diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml
index bbe4b76775d..1938aab5007 100644
--- a/.github/ISSUE_TEMPLATE/bug_report.yml
+++ b/.github/ISSUE_TEMPLATE/bug_report.yml
@@ -6,8 +6,11 @@ body:
- type: markdown
attributes:
value: |
- Thanks for taking the time to fill out this bug report!
-
+ Thanks for taking the time to file a bug report!
+
+ > ⚠️ **Auto-triage notice for external contributors:**
+ > Bug reports without **clear reproduction steps, expected vs. actual behavior, and a screenshot or terminal/log output** are auto-closed by our LLM triage bot with an explanation of what was missing. You can fill in the missing details and reopen at any time — the bot will re-evaluate. Internal BerriAI contributors are exempt.
+
**💡 Tip:** See our [Troubleshooting Guide](https://docs.litellm.ai/docs/troubleshoot) for what information to include.
- type: checkboxes
id: duplicate-check
@@ -20,21 +23,33 @@ body:
- type: textarea
id: what-happened
attributes:
- label: What happened?
- description: Also tell us, what did you expect to happen?
- placeholder: Tell us what you see!
- value: "A bug happened!"
+ label: What happened? (Actual behavior)
+ description: A clear description of what is happening today, with the bug.
+ placeholder: e.g. "Calling completion() with model=gpt-4o-mini returns an empty string."
+ validations:
+ required: true
+ - type: textarea
+ id: expected-behavior
+ attributes:
+ label: What did you expect to happen? (Expected behavior)
+ description: A clear description of what you expected to happen. **Required.**
+ placeholder: e.g. "I expected completion() to return the model's response text."
validations:
required: true
- type: textarea
id: steps-to-reproduce
attributes:
- label: Steps to Reproduce
- description: Please provide detailed steps to reproduce this bug(A curl/python code to reproduce the bug)
+ label: Steps to reproduce
+ description: |
+ Provide a minimal reproduction. Include a runnable Python snippet or a
+ `curl` command, your config.yaml if relevant, and the exact LiteLLM
+ version + Python version. Reports without a runnable reproduction are
+ auto-closed.
placeholder: |
- 1. config.yaml file/ .env file/ etc.
- 2. Run the following code...
- 3. Observe the error...
+ 1. Create `config.yaml` with: ...
+ 2. Start the proxy with: `litellm --config config.yaml --port 4000`
+ 3. Run this Python / curl: ...
+ 4. Observe: ...
value: |
1.
2.
@@ -44,9 +59,15 @@ body:
- type: textarea
id: logs
attributes:
- label: Relevant log output
- description: Please copy and paste any relevant log output. This will be automatically formatted into code, so no need for backticks.
+ label: Relevant log output / screenshot
+ description: |
+ **Required.** Paste the full traceback, stderr, proxy logs, or attach a
+ screenshot showing the bug. For UI bugs a screenshot or screen
+ recording is mandatory. Without proof of the bug, the issue is
+ auto-closed.
render: shell
+ validations:
+ required: true
- type: dropdown
id: component
attributes:
@@ -63,14 +84,14 @@ body:
- type: input
id: version
attributes:
- label: What LiteLLM version are you on ?
+ label: What LiteLLM version are you on ?
placeholder: v1.53.1
validations:
required: true
- type: input
id: contact
attributes:
- label: Twitter / LinkedIn details
+ label: Twitter / LinkedIn details
description: We announce new features on Twitter + LinkedIn. If this issue leads to an announcement, and you'd like a mention, we'll gladly shout you out!
placeholder: ex. @krrish_dh / https://www.linkedin.com/in/krish-d/
validations:
diff --git a/.github/ISSUE_TEMPLATE/feature_request.yml b/.github/ISSUE_TEMPLATE/feature_request.yml
index 4cc42901897..9844032a98b 100644
--- a/.github/ISSUE_TEMPLATE/feature_request.yml
+++ b/.github/ISSUE_TEMPLATE/feature_request.yml
@@ -1,4 +1,4 @@
-name: 🚀 Feature Request
+name: 🚀 Feature Request
description: Submit a proposal/request for a new LiteLLM feature.
title: "[Feature]: "
labels: ["enhancement"]
@@ -6,7 +6,10 @@ body:
- type: markdown
attributes:
value: |
- Thanks for making LiteLLM better!
+ Thanks for making LiteLLM better!
+
+ > ⚠️ **Auto-triage notice for external contributors:**
+ > Feature requests need (1) a clear description of the proposed feature, (2) the motivation / use case with a concrete example, and (3) what success looks like. Vague requests are auto-closed by our LLM triage bot with an explanation. Fill in the missing details and reopen at any time — the bot will re-evaluate. Internal BerriAI contributors are exempt.
- type: checkboxes
id: duplicate-check
attributes:
@@ -18,16 +21,25 @@ body:
- type: textarea
id: the-feature
attributes:
- label: The Feature
- description: A clear and concise description of the feature proposal
- placeholder: Tell us what you want!
+ label: The feature
+ description: A clear and concise description of the feature proposal. What should LiteLLM do that it doesn't today?
+ placeholder: e.g. "Support per-team max_input_tokens overrides on the proxy."
validations:
required: true
- type: textarea
id: motivation
attributes:
- label: Motivation, pitch
- description: Please outline the motivation for the proposal. Is your feature request related to a specific problem? e.g., "I'm working on X and would like Y to be possible". If this is related to another GitHub issue, please link here too.
+ label: Motivation, pitch, and concrete example
+ description: |
+ **Required.** Why is this needed? Include a concrete use case — what
+ you're trying to accomplish, what's blocked today, and what success
+ would look like (ideally with an example config / API call / UI flow).
+ If this is related to another GitHub issue, link it here too.
+ placeholder: |
+ I'm running a multi-tenant proxy where team A processes long docs and
+ team B only does short chats. Today I have to spin up two proxies.
+ With this feature I could set max_input_tokens per team and route in one
+ proxy. Example config: ...
validations:
required: true
- type: dropdown
@@ -56,7 +68,7 @@ body:
- type: input
id: contact
attributes:
- label: Twitter / LinkedIn details
+ label: Twitter / LinkedIn details
description: We announce new features on Twitter + LinkedIn. When this is announced, and you'd like a mention, we'll gladly shout you out!
placeholder: ex. @krrish_dh / https://www.linkedin.com/in/krish-d/
validations:
diff --git a/.github/pull_request_template.md b/.github/pull_request_template.md
index f9ce9e5dcb8..a9c41d4b231 100644
--- a/.github/pull_request_template.md
+++ b/.github/pull_request_template.md
@@ -1,10 +1,66 @@
+
+
## Relevant issues
-
+
## Linear ticket
-
+
+
+## Problem description
+
+
+
+## Expected vs. actual behavior
+
+
+
+## QA proof
+
+
## Pre-Submission checklist
@@ -13,7 +69,7 @@
- [ ] I have Added testing in the [`tests/test_litellm/`](https://github.com/BerriAI/litellm/tree/main/tests/test_litellm) directory, **Adding at least 1 test is a hard requirement** - [see details](https://docs.litellm.ai/docs/extras/contributing_code)
- [ ] My PR passes all unit tests on [`make test-unit`](https://docs.litellm.ai/docs/extras/contributing_code)
- [ ] My PR's scope is as isolated as possible, it only solves 1 specific problem
-- [ ] I have requested a Greptile review by commenting `@greptileai` and received a **Confidence Score of at least 4/5** before requesting a maintainer review
+- [ ] I have received a Greptile **Confidence Score of at least 4/5** before requesting a maintainer review (Greptile reviews automatically on open; comment `@greptileai` to re-trigger after pushing fixes)
## Delays in PR merge?
@@ -36,13 +92,6 @@ If you're seeing a delay in your PR being merged, ping the LiteLLM Team on [Slac
- [ ] **Merge / cherry-pick CI run**
Links:
-## Screenshots / Proof of Fix
-
-
-
## Type
diff --git a/.github/scripts/close_low_quality_prs.py b/.github/scripts/close_low_quality_prs.py
new file mode 100644
index 00000000000..746807f4a58
--- /dev/null
+++ b/.github/scripts/close_low_quality_prs.py
@@ -0,0 +1,645 @@
+#!/usr/bin/env python3
+"""
+Auto-close low-quality pull requests.
+
+Closes open PRs (including drafts, regardless of age) that satisfy ALL of:
+ 1. Have a Greptile (`greptile-apps`) review comment whose latest
+ "Confidence Score: X/5" is below the configured threshold (default: 4).
+ 2. Are authored by an external OSS contributor (internal BerriAI
+ contributors are exempt).
+ 3. Do not carry an opt-out label (default: "do not close").
+
+`--min-age-days` is retained as an opt-in safety net for one-off backfill
+runs (default: 0). The team's intent is that the count of open PRs equals
+the count of PRs internal collaborators need to action on, so neither age
+nor draft status acts as a free pass.
+
+For each match, the script posts an explanatory comment and closes the PR.
+Because OSS contributors *cannot* reopen a PR closed by the bot/maintainer
+(GitHub limitation), the close-comment instructs them to push their fixes
+and **open a fresh PR**, or to comment `@agent-shin reconsider` on the
+closed PR to have the LLM judge re-evaluate (and reopen on pass).
+
+Requires the `gh` CLI to be authenticated.
+
+Usage examples:
+ # Dry run (default) - prints what would be closed
+ python3 close_low_quality_prs.py
+
+ # Actually close matching PRs
+ python3 close_low_quality_prs.py --close
+
+ # Restrict to PRs at least N days old (one-off backfill safety net)
+ python3 close_low_quality_prs.py --min-age-days 7 --min-score 4 --close
+"""
+
+from __future__ import annotations
+
+import argparse
+import datetime as dt
+import json
+import os
+import re
+import subprocess
+import sys
+from typing import Iterable
+
+# Greptile's GitHub App appears as `greptile-apps[bot]` in REST API comments
+# and `greptile-apps` in `gh pr view --json` output. Accept either form.
+GREPTILE_BOT_LOGINS = frozenset({"greptile-apps", "greptile-apps[bot]"})
+
+# Matches lines like:
+#
Confidence Score: 3/5
+# **Confidence Score: 4/5**
+# Confidence Score: 5 / 5
+SCORE_PATTERN = re.compile(
+ r"confidence\s*score\s*[:\-]?\s*(\d+)\s*/\s*5",
+ re.IGNORECASE,
+)
+
+# `author_association` values for internal BerriAI contributors who should be
+# exempt from auto-triage.
+INTERNAL_AUTHOR_ASSOCIATIONS = frozenset({"OWNER", "MEMBER", "COLLABORATOR"})
+
+# Default labels that exempt a PR from auto-close. Defined at module scope (not
+# as a mutable argparse default) so that `--optout-label foo` REPLACES the
+# defaults instead of appending to them — the argparse `action="append"` +
+# `default=[...]` combination silently mutates the shared default list.
+DEFAULT_OPTOUT_LABELS = ("do not close", "keep open", "wip")
+
+# HTML marker appended to grace-period warning comments. Shared with the
+# Agent Shin LLM-judge script (`triage_with_llm.py`) so a warning posted
+# by either path is recognized by both: the LLM judge can see "Greptile
+# already warned this contributor 12 hours ago" and skip re-warning, and
+# the Greptile closer can see "Agent Shin already warned" and close on
+# the next run if Greptile still has a low score.
+GRACE_COMMENT_MARKER = ""
+
+# Length of the grace period between the warning comment and the actual
+# auto-close. Set to 24 hours so the contributor has at least one full
+# working day across any time zone to push fixes or comment
+# `@agent-shin reconsider`. Mirrors the constant of the same name in
+# `triage_with_llm.py` — keep them in sync if either changes.
+GRACE_PERIOD_SECONDS = 86400
+
+# Default login of the GitHub identity that performs Agent Shin's writes;
+# used for matching the author of a grace-warning comment so we don't
+# count somebody quoting the marker. The env override
+# `AGENT_SHIN_BOT_LOGIN` mirrors `triage_with_llm.py`.
+AGENT_SHIN_DEFAULT_BOT_LOGIN = "github-actions[bot]"
+
+# Logins (case-insensitive) that bypass BOTH the 1-day grace period AND
+# the dry-run gating. Mirrors `IMMEDIATE_CLOSE_LOGINS` in
+# `triage_with_llm.py`. Used for dogfooding the bot from external test
+# accounts that have no push permissions to the repo.
+IMMEDIATE_CLOSE_LOGINS = frozenset({"swiftwinds"})
+
+
+def gh(*args: str) -> str:
+ """Run a `gh` CLI command and return stdout. Raises on non-zero exit."""
+ result = subprocess.run(
+ ["gh", *args],
+ capture_output=True,
+ text=True,
+ check=True,
+ )
+ return result.stdout
+
+
+def fetch_open_prs(repo: str | None) -> list[dict]:
+ """Fetch all open PRs (number, createdAt, isDraft, labels, author).
+
+ Includes drafts: `gh pr list --state open` returns both ready-for-review
+ and draft PRs by default. This is the desired behavior — drafts are not
+ a free pass; the internal-collaborator open-PR queue should reflect every
+ PR that needs human attention regardless of draft status.
