+  });
+});