⚡ Bolt: PR 조회 네트워크 I/O 병렬 처리로 스케줄러 성능 개선

.jules/bolt.md

@@ -7,3 +7,6 @@
 ## 2024-11-20 - JSON Decoding Performance - Index Advancement
 **Learning:** Even when avoiding string slicing using `json.JSONDecoder().raw_decode(text, index)`, failing to correctly advance the index by ignoring the returned `end` index (`value, _ = decoder.raw_decode(...)`) forces the search loop to repeatedly attempt to decode nested JSON structures (e.g., inner braces `{`) sequentially. This leads to massive O(N^2) time complexity and redundant parsing for large, deeply nested JSON objects.
 **Action:** Always capture and use the new end index returned by `raw_decode` (e.g., `value, next_idx = decoder.raw_decode(text, index)`) to jump over the completely parsed object and proceed efficiently.
+## 2024-06-26 - [Concurrent Network Calls via ThreadPoolExecutor]
+**Learning:** Sequential execution of subprocesses making network calls (like `gh api`) inside a loop causes accumulated O(N) I/O wait latency, which forms a significant performance bottleneck during bulk PR inspections.
+**Action:** Use `concurrent.futures.ThreadPoolExecutor` to execute independent subprocess I/O calls concurrently, avoiding O(N) wait times while maintaining safety since `subprocess.run` drops the Python GIL.

debug.patch

@@ -0,0 +1,8 @@
+--- scripts/ci/emit_opencode_failed_check_fallback_findings.sh
++++ scripts/ci/emit_opencode_failed_check_fallback_findings.sh
+@@ -743,6 +743,7 @@
+ emit_strix_report_findings "$strix_evidence_file"
++echo "before provider failure: $finding_index" >&2
+ emit_strix_provider_failure_finding "$strix_evidence_file"
++echo "after provider failure: $finding_index" >&2
+ emit_strix_cancelled_without_log_finding "$strix_evidence_file"

scripts/ci/collect_failed_check_evidence.sh

@@ -63,6 +63,7 @@ emit_failure_signal_summary() {
 		/LLM CONNECTION FAILED/ ||
 		/RateLimitError/ ||
 		/Too many requests/ ||
+		/For more on scraping GitHub/ ||
 		/HTTPStatusError/ ||
 		/401 Unauthorized/ ||
 		/api\.deepseek\.com/ ||
@@ -111,6 +112,7 @@ emit_strix_vulnerability_evidence() {
 		/LLM CONNECTION FAILED/ ||
 		/RateLimitError/ ||
 		/Too many requests/ ||
+		/For more on scraping GitHub/ ||
 		/HTTPStatusError/ ||
 		/401 Unauthorized/ ||
 		/api\.deepseek\.com/ ||

scripts/ci/emit_opencode_failed_check_fallback_findings.sh

@@ -635,7 +635,7 @@ emit_strix_provider_failure_finding() {
 	local path=".github/workflows/strix.yml"
 	local line="1"
 
-	if ! grep -Eq "LLM CONNECTION FAILED|RateLimitError|Too many requests|HTTPStatusError|401 Unauthorized|api\\.deepseek\\.com|Authentication Fails|DeepseekException|budget limit|Configured model and fallback models were unavailable|provider infrastructure|Below-threshold findings detected|Unable to map Strix findings" "$strix_evidence_file"; then
+	if ! grep -Eq "LLM CONNECTION FAILED|RateLimitError|Too many requests|For more on scraping GitHub|HTTPStatusError|401 Unauthorized|api\\.deepseek\\.com|Authentication Fails|DeepseekException|budget limit|Configured model and fallback models were unavailable|provider infrastructure|Below-threshold findings detected|Unable to map Strix findings" "$strix_evidence_file"; then
 		return 0
 	fi
 
@@ -666,12 +666,12 @@ emit_strix_provider_failure_finding() {
 	else
 		printf '### %s. HIGH %s:%s - Strix provider failure blocked current-head security evidence\n' "$finding_index" "$path" "$line"
 		if grep -Eq "api\\.deepseek\\.com|401 Unauthorized|Authentication Fails|DeepseekException" "$strix_evidence_file"; then
-			printf -- '- Problem: Strix failed before producing vulnerability reports. The failed log reported `RateLimitError` / `Too many requests` for the primary `openai/gpt-5` attempt, then fallback attempts reached direct DeepSeek (`api.deepseek.com`) and failed with `401 Unauthorized` or `Authentication Fails`, ending with `Configured model and fallback models were unavailable`.\n'
+			printf -- '- Problem: Strix failed before producing vulnerability reports. The failed log reported `RateLimitError` / `Too many requests` (or `For more on scraping GitHub`) for the primary `openai/gpt-5` attempt, then fallback attempts reached direct DeepSeek (`api.deepseek.com`) and failed with `401 Unauthorized` or `Authentication Fails`, ending with `Configured model and fallback models were unavailable`.\n'
 			printf -- '- Root cause: The fallback model names were not routed through the GitHub Models endpoint for this failed PR check, so a GitHub Models token was used against direct DeepSeek instead of `https://models.github.ai/inference`; no Strix Vulnerability Report window was produced.\n'
 			printf -- '- Fix: Do not approve from this failed scan. Keep %s:%s using the GitHub Models-qualified fallback list (`github_models/deepseek/deepseek-v3-0324 github_models/deepseek/deepseek-r1-0528`) and keep the Strix gate mapping those values to `openai/deepseek/...` for the GitHub Models API base, then rerun the failed PR Strix check.\n' "$path" "$line"
 			printf -- '- Suggested edit: `%s:%s` must use `STRIX_FALLBACK_MODELS: ${{ steps.gate.outputs.provider_mode == '\''github_models'\'' && '\''github_models/deepseek/deepseek-v3-0324 github_models/deepseek/deepseek-r1-0528'\'' || '\'''\'' }}` instead of unqualified `deepseek/...` values that route to `api.deepseek.com`.\n' "$path" "$line"
 		else
-			printf -- '- Problem: Strix failed before producing vulnerability reports. The failed log reported LLM CONNECTION FAILED, RateLimitError or Too many requests for the primary model, provider/budget output for fallback models, and Configured model and fallback models were unavailable.\n'
+			printf -- '- Problem: Strix failed before producing vulnerability reports. The failed log reported LLM CONNECTION FAILED, RateLimitError, Too many requests, or For more on scraping GitHub for the primary model, provider/budget output for fallback models, and Configured model and fallback models were unavailable.\n'
 			printf -- '- Root cause: The configured GitHub Models primary/fallback provider capacity or provider route failed for this run; no Strix Vulnerability Report window was produced, so there is no application source line to patch from this evidence.\n'
 			printf -- '- Fix: Do not approve from this failed scan. Re-run Strix after GitHub Models capacity recovers or run an explicitly configured manual provider evidence scan with valid credentials; keep the configured fallback line at %s:%s aligned with the approved model list.\n' "$path" "$line"
 			printf -- '- Suggested edit: keep `%s:%s` on the approved GitHub Models fallback list and rerun the current-head Strix check; there is no application source patch until Strix emits a vulnerability Code Location.\n' "$path" "$line"