+ """
+ repo_args = ["--repo", repo] if repo else []
+ fields = "number,title,createdAt,isDraft,labels,author,url"
+ raw = gh(
+ "pr",
+ "list",
+ "--state",
+ "open",
+ "--limit",
+ "1000",
+ "--json",
+ fields,
+ *repo_args,
+ )
+ return json.loads(raw)
+
+
+def fetch_pr_author_association(pr_number: int, repo: str | None) -> str:
+ """Return the GitHub `author_association` for a PR, uppercase.
+
+ Values: OWNER, MEMBER, COLLABORATOR, CONTRIBUTOR, FIRST_TIME_CONTRIBUTOR,
+ FIRST_TIMER, MANNEQUIN, NONE. Returns "" on lookup failure.
+ """
+ endpoint = (
+ f"repos/{repo}/pulls/{pr_number}"
+ if repo
+ else f"repos/{{owner}}/{{repo}}/pulls/{pr_number}"
+ )
+ try:
+ data = json.loads(gh("api", endpoint))
+ except subprocess.CalledProcessError:
+ return ""
+ return (data.get("author_association") or "").upper()
+
+
+def is_external_pr_author(pr: dict, repo: str | None) -> bool:
+ """Return True if the PR author is an external OSS contributor.
+
+ Internal = `OWNER` / `MEMBER` / `COLLABORATOR` association, or a bot login.
+ """
+ login = ((pr.get("author") or {}).get("login") or "").lower()
+ if login.endswith("[bot]") or login in {"dependabot", "github-actions"}:
+ return False
+ association = fetch_pr_author_association(pr["number"], repo)
+ # Fail-safe: if the API lookup failed (empty string), treat the author as
+ # internal so we don't auto-close their PR. Auto-close is destructive, so
+ # an unknown association should never make a PR eligible for closing.
+ if not association or association in INTERNAL_AUTHOR_ASSOCIATIONS:
+ return False
+ return True
+
+
+def fetch_pr_comments(pr_number: int, repo: str | None) -> list[dict]:
+ """Fetch issue-level comments on a PR (where Greptile posts its summary)."""
+ endpoint = (
+ f"repos/{repo}/issues/{pr_number}/comments?per_page=100"
+ if repo
+ else f"repos/{{owner}}/{{repo}}/issues/{pr_number}/comments?per_page=100"
+ )
+ raw = gh("api", "--paginate", endpoint)
+ comments: list[dict] = []
+ for line in raw.strip().splitlines():
+ line = line.strip()
+ if not line:
+ continue
+ parsed = json.loads(line)
+ if isinstance(parsed, list):
+ comments.extend(parsed)
+ else:
+ comments.append(parsed)
+ return comments
+
+
+def extract_greptile_score(comments: Iterable[dict]) -> tuple[int, dict] | None:
+ """Return (score, comment) for the most recent Greptile-authored comment
+ that contains a "Confidence Score: X/5". Returns None if no such comment.
+
+ "Most recent" is determined by the comment's `updated_at` (falling back to
+ `created_at`), so re-reviews override earlier passes.
+ """
+ candidates: list[tuple[str, int, dict]] = []
+ for comment in comments:
+ user = (comment.get("user") or {}).get("login", "")
+ if user not in GREPTILE_BOT_LOGINS:
+ continue
+ body = comment.get("body") or ""
+ match = SCORE_PATTERN.search(body)
+ if not match:
+ continue
+ score = int(match.group(1))
+ timestamp = comment.get("updated_at") or comment.get("created_at") or ""
+ candidates.append((timestamp, score, comment))
+
+ if not candidates:
+ return None
+
+ candidates.sort(key=lambda triple: triple[0])
+ _, score, comment = candidates[-1]
+ return score, comment
+
+
+def parse_iso8601(value: str) -> dt.datetime:
+ """Parse a GitHub ISO-8601 timestamp into a timezone-aware datetime."""
+ return dt.datetime.fromisoformat(value.replace("Z", "+00:00"))
+
+
+def has_optout_label(pr: dict, optout_labels: set[str]) -> bool:
+ labels = {label.get("name", "").lower() for label in pr.get("labels", [])}
+ return bool(labels & {lbl.lower() for lbl in optout_labels})
+
+
+def seconds_since_last_grace_warning(
+ comments: Iterable[dict],
+ *,
+ bot_login: str | None = None,
+ now: dt.datetime | None = None,
+) -> float | None:
+ """Return seconds since the bot's most recent grace-period warning, or
+ None if no such warning has ever been posted on this PR.
+
+ Detects warnings by matching `GRACE_COMMENT_MARKER` in comments
+ authored by the bot identity. Operates on an already-fetched
+ comments list (avoids a second `gh api` call when the caller has
+ already pulled the page for Greptile-score extraction).
+
+ `now` is injectable so callers (and tests) can pin the reference
+ time. The closer runs everything against a single `now` snapshot
+ captured at the top of `main()` so age calculations stay consistent
+ across many PRs in a single run.
+ """
+ expected_login = (
+ bot_login
+ or os.environ.get("AGENT_SHIN_BOT_LOGIN")
+ or AGENT_SHIN_DEFAULT_BOT_LOGIN
+ ).lower()
+ latest: dt.datetime | None = None
+ for comment in comments:
+ author = ((comment.get("user") or {}).get("login") or "").lower()
+ if author != expected_login:
+ continue
+ body = comment.get("body") or ""
+ if GRACE_COMMENT_MARKER not in body:
+ continue
+ created = comment.get("created_at")
+ if not created:
+ continue
+ try:
+ ts = parse_iso8601(created)
+ except ValueError:
+ continue
+ if latest is None or ts > latest:
+ latest = ts
+ if latest is None:
+ return None
+ reference = now if now is not None else dt.datetime.now(dt.timezone.utc)
+ return (reference - latest).total_seconds()
+
+
+def format_grace_warning_comment(score: int, threshold: int) -> str:
+ """Comment posted on the FIRST low-Greptile-score detection — gives
+ the contributor a 1-day grace window before the auto-close fires on
+ the next daily cron run.
+
+ Mirrors `format_grace_warning_pr_comment` in
+ `triage_with_llm.py` in spirit (1-day grace + escape hatches), but
+ framed around Greptile's confidence score instead of the LLM judge's
+ rubric since the close trigger here is the Greptile signal.
+ """
+ return (
+ "👋 Hi, thanks for the PR! I'm **Agent Shin**, the automated triage bot for this repository.\n"
+ "\n"
+ "Heads up — Greptile's most recent review scored this PR "
+ f"**{score}/5**, below our merge bar of **{threshold}/5**.\n"
+ "\n"
+ "⏳ **You have 1 day to address Greptile's feedback before this PR is auto-closed.** "
+ "We close low-confidence PRs aggressively to keep the review queue manageable for "
+ "maintainers and contributors alike. **This isn't a rejection of the idea.**\n"
+ "\n"
+ "During the grace period:\n"
+ "\n"
+ "1. Push fixes that address Greptile's feedback (continue using your existing branch is fine).\n"
+ "2. Either:\n"
+ " - Comment `@greptileai` to request a fresh Greptile review. If the new score is "
+ f"**{threshold}/5 or higher**, the PR stays open.\n"
+ " - Or comment `@agent-shin reconsider` to have Agent Shin re-evaluate the PR description.\n"
+ "\n"
+ "If this PR is auto-closed in 24 hours, you'll still have options:\n"
+ "\n"
+ "- Comment `@agent-shin reconsider` after pushing fixes — Agent Shin will re-run triage "
+ "and reopen the PR if it now meets the bar.\n"
+ "- Comment `@greptileai` to request a re-review — that works **even after the PR is closed**.\n"
+ "\n"
+ "Thanks for contributing to LiteLLM. We know auto-closures can sting; the goal is to keep "
+ "the project healthy, not to dismiss your work.\n"
+ "\n"
+ f"{GRACE_COMMENT_MARKER}"
+ )
+
+
+def post_grace_warning(
+ pr: dict,
+ score: int,
+ threshold: int,
+ repo: str | None,
+ dry_run: bool,
+) -> None:
+ """Post the 1-day grace-period warning comment on `pr`.
+
+ The warning carries `GRACE_COMMENT_MARKER` so subsequent runs can
+ detect that the contributor has already been told about the
+ pending close. Does NOT close the PR — the close happens on the
+ next eligible run after `GRACE_PERIOD_SECONDS` elapses (handled
+ by `close_pr`).
+ """
+ pr_number = pr["number"]
+ repo_args = ["--repo", repo] if repo else []
+
+ if dry_run:
+ print(
+ f" [DRY RUN] Would post grace warning to PR #{pr_number} "
+ f"(greptile={score}/5): {pr['title']}"
+ )
+ return
+
+ comment_body = format_grace_warning_comment(score, threshold)
+ gh("pr", "comment", str(pr_number), "--body", comment_body, *repo_args)
+ print(f" Posted grace warning on PR #{pr_number} (greptile={score}/5)")
+
+
+def close_pr(
+ pr: dict,
+ score: int,
+ threshold: int,
+ age_days: int,
+ repo: str | None,
+ dry_run: bool,
+ label: str | None,
+ grace_period_elapsed: bool = True,
+) -> None:
+ """Post the explanatory comment and close the PR."""
+ pr_number = pr["number"]
+ repo_args = ["--repo", repo] if repo else []
+
+ if dry_run:
+ print(
+ f" [DRY RUN] Would close PR #{pr_number} "
+ f"(age={age_days}d, greptile={score}/5): {pr['title']}"
+ )
+ return
+
+ score_sentence = (
+ f"Greptile's most recent review scored this PR **{score}/5**, below "
+ f"our merge bar of **{threshold}/5**, and the 1-day grace period since "
+ "the warning has elapsed.\n\n"
+ if grace_period_elapsed
+ else (
+ f"Greptile's most recent review scored this PR **{score}/5**, "
+ f"below our merge bar of **{threshold}/5**.\n\n"
+ )
+ )
+ comment_body = (
+ f"Closing as part of automated PR triage.\n\n"
+ f"{score_sentence}"
+ "We close low-confidence PRs aggressively to keep the review queue "
+ "manageable for maintainers and contributors alike. **This is not a "
+ "rejection of the idea** — to bring this back:\n\n"
+ "1. Push the fixes that address Greptile's feedback (continue using "
+ "your existing branch is fine).\n"
+ "2. **Open a new PR** with the updated branch. Greptile will review "
+ "it again, and if it scores "
+ f"**{threshold}/5 or higher** a maintainer will take another look.\n\n"
+ "_Why open a new PR instead of reopening this one?_ GitHub does not "
+ "let external contributors reopen a PR that was closed by a bot or "
+ "maintainer, so a fresh PR is the most reliable path forward. If you "
+ "would prefer this exact PR re-evaluated, comment "
+ "`@agent-shin reconsider` once you've pushed the fixes — Agent Shin "
+ "will re-run triage and reopen this PR if it now meets the bar. "
+ "You can also comment `@greptileai` to request a fresh Greptile "
+ "review — that works **even after the PR is closed**.\n\n"
+ "Thanks for contributing to LiteLLM. We know auto-closures can sting; "
+ "the goal is to keep the project healthy, not to dismiss your work."
+ )
+ gh("pr", "comment", str(pr_number), "--body", comment_body, *repo_args)
+
+ if label:
+ try:
+ gh("pr", "edit", str(pr_number), "--add-label", label, *repo_args)
+ except subprocess.CalledProcessError as exc:
+ stderr = (exc.stderr or "").strip()
+ print(f" warn: failed to add label '{label}' to #{pr_number}: {stderr}")
+
+ gh("pr", "close", str(pr_number), *repo_args)
+ print(f" Closed PR #{pr_number} (greptile={score}/5, age={age_days}d)")
+
+
+def evaluate_pr(
+ pr: dict,
+ now: dt.datetime,
+ min_age_days: int,
+ min_score: int,
+ repo: str | None,
+ optout_labels: set[str],
+) -> tuple[str, int | None, int | None]:
+ """Decide what to do with `pr` on this triage run.
+
+ Returns (action, score_or_none, age_days_or_none) where action is one of:
+ "skip-too-young", "skip-optout-label", "skip-internal",
+ "skip-no-greptile-score", "skip-score-ok",
+ "warn-grace", "skip-in-grace-period", or "close".
+
+ Drafts are NOT skipped — the goal is "open PR count == PRs internal
+ collaborators need to action on", and a draft that Greptile scored <4/5
+ is still in that queue. Authors can opt out via the `wip` label (see
+ `DEFAULT_OPTOUT_LABELS`) if they need to keep a long-lived draft open.
+
+ Grace-period semantics: the first time a PR fails the rubric, the
+ action is `warn-grace` — the caller should post a warning comment but
+ NOT close the PR. On a subsequent run, if the warning is still less
+ than `GRACE_PERIOD_SECONDS` old AND the PR still fails, the action is
+ `skip-in-grace-period`. Once the warning ages out and the rubric is
+ still failing, the action is `close`.
+
+ Grace is bypassed for `IMMEDIATE_CLOSE_LOGINS` (test/dogfood
+ accounts), which always go straight to `close` on the first failing
+ run so the bot is dogfoodable end-to-end without a 24h delay.
