ci: make feedback backfills observable

.github/scripts/sweep-recent-review-feedback.rb

@@ -126,9 +126,12 @@ def search_recent_prs(owner:, since:, limit: 100)
     JSON.parse(stdout)
   end
 
-  def feedback_items(owner:, since:, min_severity:, pr_limit: 100)
-    search_recent_prs(owner: owner, since: since, limit: pr_limit).flat_map do |pr|
+  def feedback_items(owner:, since:, min_severity:, pr_limit: 100, progress: false)
+    prs = search_recent_prs(owner: owner, since: since, limit: pr_limit)
+    warn "review feedback sweep: inspecting #{prs.length} merged PRs since #{since}" if progress
+    prs.each_with_index.flat_map do |pr, index|
       repo = pr.fetch("repository").fetch("nameWithOwner")
+      warn "review feedback sweep: #{index + 1}/#{prs.length} #{repo}##{pr.fetch("number")}" if progress
       payload = EvalOpsReviewThreadGuard.fetch_payload(repo: repo, pr: pr.fetch("number"))
       EvalOpsReviewThreadGuard.blocking_feedback(payload, min_severity: min_severity).map do |item|
         item.merge(
@@ -925,6 +928,7 @@ def upsert_issue(repo:, title:, body:)
     guardrail_repo_issues: false,
     weekly_report_issue_repo: nil,
     weekly_report_issue_title: EvalOpsReviewFeedbackSweep::DEFAULT_WEEKLY_REPORT_TITLE,
+    progress: false,
     pr_limit: 100,
     dry_run: false
   }
@@ -944,6 +948,7 @@ def upsert_issue(repo:, title:, body:)
     parser.on("--guardrail-lifecycle-json-output PATH", "Write guardrail issue lifecycle JSON to this path") { |value| options[:guardrail_lifecycle_json_output] = value }
     parser.on("--weekly-report-issue-repo OWNER/REPO", "Create or comment on this issue repo with the guardrail report") { |value| options[:weekly_report_issue_repo] = value }
     parser.on("--weekly-report-issue-title TITLE", "Issue title for the guardrail report") { |value| options[:weekly_report_issue_title] = value }
+    parser.on("--progress", "Print PR inspection progress to stderr") { options[:progress] = true }
     parser.on("--dry-run", "Print report and skip issue writes") { options[:dry_run] = true }
   end.parse!
 
@@ -957,7 +962,8 @@ def upsert_issue(repo:, title:, body:)
     owner: options.fetch(:owner),
     since: since,
     min_severity: options.fetch(:min_severity),
-    pr_limit: options.fetch(:pr_limit)
+    pr_limit: options.fetch(:pr_limit),
+    progress: options.fetch(:progress)
   )
   ledger = EvalOpsReviewFeedbackSweep.ledger_json(
     items,

.github/workflows/review-feedback-backfill.yml

@@ -16,7 +16,7 @@ on:
       pr_limit:
         description: "Maximum merged PRs to inspect"
         required: false
-        default: "1000"
+        default: "250"
 
 permissions:
   contents: read
@@ -31,7 +31,7 @@ jobs:
       GH_TOKEN: ${{ secrets.EVALOPS_REVIEW_GUARD_TOKEN || secrets.EVALOPS_ORG_READ_TOKEN }}
       SINCE_HOURS: ${{ inputs.since_hours || '720' }}
       MIN_SEVERITY: ${{ inputs.min_severity || 'high' }}
-      PR_LIMIT: ${{ inputs.pr_limit || '1000' }}
+      PR_LIMIT: ${{ inputs.pr_limit || '250' }}
     steps:
       - uses: actions/checkout@v5
 
@@ -51,6 +51,7 @@ jobs:
             --since-hours "${SINCE_HOURS}" \
             --pr-limit "${PR_LIMIT}" \
             --min-severity "${MIN_SEVERITY}" \
+            --progress \
             --json-output review-feedback-30d-ledger.json \
             --guardrail-backlog-output review-feedback-30d-guardrail-backlog.md \
             --guardrail-backlog-json-output review-feedback-30d-guardrail-backlog.json \

test/sweep_recent_review_feedback_test.rb

@@ -86,6 +86,62 @@ def test_ledger_json_records_empty_sweeps
     assert_equal [], ledger.fetch("findings")
   end
 
+  def test_feedback_items_progress_reports_search_and_pr_numbers
+    prs = [
+      {
+        "repository" => {
+          "nameWithOwner" => "evalops/deploy"
+        },
+        "number" => 2390,
+        "title" => "fix parser",
+        "url" => "https://github.com/evalops/deploy/pull/2390",
+        "closedAt" => "2026-05-10T05:00:00Z"
+      }
+    ]
+    payload = {
+      "repository" => {
+        "pullRequest" => {
+          "reviewThreads" => {
+            "nodes" => []
+          },
+          "comments" => {
+            "nodes" => []
+          },
+          "reviews" => {
+            "nodes" => []
+          }
+        }
+      }
+    }
+    original_search = EvalOpsReviewFeedbackSweep.method(:search_recent_prs)
+    original_fetch = EvalOpsReviewThreadGuard.method(:fetch_payload)
+    EvalOpsReviewFeedbackSweep.define_singleton_method(:search_recent_prs) { |owner:, since:, limit:| prs }
+    EvalOpsReviewThreadGuard.define_singleton_method(:fetch_payload) { |repo:, pr:| payload }
+
+    _stdout, stderr = capture_io do
+      assert_equal(
+        [],
+        EvalOpsReviewFeedbackSweep.feedback_items(
+          owner: "evalops",
+          since: "2026-04-10",
+          min_severity: "high",
+          pr_limit: 10,
+          progress: true
+        )
+      )
+    end
+
+    assert_includes stderr, "review feedback sweep: inspecting 1 merged PRs since 2026-04-10"
+    assert_includes stderr, "review feedback sweep: 1/1 evalops/deploy#2390"
+  ensure
+    EvalOpsReviewFeedbackSweep.define_singleton_method(:search_recent_prs) do |owner:, since:, limit: 100|
+      original_search.call(owner: owner, since: since, limit: limit)
+    end
+    EvalOpsReviewThreadGuard.define_singleton_method(:fetch_payload) do |repo:, pr:|
+      original_fetch.call(repo: repo, pr: pr)
+    end
+  end
+
   def test_guardrail_backlog_ranks_recurring_feedback_classes
     ledger = {
       "schema_version" => "evalops.review_feedback_ledger.v1",