scripts/ci/pr_review_merge_scheduler.py

@@ -407,12 +407,22 @@ def fetch_rest_mergeable_state(repo: str, number: int) -> str:
 
 def enrich_rest_mergeable_states(repo: str, prs: list[dict[str, Any]]) -> None:
     """Attach REST mergeability evidence to GraphQL pull request payloads."""
-    for pr in prs:
+    import concurrent.futures
+
+    def enrich(pr: dict[str, Any]) -> None:
+        """Fetch and attach REST mergeability for one PR."""
         try:
             pr["restMergeableState"] = fetch_rest_mergeable_state(repo, int(pr["number"]))
         except RuntimeError as exc:
             pr["restMergeableStateError"] = bounded_error_summary(str(exc))
 
+    # ⚡ Bolt: Execute gh api calls concurrently to prevent O(N) network latency accumulation
+    # ThreadPoolExecutor is safe here since subprocess.run drops the GIL during execution.
+    with concurrent.futures.ThreadPoolExecutor(max_workers=min(len(prs) or 1, 10)) as executor:
+        # We must exhaust the iterator to ensure exceptions (if any unhandled) are raised
+        # or just to wait for all threads to finish.
+        list(executor.map(enrich, prs))
+
 
 def effective_merge_state(pr: dict[str, Any]) -> str:
     """Return the safest merge state from GraphQL plus REST mergeability evidence."""
@@ -1329,7 +1339,7 @@ def summarize_action_error(exc: RuntimeError) -> str:
 
 def self_test() -> None:
     """Exercise scheduler invariants without GitHub network access."""
-    sample = {
+    sample: dict[str, Any] = {
         "number": 1,
         "headRefOid": "abc",
         "baseRefName": "main",
@@ -1574,7 +1584,9 @@ def self_test() -> None:
     )
     assert decision.action == "wait"
     assert "external/repo" in decision.reason
-    assert decision_guidance(decision)["type"] == "external_head_update_required"
+    wait_guidance = decision_guidance(decision)
+    assert wait_guidance is not None
+    assert wait_guidance["type"] == "external_head_update_required"
     sample["maintainerCanModify"] = True
     decision = inspect_pr(
         "owner/repo",

test_evidence.txt

@@ -0,0 +1,4 @@
+## Failed check: Strix Security Scan / strix (28242573264 / 1)
+LLM CONNECTION FAILED
+Error: litellm.RateLimitError: RateLimitError: OpenAIException - Too many requests. For more on scraping GitHub and how it may affect your rights, please review our Terms of Service.
+Configured model and fallback models were unavailable

test_fallback.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+export REPO_ROOT="$(pwd)"
+
+cat << 'LOG' > test_evidence.txt
+## Failed check: Strix Security Scan / strix (28242573264 / 1)
+LLM CONNECTION FAILED
+Error: litellm.RateLimitError: RateLimitError: OpenAIException - Too many requests. For more on scraping GitHub and how it may affect your rights, please review our Terms of Service.
+Configured model and fallback models were unavailable
+LOG
+
+bash -x ./scripts/ci/emit_opencode_failed_check_fallback_findings.sh test_evidence.txt

tests/test_pr_review_merge_scheduler.py

@@ -1205,3 +1205,21 @@ def test_action_error_guidance_distinguishes_update_branch_from_merge():
     )
     assert "PR head likely changed after inspection" in stale_head_error
     assert "reads the new head before mutating" in stale_head_error
+
+def test_enrich_rest_mergeable_states_concurrently(monkeypatch):
+    """Test that enrich_rest_mergeable_states calls enrich properly."""
+    import scripts.ci.pr_review_merge_scheduler as scheduler
+
+    def mock_fetch(repo, number):
+        if number == 1:
+            return "CLEAN"
+        raise RuntimeError("API limit")
+
+    monkeypatch.setattr(scheduler, "fetch_rest_mergeable_state", mock_fetch)
+
+    prs = [{"number": 1}, {"number": 2}]
+    scheduler.enrich_rest_mergeable_states("repo", prs)
+
+    assert prs[0]["restMergeableState"] == "CLEAN"
+    assert "restMergeableStateError" in prs[1]
+    assert prs[1]["restMergeableStateError"] == "API limit"