+ """
+ if has_optout_label(pr, optout_labels):
+ return ("skip-optout-label", None, None)
+
+ created = parse_iso8601(pr["createdAt"])
+ age_days = (now - created).days
+ # `min_age_days` defaults to 0 (close as soon as Greptile scores low).
+ # Set a positive value via --min-age-days for one-off backfill runs that
+ # want to skip very-young PRs.
+ if min_age_days > 0 and age_days < min_age_days:
+ return ("skip-too-young", None, age_days)
+
+ # Only auto-close external OSS contributors. Internal contributors
+ # (BerriAI org members) handle their own backlog.
+ if not is_external_pr_author(pr, repo):
+ return ("skip-internal", None, age_days)
+
+ comments = fetch_pr_comments(pr["number"], repo)
+ extraction = extract_greptile_score(comments)
+ if extraction is None:
+ return ("skip-no-greptile-score", None, age_days)
+
+ score, _ = extraction
+ if score >= min_score:
+ return ("skip-score-ok", score, age_days)
+
+ login = ((pr.get("author") or {}).get("login") or "").lower()
+ if login in IMMEDIATE_CLOSE_LOGINS:
+ return ("close", score, age_days)
+
+ grace_age = seconds_since_last_grace_warning(comments, now=now)
+ if grace_age is None:
+ return ("warn-grace", score, age_days)
+ if grace_age < GRACE_PERIOD_SECONDS:
+ return ("skip-in-grace-period", score, age_days)
+
+ return ("close", score, age_days)
+
+
+def main() -> int:
+ parser = argparse.ArgumentParser(description=__doc__)
+ parser.add_argument(
+ "--repo",
+ type=str,
+ default=None,
+ help="Repository (owner/repo). Auto-detected if omitted.",
+ )
+ parser.add_argument(
+ "--min-age-days",
+ type=int,
+ default=0,
+ help=(
+ "Minimum age (in days) before a PR is eligible. Default 0 = "
+ "close as soon as Greptile flags it. Set a positive value for "
+ "one-off backfill runs that want to spare very-young PRs."
+ ),
+ )
+ parser.add_argument(
+ "--min-score",
+ type=int,
+ default=4,
+ choices=range(1, 6),
+ help="Greptile score below which a PR is closed (default: 4 -> closes <4/5).",
+ )
+ parser.add_argument(
+ "--optout-label",
+ action="append",
+ default=None,
+ help=(
+ "Label(s) that exempt a PR from auto-close. Repeat to add more. "
+ "Case-insensitive. When omitted, defaults to "
+ f"{list(DEFAULT_OPTOUT_LABELS)!r}; passing this flag REPLACES the "
+ "defaults (argparse `append` with a mutable default would append "
+ "instead, which we explicitly avoid)."
+ ),
+ )
+ parser.add_argument(
+ "--close-label",
+ type=str,
+ default=None,
+ help=(
+ "Optional label to add to PRs that get auto-closed "
+ "(e.g. 'auto-closed-low-quality'). Must already exist on the repo."
+ ),
+ )
+ parser.add_argument(
+ "--close",
+ action="store_true",
+ help="Actually close matching PRs (default is dry-run).",
+ )
+ parser.add_argument(
+ "--limit",
+ type=int,
+ default=None,
+ help="Maximum number of PRs to close in one run (safety net).",
+ )
+ args = parser.parse_args()
+
+ dry_run = not args.close
+ if dry_run:
+ print("=== DRY RUN MODE (pass --close to actually close PRs) ===\n")
+
+ print("Fetching open PRs...")
+ prs = fetch_open_prs(args.repo)
+ print(f"Found {len(prs)} open PRs.\n")
+
+ now = dt.datetime.now(dt.timezone.utc)
+ optout_labels = set(args.optout_label or DEFAULT_OPTOUT_LABELS)
+
+ closed = 0
+ summary = {
+ "close": 0,
+ "warn-grace": 0,
+ "skip-in-grace-period": 0,
+ "skip-too-young": 0,
+ "skip-optout-label": 0,
+ "skip-internal": 0,
+ "skip-no-greptile-score": 0,
+ "skip-score-ok": 0,
+ }
+
+ for pr in sorted(prs, key=lambda p: p["createdAt"]):
+ action, score, age_days = evaluate_pr(
+ pr,
+ now,
+ args.min_age_days,
+ args.min_score,
+ args.repo,
+ optout_labels,
+ )
+ summary[action] = summary.get(action, 0) + 1
+
+ # Per-PR dry-run override: `IMMEDIATE_CLOSE_LOGINS` accounts (e.g.
+ # SwiftWinds) always run in real-close mode regardless of the
+ # global `--close` flag. Lets a maintainer dogfood the bot from
+ # an external account while the rest of the open-PR queue stays
+ # on the safe dry-run default.
+ author_login = ((pr.get("author") or {}).get("login") or "").lower()
+ is_immediate = author_login in IMMEDIATE_CLOSE_LOGINS
+ pr_dry_run = dry_run and not is_immediate
+
+ if action == "warn-grace":
+ assert score is not None
+ print(
+ f"#{pr['number']}: \"{pr['title']}\" "
+ f"(age={age_days}d, greptile={score}/5) -> warn-grace"
+ )
+ post_grace_warning(
+ pr,
+ score=score,
+ threshold=args.min_score,
+ repo=args.repo,
+ dry_run=pr_dry_run,
+ )
+ continue
+
+ if action != "close":
+ continue
+
+ assert score is not None and age_days is not None
+ print(
+ f"#{pr['number']}: \"{pr['title']}\" "
+ f"(age={age_days}d, greptile={score}/5) -> close"
+ + (" [immediate-close login]" if is_immediate else "")
+ )
+ close_pr(
+ pr,
+ score=score,
+ threshold=args.min_score,
+ age_days=age_days,
+ repo=args.repo,
+ dry_run=pr_dry_run,
+ label=args.close_label,
+ grace_period_elapsed=not is_immediate,
+ )
+
+ if not pr_dry_run:
+ closed += 1
+ if args.limit is not None and closed >= args.limit:
+ print(f"\nReached --limit={args.limit}; stopping.")
+ break
+
+ print("\n=== Summary ===")
+ for key, value in summary.items():
+ print(f" {key:28s} {value}")
+ # `IMMEDIATE_CLOSE_LOGINS` PRs are closed even in global dry-run mode, so
+ # report actual closures alongside the dry-run "would close" count to avoid
+ # misleading operators into thinking no writes occurred.
+ would_close = summary["close"] - closed
+ if dry_run:
+ if closed:
+ print(f"\nTotal closed: {closed}; would close: {would_close}")
+ else:
+ print(f"\nTotal would close: {would_close}")
+ else:
+ print(f"\nTotal closed: {closed}")
+ print(
+ f"Total {'would warn (grace)' if dry_run else 'warned (grace)'}: "
+ f"{summary['warn-grace']}"
+ )
+ return 0
+
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/.github/scripts/triage_with_llm.py b/.github/scripts/triage_with_llm.py
new file mode 100644
index 00000000000..2442c061b98
--- /dev/null
+++ b/.github/scripts/triage_with_llm.py
@@ -0,0 +1,1126 @@
+#!/usr/bin/env python3
+"""
+Agent Shin — LLM-as-judge triage for external OSS pull requests and issues.
+
+Evaluates a single PR or issue against the contribution rubric and, when the
+LLM judge marks it as failing, posts an explanatory comment + closes the
+PR/issue. Re-triggers on `reopened` so contributors can iterate back in by
+filling in the missing pieces and reopening.
+
+Internal BerriAI contributors (`author_association` in {OWNER, MEMBER,
+COLLABORATOR}) and bot accounts are skipped entirely.
+
+Usage:
+ triage_with_llm.py --repo owner/repo --pr 1234
+ triage_with_llm.py --repo owner/repo --issue 5678
+ triage_with_llm.py --repo owner/repo --pr 1234 --close # actually close
+ triage_with_llm.py --repo owner/repo --pr 1234 --print-prompt # show prompt
+
+Defaults are SAFE: without `--close` the script writes a verdict to stdout (and,
+when running in GitHub Actions, to $GITHUB_STEP_SUMMARY) but takes no GitHub
+write actions.
+
+Environment:
+ GH_TOKEN / GITHUB_TOKEN - for `gh` CLI auth (auto-set in Actions)
+ OPENAI_API_KEY - required when --close is passed
+ OPENAI_BASE_URL - optional (route to any OpenAI-compatible API)
+ TRIAGE_MODEL - optional model override (default: gpt-5.4-mini)
+"""
+
+from __future__ import annotations
+
+import argparse
+import datetime as dt
+import json
+import os
+import re
+import subprocess
+import sys
+import textwrap
+from typing import Any
+
+DEFAULT_MODEL = "gpt-5.4-mini"
+
+INTERNAL_ASSOCIATIONS = frozenset({"OWNER", "MEMBER", "COLLABORATOR"})
+
+# Login of the account that performs Agent Shin's GitHub writes. When the
+# workflow uses `secrets.GITHUB_TOKEN` (our default), the closure / reopen
+# event's `actor.login` is `github-actions[bot]`. The env override exists
+# for local debugging and for repos that wire Agent Shin to a PAT.
+AGENT_SHIN_DEFAULT_BOT_LOGIN = "github-actions[bot]"
+
+# HTML marker appended to every reconsider verdict comment. We grep for this
+# on subsequent reconsider triggers to enforce a short cooldown so that
+# repeated `@agent-shin reconsider` comments don't burn CI/LLM budget.
+# Using a unique HTML comment keeps the marker invisible to humans while
+# being trivially greppable from a comments-list API response.
+RECONSIDER_COMMENT_MARKER = ""
+
+# Minimum gap between two reconsider verdicts on the same PR/issue. Set to
+# 10 minutes — long enough that a contributor can't trivially spam the
+# trigger, short enough that a genuine "I just pushed a fix and reupdated
+# the body" iteration loop isn't punished.
+RECONSIDER_RATE_LIMIT_SECONDS = 600
+
+# HTML marker appended to the grace-period warning comment posted on the
+# first low-quality detection. We grep for this on subsequent triage runs
+# to (a) detect that a warning was already posted (so we don't spam the
+# contributor with duplicate warnings) and (b) measure how long ago it
+# was posted so we know when the grace period has elapsed.
+GRACE_COMMENT_MARKER = ""
+
+# Length of the grace period between the warning comment and the actual
+# auto-close. Set to 24 hours so the contributor has at least one full
+# working day across any time zone to push fixes or comment
+# `@agent-shin reconsider`.
+GRACE_PERIOD_SECONDS = 86400
+
+# Logins (case-insensitive) that bypass BOTH the 1-day grace period AND
+# the dry-run / `AGENT_SHIN_ENABLED` workflow gating — every Agent Shin
+# verdict against a PR/issue from one of these accounts is treated as a
+# real run with immediate close on fail. Useful for dogfooding the bot
+# from an external account that has no push permissions to the repo.
+# Listed lower-case so callers compare via `login.lower() in ...`.
+IMMEDIATE_CLOSE_LOGINS = frozenset({"swiftwinds"})
+
+# Model families that require `reasoning_effort` to be set, and that reject
+# `temperature != 1` unless `reasoning_effort` is "none". For these models we
+# pass `reasoning_effort="none"` so a `temperature=0` deterministic judgment
+# is still accepted. See litellm/llms/openai/chat/gpt_5_transformation.py for
+# the full set of constraints LiteLLM applies to these models.
+GPT5_FAMILY_PREFIX = "gpt-5"
+
+# Regexes for picking off "obvious passes" without burning LLM tokens.
+#
+# Keep this list to GitHub's documented PR-closing keywords only
+# (https://docs.github.com/issues/tracking-your-work-with-issues/linking-a-pull-request-to-an-issue).
+# Casual mentions like "see #1234" or "ref #1234" are intentionally NOT
+# auto-passed — they should fall through to the LLM judge, which has the
+# stricter rubric "a bare issue number without a closing keyword counts only
+# if it's clearly the related issue (not a passing mention)".
+LINKED_ISSUE_PATTERN = re.compile(
+ r"\b(?:fixes|fix|fixed|closes|close|closed|resolves|resolve|resolved)\s+"
+ r"(?:#\d+|https?://github\.com/[\w.-]+/[\w.-]+/issues/\d+)",
+ re.IGNORECASE,
+)
+HTML_COMMENT_PATTERN = re.compile(r"", re.DOTALL)
+
+
+# ---------------------------------------------------------------------------
+# gh helpers
+
+
+def gh(*args: str) -> str:
+ """Run a `gh` CLI command and return stdout. Raises on non-zero exit."""
+ result = subprocess.run(
+ ["gh", *args],
+ capture_output=True,
+ text=True,
+ check=True,
+ )
+ return result.stdout
+
+
+def fetch_pr(repo: str, number: int) -> dict:
+ """Return the full GitHub REST representation of a PR."""
+ return json.loads(gh("api", f"repos/{repo}/pulls/{number}"))
+
+
+def fetch_issue(repo: str, number: int) -> dict:
+ """Return the full GitHub REST representation of an issue."""
