Fix lint: shellcheck SC2002 and PYFMT

ACTIONLINT flagged two useless 'cat file | tee' pipes; use tee's stdin
redirection instead. PYFMT wanted three expressions wrapped -- applied via
black + usort. Also strip() the Buildkite token and report the HTTP status on
fetch failure, after the first CI run 401'd on every cluster with no
indication of why.

Author: atalman

.github/workflows/vllm-torch-nightly-triage.yml

@@ -80,8 +80,7 @@ jobs:
         if: steps.report.outputs.has_report == 'true'
         run: |
           set -euo pipefail
-          cat "${RUNNER_TEMP}/report.md"
-          cat "${RUNNER_TEMP}/report.md" >> "${GITHUB_STEP_SUMMARY}"
+          tee -a "${GITHUB_STEP_SUMMARY}" < "${RUNNER_TEMP}/report.md"
 
       - name: No comparable build
         if: steps.report.outputs.has_report != 'true'
@@ -173,7 +172,7 @@ jobs:
         run: |
           set -euo pipefail
           if [[ -s "${RUNNER_TEMP}/findings.md" ]]; then
-            cat "${RUNNER_TEMP}/findings.md" | tee -a "${GITHUB_STEP_SUMMARY}"
+            tee -a "${GITHUB_STEP_SUMMARY}" < "${RUNNER_TEMP}/findings.md"
           else
             echo "Agent produced no findings file." | tee -a "${GITHUB_STEP_SUMMARY}"
           fi

tools/torchci/vllm_torch_nightly_triage.py

@@ -34,7 +34,6 @@
 
 from torchci.clickhouse import get_clickhouse_client
 
-
 VLLM_REPO = "https://github.com/vllm-project/vllm.git"
 PIPELINE = "CI"
 
@@ -97,7 +96,9 @@ def find_latest_pair(
     )
 
     def as_dict(row: Tuple) -> Dict[str, Any]:
-        return dict(zip(("number", "title", "commit", "created_at", "state", "url"), row))
+        return dict(
+            zip(("number", "title", "commit", "created_at", "state", "url"), row)
+        )
 
     parsed = [as_dict(r) for r in builds]
     nightlies = [b for b in parsed if b["title"].startswith(TORCH_NIGHTLY_MSG)]
@@ -228,10 +229,14 @@ def render(
     for key, jobs in ordered:
         states = sorted({j["state"] for j in jobs})
         exits = sorted({str(j["exit_status"]) for j in jobs})
-        out.append(f"<details><summary><b>{key}</b> — {len(jobs)} job(s), "
-                   f"{'/'.join(states)}, exit {','.join(exits)}</summary>\n")
+        out.append(
+            f"<details><summary><b>{key}</b> — {len(jobs)} job(s), "
+            f"{'/'.join(states)}, exit {','.join(exits)}</summary>\n"
+        )
         for j in sorted(jobs, key=lambda x: x["name"]):
-            out.append(f"- [{j['name']}]({j['url']}) — `{j['state']}` exit `{j['exit_status']}`")
+            out.append(
+                f"- [{j['name']}]({j['url']}) — `{j['state']}` exit `{j['exit_status']}`"
+            )
         out.append("\n</details>")
 
     out.append("\n### Infrastructure check\n")
@@ -299,6 +304,16 @@ def fetch_cluster_logs(
         try:
             with urllib.request.urlopen(req, timeout=120) as resp:
                 body = resp.read().decode("utf-8", errors="replace")
+        except urllib.error.HTTPError as exc:
+            hint = ""
+            if exc.code == 401:
+                hint = (
+                    "  (401 => token invalid for this org. Check it has read_builds "
+                    "and read_build_logs, that the vllm organization is selected, and "
+                    "that the value has no trailing newline.)"
+                )
+            print(f"skip {key}: HTTP {exc.code} {exc.reason}{hint}", file=sys.stderr)
+            continue
         except urllib.error.URLError as exc:
             print(f"skip {key}: {exc}", file=sys.stderr)
             continue
@@ -366,7 +381,10 @@ def main() -> int:
     if args.logs_dir and buckets["regressed"]:
         import os as _os
 
-        token = _os.environ.get("BUILDKITE_TOKEN", "")
+        # .strip() matters: a trailing newline in the value (easy to introduce when
+        # pasting a token into a secret) makes the Authorization header invalid and
+        # every request 401s.
+        token = _os.environ.get("BUILDKITE_TOKEN", "").strip()
         if not token:
             # Not fatal: the report above is still useful without logs.
             print("BUILDKITE_TOKEN unset; skipping log fetch", file=sys.stderr)