+ return json.loads(gh("api", f"repos/{repo}/issues/{number}"))
+
+
+def post_comment(repo: str, number: int, body: str) -> None:
+ """Post an issue-style comment (works for both issues and PRs)."""
+ gh(
+ "api",
+ f"repos/{repo}/issues/{number}/comments",
+ "-X",
+ "POST",
+ "-f",
+ f"body={body}",
+ )
+
+
+def close_pr(repo: str, number: int) -> None:
+ """Close a pull request (state=closed)."""
+ gh(
+ "api",
+ f"repos/{repo}/pulls/{number}",
+ "-X",
+ "PATCH",
+ "-f",
+ "state=closed",
+ )
+
+
+def reopen_pr(repo: str, number: int) -> None:
+ """Reopen a previously-closed pull request (state=open).
+
+ Used by the `@agent-shin reconsider` comment-trigger flow: the bot has
+ write access via GH_TOKEN, so it can reopen on the contributor's behalf
+ even though GitHub doesn't let the OSS author do it themselves.
+ """
+ gh(
+ "api",
+ f"repos/{repo}/pulls/{number}",
+ "-X",
+ "PATCH",
+ "-f",
+ "state=open",
+ )
+
+
+def close_issue(repo: str, number: int, *, not_planned: bool = True) -> None:
+ """Close an issue, marking state_reason=not_planned by default."""
+ args = [
+ "api",
+ f"repos/{repo}/issues/{number}",
+ "-X",
+ "PATCH",
+ "-f",
+ "state=closed",
+ ]
+ if not_planned:
+ args.extend(["-f", "state_reason=not_planned"])
+ gh(*args)
+
+
+def reopen_issue(repo: str, number: int) -> None:
+ """Reopen a previously-closed issue (state=open, state_reason=reopened)."""
+ gh(
+ "api",
+ f"repos/{repo}/issues/{number}",
+ "-X",
+ "PATCH",
+ "-f",
+ "state=open",
+ "-f",
+ "state_reason=reopened",
+ )
+
+
+def _iter_paginated_json(*api_args: str) -> Any:
+ """Yield JSON objects from `gh api --paginate ... -q '.[]'`.
+
+ `gh api --paginate` on a JSON-array endpoint concatenates pages into
+ one stream; `-q '.[]'` flattens that stream into newline-delimited
+ objects (jq-style). This keeps memory bounded for chatty endpoints
+ like issue events/comments on long-lived PRs.
+ """
+ raw = gh("api", "--paginate", *api_args, "-q", ".[]")
+ for line in raw.splitlines():
+ line = line.strip()
+ if not line:
+ continue
+ try:
+ yield json.loads(line)
+ except json.JSONDecodeError:
+ # A malformed line should not blow up the whole guard. Skip and
+ # carry on — at worst the guard fail-closes (returns False /
+ # None) and the caller treats it as "unknown".
+ continue
+
+
+def fetch_last_close_actor(repo: str, number: int) -> str | None:
+ """Return the login of the actor who most recently closed this PR/issue.
+
+ Returns None if no `closed` event is found (unusual for a closed item,
+ but possible if the events API returns nothing — in which case the
+ bot-closed guard should fail-closed, i.e. refuse to reopen).
+ """
+ last: str | None = None
+ for event in _iter_paginated_json(f"repos/{repo}/issues/{number}/events"):
+ if event.get("event") == "closed":
+ last = (event.get("actor") or {}).get("login")
+ return last
+
+
+def was_closed_by_agent_shin(
+ repo: str, number: int, *, bot_login: str | None = None
+) -> bool:
+ """Return True iff the PR/issue was most-recently closed by Agent Shin.
+
+ This is the guard that stops `@agent-shin reconsider` from being used
+ to override a maintainer's closure for non-rubric reasons (security,
+ duplicate, design rejection, etc.). The check is intentionally
+ fail-closed: any uncertainty about who closed the item must be
+ treated as "not the bot" so the destructive reopen path stays gated.
+ """
+ expected = (
+ bot_login
+ or os.environ.get("AGENT_SHIN_BOT_LOGIN")
+ or AGENT_SHIN_DEFAULT_BOT_LOGIN
+ ).lower()
+ actor = fetch_last_close_actor(repo, number)
+ if not actor:
+ return False
+ return actor.lower() == expected
+
+
+def _seconds_since_latest_marker_comment(
+ repo: str,
+ number: int,
+ *,
+ marker: str,
+ bot_login: str | None = None,
+) -> float | None:
+ """Shared helper: return seconds since the bot's most recent comment
+ that contains the given HTML marker, or None if no such comment exists.
+
+ Used by both the reconsider-verdict cooldown and the grace-period
+ warning detection — keeping the iteration logic centralized stops the
+ two paths from drifting (e.g. one fixing a tz parsing bug and the
+ other forgetting to mirror it).
+ """
+ expected_login = (
+ bot_login
+ or os.environ.get("AGENT_SHIN_BOT_LOGIN")
+ or AGENT_SHIN_DEFAULT_BOT_LOGIN
+ ).lower()
+ latest: dt.datetime | None = None
+ for comment in _iter_paginated_json(f"repos/{repo}/issues/{number}/comments"):
+ author = ((comment.get("user") or {}).get("login") or "").lower()
+ if author != expected_login:
+ continue
+ body = comment.get("body") or ""
+ if marker not in body:
+ continue
+ created = comment.get("created_at")
+ if not created:
+ continue
+ try:
+ ts = dt.datetime.fromisoformat(created.replace("Z", "+00:00"))
+ except ValueError:
+ continue
+ if latest is None or ts > latest:
+ latest = ts
+ if latest is None:
+ return None
+ return (dt.datetime.now(dt.timezone.utc) - latest).total_seconds()
+
+
+def seconds_since_last_reconsider_verdict(
+ repo: str, number: int, *, bot_login: str | None = None
+) -> float | None:
+ """Return seconds since the bot's most recent reconsider verdict comment.
+
+ Detects comments by matching the HTML marker `RECONSIDER_COMMENT_MARKER`
+ appended by `format_reopen_comment` and
+ `format_reconsider_still_failing_comment`. Returns None when the bot
+ has never posted a reconsider verdict on this PR/issue (or when the
+ only matching comments are missing a `created_at` timestamp, which
+ shouldn't happen on a real GitHub response).
+ """
+ return _seconds_since_latest_marker_comment(
+ repo, number, marker=RECONSIDER_COMMENT_MARKER, bot_login=bot_login
+ )
+
+
+def seconds_since_last_grace_warning(
+ repo: str, number: int, *, bot_login: str | None = None
+) -> float | None:
+ """Return seconds since the bot's most recent grace-period warning.
+
+ Detects warning comments by matching the HTML marker
+ `GRACE_COMMENT_MARKER` appended by `format_grace_warning_pr_comment`
+ and `format_grace_warning_issue_comment`. Returns None when no
+ grace warning has ever been posted on this PR/issue — that's the
+ "first low-quality detection" signal that drives the warning path.
+ """
+ return _seconds_since_latest_marker_comment(
+ repo, number, marker=GRACE_COMMENT_MARKER, bot_login=bot_login
+ )
+
+
+# ---------------------------------------------------------------------------
+# Author classification
+
+
+def is_internal_contributor(item: dict) -> bool:
+ """Return True if the PR/issue author should be exempted from triage.
+
+ Fail-safe: if `author_association` is missing or empty (which should never
+ happen on a successful GitHub REST response but is possible on schema
+ changes or partial responses), treat the author as INTERNAL so the
+ destructive close path never fires on an unknown contributor. This matches
+ the sibling `is_external_pr_author` in `close_low_quality_prs.py`.
+ """
+ login = ((item.get("user") or {}).get("login") or "").lower()
+ if login.endswith("[bot]") or login in {"dependabot", "github-actions"}:
+ return True
+ association = (item.get("author_association") or "").upper()
+ if not association or association in INTERNAL_ASSOCIATIONS:
+ return True
+ return False
+
+
+# ---------------------------------------------------------------------------
+# Prompt construction
+
+
+def strip_html_comments(text: str) -> str:
+ """Remove HTML comments — template placeholder text shouldn't fool the judge."""
+ return HTML_COMMENT_PATTERN.sub("", text or "")
+
+
+def has_linked_issue(text: str) -> bool:
+ """Heuristic: does this body link to an open issue (Fixes #123 etc.)?"""
+ return bool(LINKED_ISSUE_PATTERN.search(strip_html_comments(text or "")))
+
+
+def build_pr_prompt(*, title: str, body: str) -> str:
+ cleaned_body = strip_html_comments(body or "").strip() or "(empty)"
+ # Dedent the static template *before* interpolating dynamic fields so that
+ # multi-line bodies (whose 2nd+ lines start at column 0) don't defeat the
+ # common-indent computation in textwrap.dedent.
+ template = textwrap.dedent(
+ """
+ You are "Agent Shin", the OSS triage bot for the LiteLLM open-source
+ repository (BerriAI/litellm). Decide whether this external pull request
+ meets the project's contribution standards.
+
+ The PR PASSES triage if it satisfies AT LEAST ONE of:
+
+ (A) It links to a related GitHub issue. Acceptable forms:
+ "Fixes #1234", "Closes #1234", "Resolves #1234",
+ "Refs https://github.com/BerriAI/litellm/issues/1234". A bare
+ issue number without a closing keyword counts only if it's
+ clearly the related issue (not a passing mention).
+
+ (B) The PR body contains ALL of:
+ - A clear problem description (what bug or missing feature this
+ addresses, beyond the title).
+ - Expected vs. actual behavior (or, for features, "what's
+ possible now vs. with this PR").
+ - Visual QA proof: before/after screenshots, a screen recording,
+ terminal output, log output, or test output demonstrating the
+ fix or feature works end-to-end. Saying "I tested it" is NOT
+ proof.
+
+ Bias toward PASS when the PR has structure and context — only FAIL when
+ the body is empty, copy-paste filler from the template, or genuinely
+ missing both a linked issue AND the core elements of (B).
+
+ Respond with a single JSON object, no prose:
+
+ {{
+ "verdict": "pass" | "fail",
+ "linked_issue": boolean,
+ "has_problem_description": boolean,
+ "has_expected_vs_actual": boolean,
+ "has_qa_proof": boolean,
+ "missing": ["plain-english strings naming what is missing"],
+ "explanation": "1-2 sentence reasoning for the team to skim"
+ }}
+
+ ---
+ PR title: {title}
+
+ PR body:
+ ---
+ {cleaned_body}
+ ---
+ """
+ ).strip()
+ return template.format(title=title, cleaned_body=cleaned_body)
+
+
+def build_issue_prompt(*, title: str, body: str) -> str:
+ cleaned_body = strip_html_comments(body or "").strip() or "(empty)"
+ # Dedent the static template *before* interpolating dynamic fields so that
+ # multi-line bodies (whose 2nd+ lines start at column 0) don't defeat the
+ # common-indent computation in textwrap.dedent.
+ template = textwrap.dedent(
+ """
+ You are "Agent Shin", the OSS triage bot for the LiteLLM open-source
+ repository (BerriAI/litellm). Decide whether this GitHub issue meets
+ the project's reporting standards.
+
+ For a BUG REPORT the issue PASSES triage when it contains ALL of:
+ - A clear reproduction (steps, runnable code snippet, curl command,
+ or example config the maintainer can paste into their machine).
+ - Screenshot, terminal output, traceback, or log output as proof of
+ the bug.
+ - Expected vs. actual behavior.
+
+ For a FEATURE REQUEST the issue PASSES triage when it contains ALL of:
+ - A clear description of the proposed feature (what should LiteLLM do
+ that it does not today).
+ - Motivation / use case with a concrete example (config, API call,
+ UI flow, or scenario showing what's blocked today).
+
+ Bias toward PASS when the issue has structure and context — only FAIL
+ when the body is empty, copy-paste template placeholder text, or a
+ one-line "X is broken" with no detail. Asking clarifying questions is
+ OK content; mark such issues PASS.
+
+ Respond with a single JSON object, no prose:
+
+ {{
+ "verdict": "pass" | "fail",
+ "kind": "bug" | "feature" | "other",
+ "has_repro": boolean,
+ "has_proof": boolean,
+ "has_expected_vs_actual": boolean,
+ "has_motivation_example": boolean,
+ "missing": ["plain-english strings naming what is missing"],
+ "explanation": "1-2 sentence reasoning for the team to skim"
+ }}
+
+ ---
+ Issue title: {title}
+
+ Issue body:
+ ---
+ {cleaned_body}
+ ---
+ """
+ ).strip()
+ return template.format(title=title, cleaned_body=cleaned_body)
+
+
+# ---------------------------------------------------------------------------
+# LLM call + verdict parsing
+
+
+def call_llm_judge(
+ prompt: str, *, model: str, api_key: str, base_url: str | None
+) -> str:
+ """Call an OpenAI-compatible chat completions endpoint. Returns raw text."""
+ # Import inside the function so unit tests that monkey-patch this never
+ # need the openai package installed.
+ from openai import OpenAI
+
+ client = (
+ OpenAI(api_key=api_key, base_url=base_url)
+ if base_url
+ else OpenAI(api_key=api_key)
+ )
+ kwargs: dict[str, Any] = {
+ "model": model,
+ "messages": [{"role": "user", "content": prompt}],
+ "temperature": 0,
+ "response_format": {"type": "json_object"},
+ }
+ # gpt-5.x reasoning models reject `temperature != 1` unless
+ # `reasoning_effort` is explicitly "none". Set it via `extra_body` so this
+ # works across openai SDK versions regardless of whether the SDK natively
+ # types `reasoning_effort` as a top-level chat-completions param yet.
+ if model.lower().startswith(GPT5_FAMILY_PREFIX):
+ kwargs["extra_body"] = {"reasoning_effort": "none"}
+ response = client.chat.completions.create(**kwargs)
+ return response.choices[0].message.content or ""
+
+
+def parse_verdict(raw: str) -> dict:
+ """Parse the LLM's JSON response. Tolerates ```json fences and stray text."""
+ if not raw:
+ raise ValueError("empty LLM response")
+ text = raw.strip()
+ if text.startswith("```"):
+ text = re.sub(r"^```(?:json)?\s*", "", text)
+ text = re.sub(r"\s*```$", "", text)
+ try:
+ return json.loads(text)
+ except json.JSONDecodeError:
+ match = re.search(r"\{.*\}", text, re.DOTALL)
+ if not match:
+ raise ValueError(f"could not extract JSON from LLM response: {raw[:200]}")
+ return json.loads(match.group(0))
+
+
+# ---------------------------------------------------------------------------
+# Comment composition
+
+
+def _format_missing(missing: list[str]) -> str:
+ if not missing:
+ return "- (see explanation below)"
+ return "\n".join(f"- {m}" for m in missing)
+
+
+def format_pr_close_comment(verdict: dict) -> str:
+ missing_lines = _format_missing(verdict.get("missing") or [])
+ explanation = verdict.get("explanation") or ""
+ return (
+ "👋 Hi, thanks for the PR! I'm **Agent Shin**, the automated triage bot for this repository.\n"
+ "\n"
+ "This PR is being **auto-closed** because it does not yet meet the bar described in our "
+ "[pull-request template](https://github.com/BerriAI/litellm/blob/main/.github/pull_request_template.md). "
+ "Specifically, I couldn't find:\n"
+ "\n"
+ f"{missing_lines}\n"
+ "\n"
+ f"> {explanation}\n"
+ "\n"
+ "**This isn't a rejection of the idea.** To bring this PR back:\n"
+ "\n"
+ "1. Update the PR description to either:\n"
+ " - Link a related GitHub issue (e.g. `Fixes #1234`), OR\n"
+ " - Add a clear **problem description**, **expected vs. actual behavior**, and **visual QA proof** "
+ "(before/after screenshots, a short screen recording, or terminal/log output).\n"
+ "2. Either:\n"
+ " - **Open a new PR** with the same fixes — recommended path. GitHub does not let external "
+ "contributors reopen a PR that was closed by a bot/maintainer, so a fresh PR is the most reliable way "
+ "to get back into the review queue.\n"
+ " - **Or** comment `@agent-shin reconsider` on this closed PR after updating the description. "
+ "I'll re-run the triage; if it now passes, I'll reopen this PR automatically.\n"
+ " - You can also comment `@greptileai` on this PR to request a fresh Greptile review — that "
+ "still works **even after the PR is closed**, and a higher score is one of the signals that "
+ "lifts the PR back into the queue.\n"
+ "\n"
+ "Internal BerriAI contributors: this rubric doesn't apply to you — ping a maintainer.\n"
+ "\n"
+ "_(I'm an LLM, so I'm not infallible. If you think I got this wrong, comment "
+ "`@agent-shin reconsider` or ping a maintainer — they'll override me.)_"
+ )
+
+
+def format_issue_close_comment(verdict: dict) -> str:
+ missing_lines = _format_missing(verdict.get("missing") or [])
+ explanation = verdict.get("explanation") or ""
+ return (
+ "👋 Hi, thanks for filing this! I'm **Agent Shin**, the automated triage bot for this repository.\n"
+ "\n"
+ "This issue is being **auto-closed** because it doesn't yet have enough detail for a maintainer to act on. "
+ "Specifically, I couldn't find:\n"
+ "\n"
+ f"{missing_lines}\n"
+ "\n"
+ f"> {explanation}\n"
+ "\n"
+ "**This isn't a \"won't fix\".** To bring this issue back:\n"
+ "\n"
+ "1. Edit the issue to add the missing pieces:\n"
+ " - For **bug reports**: a runnable reproduction (code / curl / config), expected vs. actual behavior, "
+ "and a screenshot / traceback / log showing the bug.\n"
+ " - For **feature requests**: a concrete description of what should change, plus a use case and example "
+ "(config / API call / UI flow).\n"
+ "2. Comment `@agent-shin reconsider` on this issue once you've updated it. "
+ "I'll re-run triage and reopen the issue if it now meets the bar. "
+ "(GitHub doesn't always let the original reporter reopen a bot-closed issue, "
+ "so the comment-based reconsider is the reliable path.)\n"
+ "\n"
+ "Internal BerriAI contributors: this rubric doesn't apply to you — ping a maintainer.\n"
+ "\n"
+ "_(I'm an LLM, so I'm not infallible. If you think I got this wrong, comment "
+ "`@agent-shin reconsider` or ping a maintainer — they'll override me.)_"
+ )
+
+
+def format_grace_warning_pr_comment(verdict: dict) -> str:
+ """Comment posted on the FIRST low-quality detection — gives the
+ contributor a 1-day grace window to fix the PR before the next
+ triage run actually closes it.
+
+ This is the "before-close" warning. On the second triage run, if the
+ grace marker is older than `GRACE_PERIOD_SECONDS` AND the PR still
+ fails the rubric, the close path runs (which posts
+ `format_pr_close_comment` and closes the PR).
+ """
+ missing_lines = _format_missing(verdict.get("missing") or [])
+ explanation = verdict.get("explanation") or ""
+ return (
+ "👋 Hi, thanks for the PR! I'm **Agent Shin**, the automated triage bot for this repository.\n"
+ "\n"
+ "Heads up — this PR does not yet meet the bar described in our "
+ "[pull-request template](https://github.com/BerriAI/litellm/blob/main/.github/pull_request_template.md). "
+ "Specifically, I couldn't find:\n"
+ "\n"
+ f"{missing_lines}\n"
+ "\n"
+ f"> {explanation}\n"
+ "\n"
+ "⏳ **You have 1 day to address this before this PR is auto-closed.** "
+ "During the grace period:\n"
+ "\n"
+ "1. Update the PR description to either:\n"
+ " - Link a related GitHub issue (e.g. `Fixes #1234`), OR\n"
+ " - Add a clear **problem description**, **expected vs. actual behavior**, and **visual QA proof** "
+ "(before/after screenshots, a short screen recording, or terminal/log output).\n"
+ "2. Comment `@agent-shin reconsider` on this PR after updating it. If your update meets the "
+ "bar, I'll skip the auto-close and a maintainer will take another look.\n"
+ "\n"
+ "If this PR is auto-closed in 24 hours, you'll still have options:\n"
+ "\n"
+ "- Comment `@agent-shin reconsider` to have me re-evaluate (and reopen the PR if it now meets the bar).\n"
+ "- Comment `@greptileai` to request a fresh Greptile review — that works **even after the PR is closed**.\n"
+ "\n"
+ "Internal BerriAI contributors: this rubric doesn't apply to you — ping a maintainer.\n"
+ "\n"
+ "_(I'm an LLM, so I'm not infallible. If you think I got this wrong, comment "
+ "`@agent-shin reconsider` or ping a maintainer — they'll override me.)_\n"
+ "\n"
+ f"{GRACE_COMMENT_MARKER}"
+ )
+
+
+def format_grace_warning_issue_comment(verdict: dict) -> str:
+ """Issue analogue of `format_grace_warning_pr_comment`."""
+ missing_lines = _format_missing(verdict.get("missing") or [])
+ explanation = verdict.get("explanation") or ""
+ return (
+ "👋 Hi, thanks for filing this! I'm **Agent Shin**, the automated triage bot for this repository.\n"
+ "\n"
+ "Heads up — this issue doesn't yet have enough detail for a maintainer to act on. "
+ "Specifically, I couldn't find:\n"
+ "\n"
+ f"{missing_lines}\n"
+ "\n"
+ f"> {explanation}\n"
+ "\n"
+ "⏳ **You have 1 day to address this before this issue is auto-closed.** "
+ "During the grace period:\n"
+ "\n"
+ "1. Edit the issue to add the missing pieces:\n"
+ " - For **bug reports**: a runnable reproduction (code / curl / config), expected vs. actual behavior, "
+ "and a screenshot / traceback / log showing the bug.\n"
+ " - For **feature requests**: a concrete description of what should change, plus a use case and example "
+ "(config / API call / UI flow).\n"
+ "2. Comment `@agent-shin reconsider` on this issue after updating it. If your update meets the bar, "
+ "I'll skip the auto-close and a maintainer will take another look.\n"
+ "\n"
+ "If this issue is auto-closed in 24 hours, you can still comment `@agent-shin reconsider` to have "
+ "me re-evaluate (and reopen the issue if it now meets the bar).\n"
+ "\n"
+ "Internal BerriAI contributors: this rubric doesn't apply to you — ping a maintainer.\n"
+ "\n"
+ "_(I'm an LLM, so I'm not infallible. If you think I got this wrong, comment "
+ "`@agent-shin reconsider` or ping a maintainer — they'll override me.)_\n"
+ "\n"
+ f"{GRACE_COMMENT_MARKER}"
+ )
+
+
+# ---------------------------------------------------------------------------
+# Step-summary helpers
+
+
+def write_step_summary(content: str) -> None:
+ """When running inside GitHub Actions, append to the step summary file."""
+ path = os.environ.get("GITHUB_STEP_SUMMARY")
+ if not path:
+ return
+ try:
+ with open(path, "a", encoding="utf-8") as handle:
+ handle.write(content)
+ if not content.endswith("\n"):
+ handle.write("\n")
+ except OSError as exc:
+ print(f"warn: failed to write step summary: {exc}", file=sys.stderr)
+
+
+# ---------------------------------------------------------------------------
+# Core orchestration
+
+
+def format_reopen_comment(kind: str) -> str:
+ """Comment posted when Agent Shin reopens after a successful reconsider."""
+ noun = "PR" if kind == "pr" else "issue"
+ # The trailing HTML marker is used by `seconds_since_last_reconsider_verdict`
+ # to enforce a cooldown between repeated `@agent-shin reconsider` triggers.
+ # Keep the marker on its own line so it doesn't disturb the rendered text.
+ return (
+ f"♻️ **Re-evaluated and reopened.** Thanks for updating the {noun}!\n"
+ "\n"
+ "Agent Shin re-ran triage on the latest description and it now meets "
+ "the bar. A maintainer will take another look soon — please don't "
+ f"close this {noun} again unless asked to.\n"
+ "\n"
+ "_(If a maintainer ends up closing this for non-rubric reasons, that "
+ "decision stands; comment `@agent-shin reconsider` again only if you "
+ "have substantively new information.)_\n"
+ "\n"
+ f"{RECONSIDER_COMMENT_MARKER}"
+ )
+
+
+def format_reconsider_still_failing_comment(kind: str, verdict: dict) -> str:
+ """Comment posted when reconsider re-runs triage but the verdict is still fail."""
+ missing_lines = _format_missing(verdict.get("missing") or [])
+ explanation = verdict.get("explanation") or ""
+ noun = "PR" if kind == "pr" else "issue"
+ # The trailing HTML marker is used by `seconds_since_last_reconsider_verdict`
+ # to enforce a cooldown between repeated `@agent-shin reconsider` triggers.
+ return (
+ f"⏸️ **Re-evaluated; this {noun} still doesn't meet the rubric.**\n"
+ "\n"
+ "Agent Shin re-ran triage on the current description but is still "
+ "missing:\n"
+ "\n"
+ f"{missing_lines}\n"
+ "\n"
+ f"> {explanation}\n"
+ "\n"
+ "Update the description with the missing pieces and comment "
+ "`@agent-shin reconsider` again, or ping a maintainer if you think "
+ "I got this wrong.\n"
+ "\n"
+ "_(I'm an LLM and I'm not infallible.)_\n"
+ "\n"
+ f"{RECONSIDER_COMMENT_MARKER}"
+ )
+
+
+def triage(
+ *,
+ repo: str,
+ kind: str,
+ number: int,
+ close: bool,
+ model: str,
+ judge: Any = None,
+ print_prompt: bool = False,
+ reconsider: bool = False,
+) -> dict:
+ """Triage a single PR or issue. Returns a result dict for logging/tests.
+
+ `judge` is an optional callable `(prompt) -> str` for tests / dry-run with
+ a stub. In production, leave it None and the script uses `call_llm_judge`.
+
+ When `reconsider=True`, the closed-state guard is skipped and a
+ fail-but-no-comment is replaced with a "still failing" comment + leave
+ closed; a pass triggers `reopen_pr`/`reopen_issue` plus a reopen comment.
+ Reconsider mode is intended for the `@agent-shin reconsider` comment
+ trigger. Like regular triage, `close=False` keeps reconsider in dry-run
+ (returns `would-reopen` / `would-reconsider-still-failing` so a local
+ operator can preview without write side effects); the workflow only
+ passes `--close` when `AGENT_SHIN_ENABLED=true`.
+
+ Reconsider mode adds two extra safety guards on top of the regular
+ triage skip-internal-author check:
+
+ 1. **Bot-closed guard.** Only reopens if the most recent close was
+ performed by the bot identity (default `github-actions[bot]`).
+ This stops a contributor from using `@agent-shin reconsider` to
+ override a maintainer's close for non-rubric reasons.
+ 2. **Rate-limit guard.** If the bot has already posted a reconsider
+ verdict on this PR/issue within `RECONSIDER_RATE_LIMIT_SECONDS`,
+ skip — repeated triggers from the same contributor shouldn't burn
+ CI minutes or LLM budget.
+ """
+ fetcher = {"pr": fetch_pr, "issue": fetch_issue}[kind]
+ item = fetcher(repo, number)
+
+ title = item.get("title") or ""
+ body = item.get("body") or ""
+ login = (item.get("user") or {}).get("login") or ""
+ association = item.get("author_association") or ""
+ state = item.get("state") or ""
+
+ base_result = {
+ "kind": kind,
+ "number": number,
+ "title": title,
+ "author": login,
+ "author_association": association,
+ "state": state,
+ "reconsider": reconsider,
+ }
+
+ # Reconsider only makes sense on a closed PR/issue. A "reconsider on an
+ # open PR" is a no-op (the regular triage flow already evaluates open
+ # PRs); return a clear skip so the workflow can short-circuit.
+ if reconsider:
+ if state != "closed":
+ return {**base_result, "action": "skip-not-closed"}
+ else:
+ if state != "open":
+ return {**base_result, "action": "skip-not-open"}
+
+ if is_internal_contributor(item):
+ return {**base_result, "action": "skip-internal-author"}
+
+ # Reconsider-only guards — these run BEFORE the LLM call so a
+ # maintainer-closed PR / rate-limited trigger never spends LLM budget.
+ if reconsider:
+ if not was_closed_by_agent_shin(repo, number):
+ return {**base_result, "action": "skip-not-bot-closed"}
+ age = seconds_since_last_reconsider_verdict(repo, number)
+ if age is not None and age < RECONSIDER_RATE_LIMIT_SECONDS:
+ return {
+ **base_result,
+ "action": "skip-rate-limited",
+ "rate_limit_age_seconds": age,
+ "rate_limit_window_seconds": RECONSIDER_RATE_LIMIT_SECONDS,
+ }
+
+ if kind == "pr":
+ prompt = build_pr_prompt(title=title, body=body)
+ # Short-circuit: if body very clearly links a related issue, just pass.
+ if has_linked_issue(body):
+ base = {
+ **base_result,
+ "action": "pass-linked-issue",
+ "verdict": {
+ "verdict": "pass",
+ "linked_issue": True,
+ "explanation": "Linked-issue regex matched; LLM was not called.",
+ },
+ }
+ if reconsider:
+ # Pass-on-reconsider -> reopen the PR with a friendly comment.
+ reopen_body = format_reopen_comment(kind)
+ if not close:
+ return {
+ **base,
+ "action": "would-reopen",
+ "comment": reopen_body,
+ }
+ post_comment(repo, number, reopen_body)
+ reopen_pr(repo, number)
+ return {
+ **base,
+ "action": "reopened",
+ "comment": reopen_body,
+ }
+ return base
+ else:
+ prompt = build_issue_prompt(title=title, body=body)
+
+ if print_prompt:
+ return {**base_result, "action": "print-prompt", "prompt": prompt}
+
+ if judge is None:
+ api_key = os.environ.get("OPENAI_API_KEY")
+ if not api_key:
+ # No key configured — never take a destructive action. Report skip.
+ return {
+ **base_result,
+ "action": "skip-no-llm-key",
+ "prompt_preview": prompt[:200],
+ }
+ base_url = os.environ.get("OPENAI_BASE_URL") or None
+
+ def judge(p: str) -> str:
+ return call_llm_judge(p, model=model, api_key=api_key, base_url=base_url)
+
+ try:
+ raw = judge(prompt)
+ verdict = parse_verdict(raw)
+ except Exception as exc: # noqa: BLE001 - judge errors must never close PRs
+ return {**base_result, "action": "skip-llm-error", "error": str(exc)}
+
+ decision = (verdict.get("verdict") or "").lower()
+
+ if reconsider:
+ # Reconsider: pass -> reopen + post reopen comment;
+ # fail -> leave closed + post a "still failing" comment so the
+ # contributor can iterate again.
+ # In dry-run (`close=False`) we return `would-*` actions instead
+ # of touching GitHub state, mirroring the regular triage flow's
+ # `would-close`. This lets a local operator preview the outcome
+ # of `python triage_with_llm.py --reconsider --pr N` without
+ # risking accidental comments or reopens.
+ if decision != "fail":
+ reopen_body = format_reopen_comment(kind)
+ if not close:
+ return {
+ **base_result,
+ "action": "would-reopen",
+ "verdict": verdict,
+ "comment": reopen_body,
+ }
+ post_comment(repo, number, reopen_body)
+ if kind == "pr":
+ reopen_pr(repo, number)
+ else:
+ reopen_issue(repo, number)
+ return {
+ **base_result,
+ "action": "reopened",
+ "verdict": verdict,
+ "comment": reopen_body,
+ }
+ still_failing = format_reconsider_still_failing_comment(kind, verdict)
+ if not close:
+ return {
+ **base_result,
+ "action": "would-reconsider-still-failing",
+ "verdict": verdict,
+ "comment": still_failing,
+ }
+ post_comment(repo, number, still_failing)
+ return {
+ **base_result,
+ "action": "reconsider-still-failing",
+ "verdict": verdict,
+ "comment": still_failing,
+ }
+
+ if decision != "fail":
+ return {**base_result, "action": "pass-llm", "verdict": verdict}
+
+ # Grace-period flow: on the first low-quality detection, post a warning
+ # comment instead of closing immediately. On a subsequent triage run
+ # (manual re-trigger, or the daily `close_low_quality_prs.py` cron
+ # finding the same PR in its own pass), if `GRACE_PERIOD_SECONDS` has
+ # elapsed since the warning AND the PR still fails the rubric, close.
+ #
+ # `IMMEDIATE_CLOSE_LOGINS` (e.g. test/dogfood accounts like SwiftWinds)
+ # bypass the grace period entirely — every fail is treated as a real
+ # close run. This is intentional: those accounts exist specifically to
+ # exercise the bot end-to-end, and waiting a day per iteration kills
+ # the feedback loop.
+ is_immediate = login.lower() in IMMEDIATE_CLOSE_LOGINS
+
+ if not is_immediate:
+ grace_age = seconds_since_last_grace_warning(repo, number)
+ if grace_age is None:
+ warning_body = (
+ format_grace_warning_pr_comment(verdict)
+ if kind == "pr"
+ else format_grace_warning_issue_comment(verdict)
+ )
+ if not close:
+ return {
+ **base_result,
+ "action": "would-warn-grace",
+ "verdict": verdict,
+ "comment": warning_body,
+ }
+ post_comment(repo, number, warning_body)
+ return {
+ **base_result,
+ "action": "warned-grace",
+ "verdict": verdict,
+ "comment": warning_body,
+ }
+ if grace_age < GRACE_PERIOD_SECONDS:
+ return {
+ **base_result,
+ "action": "skip-in-grace-period",
+ "verdict": verdict,
+ "grace_age_seconds": grace_age,
+ "grace_period_seconds": GRACE_PERIOD_SECONDS,
+ }
+
+ # Either the grace window has elapsed or this author bypasses grace —
+ # proceed to the actual close path. `--close` still gates the
+ # destructive write for the regular-author path so a local operator
+ # can preview a "would close after grace" verdict; immediate-close
+ # accounts (`is_immediate`) ignore `--close` so a workflow that
+ # forces dry-run for the global population can still take real
+ # action on those test accounts.
+ if not close and not is_immediate:
+ return {**base_result, "action": "would-close", "verdict": verdict}
+
+ comment_body = (
+ format_pr_close_comment(verdict)
+ if kind == "pr"
+ else format_issue_close_comment(verdict)
+ )
+ post_comment(repo, number, comment_body)
+ if kind == "pr":
+ close_pr(repo, number)
+ else:
+ close_issue(repo, number)
+
+ return {
+ **base_result,
+ "action": "closed",
+ "verdict": verdict,
+ "comment": comment_body,
+ "immediate_close": is_immediate,
+ }
+
+
+# ---------------------------------------------------------------------------
+# CLI
+
+
+def render_summary(result: dict) -> str:
+ """Render a human-readable summary block (used for stdout + step summary)."""
+ lines = ["## Agent Shin verdict", ""]
+ lines.append(
+ f"- **{result['kind'].upper()} #{result['number']}**: {result.get('title', '')}"
+ )
+ lines.append(
+ f"- **Author**: `{result.get('author', '')}` ({result.get('author_association', '')})"
+ )
+ lines.append(f"- **State**: {result.get('state', '')}")
+ lines.append(f"- **Action**: `{result['action']}`")
+ verdict = result.get("verdict")
+ if verdict:
+ lines.append("")
+ lines.append("```json")
+ lines.append(json.dumps(verdict, indent=2))
+ lines.append("```")
+ error = result.get("error")
+ if error:
+ lines.append("")
+ lines.append(f"_LLM error: {error}_")
+ comment = result.get("comment")
+ if comment:
+ lines.append("")
+ lines.append("### Posted comment:")
+ lines.append("")
+ lines.append("> " + comment.replace("\n", "\n> "))
+ return "\n".join(lines)
+
+
+def main() -> int:
+ parser = argparse.ArgumentParser(description=__doc__)
+ parser.add_argument("--repo", required=True, help="Repository (owner/repo).")
+ target = parser.add_mutually_exclusive_group(required=True)
+ target.add_argument("--pr", type=int, help="Pull request number to triage.")
+ target.add_argument("--issue", type=int, help="Issue number to triage.")
+ parser.add_argument(
+ "--close",
+ action="store_true",
+ help="Actually post comment + close on fail (default: dry run).",
+ )
+ parser.add_argument(
+ "--model",
+ # `os.environ.get("TRIAGE_MODEL", DEFAULT_MODEL)` would return "" when
+ # GitHub Actions exposes an unset repo variable as an empty-string env
+ # var, silently bypassing DEFAULT_MODEL and causing every call to fail
+ # as `skip-llm-error`. The `or` guard collapses empty -> default.
+ default=os.environ.get("TRIAGE_MODEL") or DEFAULT_MODEL,
+ help=f"OpenAI-compatible model name (default: {DEFAULT_MODEL}).",
+ )
+ parser.add_argument(
+ "--print-prompt",
+ action="store_true",
+ help="Print the prompt that would be sent to the judge and exit.",
+ )
+ parser.add_argument(
+ "--reconsider",
+ action="store_true",
+ help=(
+ "Re-run triage on a CLOSED PR/issue and reopen it on pass. "
+ "Used by the `@agent-shin reconsider` comment-trigger workflow. "
+ "Only invoke this from a workflow that has already gated on "
+ "AGENT_SHIN_ENABLED=true and verified the commenter is the "
+ "PR/issue author or an internal collaborator."
+ ),
+ )
+ args = parser.parse_args()
+
+ kind = "pr" if args.pr is not None else "issue"
+ number = args.pr if args.pr is not None else args.issue
+
+ result = triage(
+ repo=args.repo,
+ kind=kind,
+ number=number,
+ close=args.close,
+ model=args.model,
+ print_prompt=args.print_prompt,
+ reconsider=args.reconsider,
+ )
+
+ if result.get("action") == "print-prompt":
+ print(result["prompt"])
+ return 0
+
+ summary = render_summary(result)
+ print(summary)
+ write_step_summary(summary + "\n")
+ return 0
+
+
+if __name__ == "__main__":
+ sys.exit(main())
diff --git a/.github/workflows/close_low_quality_prs.yml b/.github/workflows/close_low_quality_prs.yml
new file mode 100644
index 00000000000..2401be84000
--- /dev/null
+++ b/.github/workflows/close_low_quality_prs.yml
@@ -0,0 +1,92 @@
+name: Close Low-Quality PRs
+
+# Auto-close any open PR (including drafts, regardless of age) authored by an
+# external OSS contributor that Greptile reviewed with a confidence score
+# below 4/5. Closures are explained in a comment that tells the contributor
+# to push fixes and open a fresh PR (since OSS authors cannot reopen a PR
+# closed by a bot/maintainer) or comment `@agent-shin reconsider` to have
+# Agent Shin re-evaluate.
+#
+# Manual one-off run:
+# gh workflow run "Close Low-Quality PRs" -f close=true
+#
+# Dry-run preview (no PRs are touched):
+# gh workflow run "Close Low-Quality PRs" -f close=false
+
+on:
+ schedule:
+ # Daily at 09:00 UTC. Pairs well with the stale-issue workflow at midnight.
+ - cron: "0 9 * * *"
+ workflow_dispatch:
+ inputs:
+ close:
+ description: "Actually close matching PRs (false = dry run)."
+ required: false
+ default: "false"
+ type: choice
+ options:
+ - "true"
+ - "false"
+ min_age_days:
+ description: "Minimum PR age in days (default 0 = no age filter)."
+ required: false
+ default: "0"
+ min_score:
+ description: "Greptile score below which a PR is closed (1-5)."
+ required: false
+ default: "4"
+ limit:
+ description: "Maximum number of PRs to close in a single run."
+ required: false
+ default: "25"
+
+permissions:
+ contents: read
+ pull-requests: write
+ issues: write
+
+jobs:
+ close-low-quality-prs:
+ if: github.repository == 'BerriAI/litellm'
+ runs-on: ubuntu-latest
+ steps:
+ - name: Checkout triage script
+ uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+ with:
+ sparse-checkout: .github/scripts
+ persist-credentials: false
+
+ - name: Set up Python
+ uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+
+ - name: Run low-quality PR closer
+ env:
+ GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+ # Scheduled runs are ALWAYS dry-run, even when AGENT_SHIN_ENABLED is
+ # "true", so the team can QA the closer's verdicts in step summaries
+ # before any contributor sees a PR closed. Real closures only happen
+ # on manual workflow_dispatch with close=true (and the variable set).
+ CLOSE_FLAG: ${{ github.event.inputs.close || 'false' }}
+ AGENT_SHIN_ENABLED: ${{ vars.AGENT_SHIN_ENABLED }}
+ MIN_AGE_DAYS: ${{ github.event.inputs.min_age_days || '0' }}
+ MIN_SCORE: ${{ github.event.inputs.min_score || '4' }}
+ LIMIT: ${{ github.event.inputs.limit || '25' }}
+ run: |
+ set -euo pipefail
+ ARGS=(
+ --repo "${{ github.repository }}"
+ --min-age-days "${MIN_AGE_DAYS}"
+ --min-score "${MIN_SCORE}"
+ --limit "${LIMIT}"
+ )
+ if [ "${AGENT_SHIN_ENABLED:-false}" != "true" ]; then
+ echo "::notice::AGENT_SHIN_ENABLED is not 'true' -> forcing dry-run regardless of close input."
+ elif [ "${GITHUB_EVENT_NAME:-}" = "workflow_dispatch" ] && [ "${CLOSE_FLAG}" = "true" ]; then
+ ARGS+=(--close)
+ echo "::notice::Running in close-on-fail mode."
+ else
+ echo "::notice::AGENT_SHIN_ENABLED is true but this trigger is dry-run (scheduled event or close=false)."
+ fi
+ python3 .github/scripts/close_low_quality_prs.py "${ARGS[@]}"
diff --git a/.github/workflows/triage_issue_with_llm.yml b/.github/workflows/triage_issue_with_llm.yml
new file mode 100644
index 00000000000..ff0497f9893
--- /dev/null
+++ b/.github/workflows/triage_issue_with_llm.yml
@@ -0,0 +1,90 @@
+name: Agent Shin — Issue triage
+
+# LLM-as-judge triage for external GitHub issues.
+#
+# DRY-RUN BY DEFAULT. See .github/workflows/triage_pr_with_llm.yml for the
+# enablement procedure — same repo variable (`AGENT_SHIN_ENABLED=true`)
+# unlocks the PR and issue triage flows together.
+
+on:
+ issues:
+ types: [opened, reopened]
+ workflow_dispatch:
+ inputs:
+ issue_number:
+ description: "Issue number to triage manually."
+ required: true
+ close:
+ description: "If true and AGENT_SHIN_ENABLED=true, actually close on fail."
+ required: false
+ default: "false"
+ type: choice
+ options:
+ - "true"
+ - "false"
+
+permissions:
+ contents: read
+ issues: write
+
+jobs:
+ triage:
+ if: github.repository == 'BerriAI/litellm'
+ runs-on: ubuntu-latest
+ steps:
+ - name: Checkout triage script
+ uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+ with:
+ sparse-checkout: .github/scripts
+ persist-credentials: false
+
+ - name: Set up Python
+ uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+
+ - name: Install LLM client
+ run: pip install --no-cache-dir "openai>=1.40.0"
+
+ - name: Run Agent Shin
+ env:
+ GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+ OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+ OPENAI_BASE_URL: ${{ vars.OPENAI_BASE_URL }}
+ TRIAGE_MODEL: ${{ vars.TRIAGE_MODEL }}
+ AGENT_SHIN_ENABLED: ${{ vars.AGENT_SHIN_ENABLED }}
+ DISPATCH_CLOSE: ${{ github.event.inputs.close }}
+ ISSUE_NUMBER: ${{ github.event.issue.number || github.event.inputs.issue_number }}
+ run: |
+ set -euo pipefail
+ ARGS=(--repo "${{ github.repository }}" --issue "${ISSUE_NUMBER}")
+ # Fail-safe gating: only the EXACT string "true" enables the
+ # destructive --close path. The workflow_dispatch input is a
+ # `choice` dropdown of "true"/"false" so the UI is constrained,
+ # but the API (`gh workflow run -f close=...`) accepts any
+ # string, and a `!= "false"` check would treat "True", "yes",
+ # "1", "TRUE", typos, and accidental whitespace as enabling
+ # closure. Mirror the Greptile closer's `= "true"` pattern.
+ if [ "${AGENT_SHIN_ENABLED:-false}" = "true" ] && [ "${DISPATCH_CLOSE:-false}" = "true" ]; then
+ ARGS+=(--close)
+ echo "::notice::Agent Shin is ENABLED and running in close-on-fail mode."
+ elif [ "${AGENT_SHIN_ENABLED:-false}" = "true" ]; then
+ echo "::notice::Agent Shin is ENABLED but this trigger is dry-run (workflow_dispatch close != 'true')."
+ else
+ echo "::notice::Agent Shin is in DRY-RUN mode (AGENT_SHIN_ENABLED is not 'true'). No comments will be posted; no issues will be closed."
+ fi
+ # Automatic `issues` events stay dry-run regardless until the team
+ # explicitly invokes workflow_dispatch with close=true.
+ if [ "${GITHUB_EVENT_NAME:-}" = "issues" ]; then
+ # filter out --close rather than substituting to "" (which would
+ # leave an empty positional arg that argparse rejects)
+ FILTERED=()
+ for arg in "${ARGS[@]}"; do
+ if [ "${arg}" != "--close" ]; then
+ FILTERED+=("${arg}")
+ fi
+ done
+ ARGS=("${FILTERED[@]}")
+ echo "::notice::issues trigger -> forcing dry-run."
+ fi
+ python3 .github/scripts/triage_with_llm.py "${ARGS[@]}"
diff --git a/.github/workflows/triage_pr_with_llm.yml b/.github/workflows/triage_pr_with_llm.yml
new file mode 100644
index 00000000000..eac7e6a56b3
--- /dev/null
+++ b/.github/workflows/triage_pr_with_llm.yml
@@ -0,0 +1,104 @@
+name: Agent Shin — PR triage
+
+# LLM-as-judge triage for external pull requests.
+#
+# DRY-RUN BY DEFAULT. Closures and public comments are gated on the repo
+# variable `AGENT_SHIN_ENABLED` being set to the string `"true"`. Until then,
+# every run only writes its verdict to the workflow step summary so the team
+# can QA the judge's decisions before flipping it on.
+#
+# To enable for real:
+# 1. Add a repo secret `OPENAI_API_KEY` (or compatible).
+# 2. Set repo variable `AGENT_SHIN_ENABLED` to `true`
+# (Settings > Secrets and variables > Actions > Variables).
+#
+# We use `pull_request_target` so the workflow has access to repo secrets
+# and runs against PRs from forks. We never check out fork code — only read
+# PR metadata via `gh api`, so this is safe.
+
+on:
+ pull_request_target:
+ types: [opened, reopened]
+ workflow_dispatch:
+ inputs:
+ pr_number:
+ description: "PR number to triage manually."
+ required: true
+ close:
+ description: "If true and AGENT_SHIN_ENABLED=true, actually close on fail."
+ required: false
+ default: "false"
+ type: choice
+ options:
+ - "true"
+ - "false"
+
+permissions:
+ contents: read
+ issues: write
+ pull-requests: write
+
+jobs:
+ triage:
+ if: github.repository == 'BerriAI/litellm'
+ runs-on: ubuntu-latest
+ steps:
+ - name: Checkout triage script
+ uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+ with:
+ sparse-checkout: .github/scripts
+ persist-credentials: false
+
+ - name: Set up Python
+ uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+
+ - name: Install LLM client
+ run: pip install --no-cache-dir "openai>=1.40.0"
+
+ - name: Run Agent Shin
+ env:
+ GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+ OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+ OPENAI_BASE_URL: ${{ vars.OPENAI_BASE_URL }}
+ TRIAGE_MODEL: ${{ vars.TRIAGE_MODEL }}
+ AGENT_SHIN_ENABLED: ${{ vars.AGENT_SHIN_ENABLED }}
+ DISPATCH_CLOSE: ${{ github.event.inputs.close }}
+ PR_NUMBER: ${{ github.event.pull_request.number || github.event.inputs.pr_number }}
+ run: |
+ set -euo pipefail
+ ARGS=(--repo "${{ github.repository }}" --pr "${PR_NUMBER}")
+ # Fail-safe gating: only the EXACT string "true" enables the
+ # destructive --close path. The workflow_dispatch input is a
+ # `choice` dropdown of "true"/"false" so the UI is constrained,
+ # but the API (`gh workflow run -f close=...`) accepts any
+ # string, and a `!= "false"` check would treat "True", "yes",
+ # "1", "TRUE", typos, and accidental whitespace as enabling
+ # closure. Mirror the Greptile closer's `= "true"` pattern.
+ if [ "${AGENT_SHIN_ENABLED:-false}" = "true" ] && [ "${DISPATCH_CLOSE:-false}" = "true" ]; then
+ ARGS+=(--close)
+ echo "::notice::Agent Shin is ENABLED and running in close-on-fail mode."
+ elif [ "${AGENT_SHIN_ENABLED:-false}" = "true" ]; then
+ echo "::notice::Agent Shin is ENABLED but this trigger is dry-run (workflow_dispatch close != 'true' or scheduled event)."
+ else
+ echo "::notice::Agent Shin is in DRY-RUN mode (AGENT_SHIN_ENABLED is not 'true'). No comments will be posted; no PRs will be closed."
+ fi
+ # On the scheduled/automatic pull_request_target trigger we default to
+ # dry-run regardless, so the team can review verdicts in the step
+ # summary before any contributor sees a comment. Only the manual
+ # workflow_dispatch path (with close=true) closes PRs.
+ if [ "${GITHUB_EVENT_NAME:-}" = "pull_request_target" ]; then
+ # strip any --close added above (filter out, don't substitute
+ # to empty string — that would leave a stray "" positional arg
+ # that argparse rejects)
+ FILTERED=()
+ for arg in "${ARGS[@]}"; do
+ if [ "${arg}" != "--close" ]; then
+ FILTERED+=("${arg}")
+ fi
+ done
+ ARGS=("${FILTERED[@]}")
+ echo "::notice::pull_request_target trigger -> forcing dry-run."
+ fi
+ python3 .github/scripts/triage_with_llm.py "${ARGS[@]}"
diff --git a/.github/workflows/triage_reconsider.yml b/.github/workflows/triage_reconsider.yml
new file mode 100644
index 00000000000..5b23a8d5697
--- /dev/null
+++ b/.github/workflows/triage_reconsider.yml
@@ -0,0 +1,130 @@
+name: Agent Shin — reconsider
+
+# Comment-trigger workflow: when the PR/issue author (or an internal
+# collaborator) comments `@agent-shin reconsider` on a CLOSED PR/issue,
+# Agent Shin re-runs LLM-judge triage on the current title+body and:
+#
+# - on PASS: posts a "re-evaluated and reopened" comment + reopens.
+# - on FAIL: posts a "still missing X" comment and leaves it closed,
+# so the contributor can iterate again.
+#
+# This exists because GitHub does NOT let an external (non-write-access)
+# OSS contributor reopen a PR/issue closed by a bot or maintainer. Without
+# this comment trigger, a contributor whose PR Agent Shin auto-closed
+# would have no path back into the review queue except opening a fresh PR
+# (which loses the original PR's history). The bot, on the other hand,
+# has write access via GH_TOKEN and can reopen on their behalf.
+#
+# DRY-RUN BY DEFAULT — gated on `vars.AGENT_SHIN_ENABLED == 'true'` just
+# like the other Agent Shin workflows. The workflow also gates on the
+# commenter being either the PR/issue author or an internal collaborator
+# (OWNER/MEMBER/COLLABORATOR) so random commenters cannot DOS the LLM
+# judge or force a reopen.
+
+on:
+ issue_comment:
+ types: [created]
+
+permissions:
+ contents: read
+ issues: write
+ pull-requests: write
+
+jobs:
+ reconsider:
+ if: |
+ github.repository == 'BerriAI/litellm'
+ && contains(github.event.comment.body, '@agent-shin reconsider')
+ runs-on: ubuntu-latest
+ steps:
+ - name: Authorize commenter
+ # Only the PR/issue author OR an internal collaborator may trigger
+ # a reconsider. Outside random commenters could otherwise spam the
+ # phrase to burn LLM budget or, if a fail-open bug were ever
+ # introduced, force a reopen on someone else's behalf.
+ #
+ # We expose the authorization decision as a step output and gate
+ # every subsequent (potentially destructive) step on it. A `run:`
+ # step with `exit 0` would NOT stop the job — only `if:` gating
+ # on a known-true output is safe here.
+ id: auth
+ env:
+ COMMENTER: ${{ github.event.comment.user.login }}
+ AUTHOR: ${{ github.event.issue.user.login }}
+ ASSOCIATION: ${{ github.event.comment.author_association }}
+ run: |
+ set -euo pipefail
+ if [ "${COMMENTER}" = "${AUTHOR}" ]; then
+ echo "::notice::Authorized: commenter is the PR/issue author."
+ echo "authorized=true" >> "$GITHUB_OUTPUT"
+ exit 0
+ fi
+ case "${ASSOCIATION}" in
+ OWNER|MEMBER|COLLABORATOR)
+ echo "::notice::Authorized: commenter is an internal collaborator (${ASSOCIATION})."
+ echo "authorized=true" >> "$GITHUB_OUTPUT"
+ ;;
+ *)
+ echo "::notice::Commenter '${COMMENTER}' (${ASSOCIATION}) is not authorized to trigger reconsider; skipping subsequent steps."
+ echo "authorized=false" >> "$GITHUB_OUTPUT"
+ ;;
+ esac
+
+ - name: Checkout triage script
+ if: steps.auth.outputs.authorized == 'true'
+ uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
+ with:
+ sparse-checkout: .github/scripts
+ persist-credentials: false
+
+ - name: Set up Python
+ if: steps.auth.outputs.authorized == 'true'
+ uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
+ with:
+ python-version: "3.12"
+
+ - name: Install LLM client
+ if: steps.auth.outputs.authorized == 'true'
+ run: pip install --no-cache-dir "openai>=1.40.0"
+
+ - name: Run Agent Shin reconsider
+ if: steps.auth.outputs.authorized == 'true'
+ env:
+ GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+ OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+ OPENAI_BASE_URL: ${{ vars.OPENAI_BASE_URL }}
+ TRIAGE_MODEL: ${{ vars.TRIAGE_MODEL }}
+ AGENT_SHIN_ENABLED: ${{ vars.AGENT_SHIN_ENABLED }}
+ # `issue_comment` events fire for both issues and PR comments.
+ # `issue.pull_request` is set iff this is a PR comment, so we use
+ # its presence to decide whether to invoke `--pr N` or `--issue N`.
+ IS_PR: ${{ github.event.issue.pull_request != null }}
+ NUMBER: ${{ github.event.issue.number }}
+ run: |
+ set -euo pipefail
+ if [ "${IS_PR}" = "true" ]; then
+ ARGS=(--repo "${{ github.repository }}" --pr "${NUMBER}" --reconsider)
+ else
+ ARGS=(--repo "${{ github.repository }}" --issue "${NUMBER}" --reconsider)
+ fi
+ # Reconsider's destructive actions (post comment + reopen) are
+ # gated on `--close`, mirroring the regular triage workflows.
+ # When AGENT_SHIN_ENABLED is not the EXACT string "true", we
+ # still run the script so its verdict + would-X action lands in
+ # the step summary for QA — but without `--close`, the script
+ # returns `would-reopen` / `would-reconsider-still-failing`
+ # instead of touching GitHub state.
+ #
+ # Use the positive `= "true"` gate (not `!= "true" -> exit`) so
+ # the workflow guardrails in
+ # tests/test_litellm/test_github_triage_workflows.py see the
+ # canonical fail-safe enable pattern. Unknown values like
+ # "True", "yes", "1", or typos fall through to the dry-run
+ # branch, which is the safe default.
+ if [ "${AGENT_SHIN_ENABLED:-false}" = "true" ]; then
+ ARGS+=(--close)
+ echo "::notice::Agent Shin reconsider ENABLED — running real triage (close=true)."
+ else
+ echo "::notice::AGENT_SHIN_ENABLED is not 'true' -> reconsider stays in dry-run (no comment, no reopen)."
+ fi
+ python3 .github/scripts/triage_with_llm.py "${ARGS[@]}"
diff --git a/tests/test_litellm/test_github_close_low_quality_prs.py b/tests/test_litellm/test_github_close_low_quality_prs.py
new file mode 100644
index 00000000000..cda28036c75
--- /dev/null
+++ b/tests/test_litellm/test_github_close_low_quality_prs.py
@@ -0,0 +1,655 @@
+"""Unit tests for `.github/scripts/close_low_quality_prs.py`.
+
+These exercise the pure logic (score extraction and per-PR evaluation) without
+hitting GitHub. Network/CLI calls are stubbed via monkeypatch.
+"""
+
+from __future__ import annotations
+
+import datetime as dt
+import importlib.util
+import sys
+from pathlib import Path
+
+import pytest
+
+SCRIPT_PATH = (
+ Path(__file__).resolve().parents[2]
+ / ".github"
+ / "scripts"
+ / "close_low_quality_prs.py"
+)
+
+
+@pytest.fixture(scope="module")
+def closer_module():
+ """Load the script as a module via its file path (it lives outside the package)."""
+ spec = importlib.util.spec_from_file_location("close_low_quality_prs", SCRIPT_PATH)
+ assert spec and spec.loader, f"Could not load spec for {SCRIPT_PATH}"
+ module = importlib.util.module_from_spec(spec)
+ sys.modules["close_low_quality_prs"] = module
+ spec.loader.exec_module(module)
+ return module
+
+
+def _greptile_comment(
+ body: str,
+ updated_at: str = "2026-05-10T00:00:00Z",
+ login: str = "greptile-apps[bot]",
+) -> dict:
+ return {
+ "user": {"login": login},
+ "body": body,
+ "created_at": updated_at,
+ "updated_at": updated_at,
+ }
+
+
+class TestExtractGreptileScore:
+ def test_should_extract_score_from_html_header(self, closer_module):
+ comments = [
+ _greptile_comment("Confidence Score: 3/5
\nSome body text.")
+ ]
+ result = closer_module.extract_greptile_score(comments)
+ assert result is not None
+ score, _ = result
+ assert score == 3
+
+ def test_should_accept_both_greptile_login_variants(self, closer_module):
+ # REST API form ("greptile-apps[bot]") and GraphQL form ("greptile-apps")
+ for login in ("greptile-apps", "greptile-apps[bot]"):
+ comments = [
+ _greptile_comment("Confidence Score: 2/5
", login=login)
+ ]
+ result = closer_module.extract_greptile_score(comments)
+ assert result is not None, f"failed to detect score for login={login}"
+ score, _ = result
+ assert score == 2
+
+ def test_should_extract_score_from_plain_text(self, closer_module):
+ comments = [_greptile_comment("Confidence Score: 5/5 — looks good!")]
+ result = closer_module.extract_greptile_score(comments)
+ assert result is not None
+ score, _ = result
+ assert score == 5
+
+ def test_should_tolerate_whitespace_and_case(self, closer_module):
+ comments = [_greptile_comment("**confidence score : 2 / 5**")]
+ result = closer_module.extract_greptile_score(comments)
+ assert result is not None
+ score, _ = result
+ assert score == 2
+
+ def test_should_pick_most_recent_comment_when_rereview_happens(self, closer_module):
+ comments = [
+ _greptile_comment(
+ "Confidence Score: 2/5", updated_at="2026-05-01T00:00:00Z"
+ ),
+ _greptile_comment(
+ "Confidence Score: 5/5", updated_at="2026-05-12T00:00:00Z"
+ ),
+ ]
+ result = closer_module.extract_greptile_score(comments)
+ assert result is not None
+ score, _ = result
+ assert score == 5
+
+ def test_should_ignore_non_greptile_authors(self, closer_module):
+ comments = [
+ {
+ "user": {"login": "some-human"},
+ "body": "Confidence Score: 1/5",
+ "created_at": "2026-05-12T00:00:00Z",
+ "updated_at": "2026-05-12T00:00:00Z",
+ }
+ ]
+ assert closer_module.extract_greptile_score(comments) is None
+
+ def test_should_return_none_when_no_score_present(self, closer_module):
+ comments = [_greptile_comment("Greptile summary without a score.")]
+ assert closer_module.extract_greptile_score(comments) is None
+
+ def test_should_return_none_for_empty_comments(self, closer_module):
+ assert closer_module.extract_greptile_score([]) is None
+
+
+class TestEvaluatePr:
+ @pytest.fixture(autouse=True)
+ def _now(self):
+ return dt.datetime(2026, 5, 17, tzinfo=dt.timezone.utc)
+
+ def _make_pr(
+ self,
+ *,
+ number: int = 1,
+ created_days_ago: int = 10,
+ is_draft: bool = False,
+ labels: list[str] | None = None,
+ author_login: str = "someone",
+ ) -> dict:
+ created = dt.datetime(2026, 5, 17, tzinfo=dt.timezone.utc) - dt.timedelta(
+ days=created_days_ago
+ )
+ return {
+ "number": number,
+ "title": f"PR #{number}",
+ "createdAt": created.isoformat().replace("+00:00", "Z"),
+ "isDraft": is_draft,
+ "labels": [{"name": lbl} for lbl in (labels or [])],
+ "author": {"login": author_login},
+ "url": f"https://example.com/pr/{number}",
+ }
+
+ @pytest.fixture(autouse=True)
+ def _external_author(self, closer_module, monkeypatch):
+ """Treat every test PR as external unless overridden."""
+ monkeypatch.setattr(
+ closer_module, "is_external_pr_author", lambda pr, repo: True
+ )
+
+ def test_should_warn_drafts_when_score_low_first_time(
+ self, closer_module, _now, monkeypatch
+ ):
+ # Drafts are NOT a free pass — the open-PR queue should reflect any
+ # PR that needs human attention regardless of draft status. Authors
+ # who need a long-lived draft can use the `wip` opt-out label.
+ # First run: warn the contributor (1-day grace), don't close yet.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 2/5")],
+ )
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(is_draft=True, created_days_ago=0),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "warn-grace"
+ assert score == 2 and age == 0
+
+ def test_should_warn_brand_new_pr_when_min_age_zero(
+ self, closer_module, _now, monkeypatch
+ ):
+ # `min_age_days=0` means no age filter — a freshly-opened PR is
+ # eligible the moment Greptile scores it below threshold. The
+ # first detection still goes through the warn-grace step rather
+ # than closing immediately, giving the contributor 24 hours to
+ # respond before the next run actually closes the PR.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 1/5")],
+ )
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=0),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "warn-grace"
+ assert score == 1 and age == 0
+
+ def test_should_skip_optout_label_case_insensitive(
+ self, closer_module, _now, monkeypatch
+ ):
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: pytest.fail("should not fetch comments for opt-outs"),
+ )
+ action, _, _ = closer_module.evaluate_pr(
+ self._make_pr(labels=["WIP"]),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels={"wip"},
+ )
+ assert action == "skip-optout-label"
+
+ def test_should_skip_too_young_when_min_age_set(
+ self, closer_module, _now, monkeypatch
+ ):
+ # The min-age-days flag is now opt-in (default 0). When a maintainer
+ # explicitly passes a positive value (e.g. for a backfill run that
+ # wants to spare brand-new PRs), the skip-too-young path still works.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: pytest.fail("should not fetch comments for young PRs"),
+ )
+ action, _, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=2),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-too-young"
+ assert age == 2
+
+ def test_should_not_skip_when_min_age_is_zero(
+ self, closer_module, _now, monkeypatch
+ ):
+ # With the new default min_age_days=0, even a 0-day-old PR is
+ # evaluated. This test pins that behavior so future refactors don't
+ # silently restore an age filter.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 5/5")],
+ )
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=0),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-score-ok"
+ assert score == 5 and age == 0
+
+ def test_should_skip_when_greptile_has_not_reviewed(
+ self, closer_module, _now, monkeypatch
+ ):
+ monkeypatch.setattr(closer_module, "fetch_pr_comments", lambda *a, **kw: [])
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=10),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-no-greptile-score"
+ assert score is None and age == 10
+
+ def test_should_skip_when_score_meets_threshold(
+ self, closer_module, _now, monkeypatch
+ ):
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 4/5")],
+ )
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=10),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-score-ok"
+ assert score == 4 and age == 10
+
+ def test_should_warn_when_old_and_low_score_no_prior_warning(
+ self, closer_module, _now, monkeypatch
+ ):
+ # Even an old PR that still has no grace warning gets one on the
+ # first eligible run — the daily cron is the natural cadence, so
+ # an existing-but-never-warned PR enters the grace flow normally.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 3/5")],
+ )
+ action, score, age = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=10),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "warn-grace"
+ assert score == 3 and age == 10
+
+ def test_should_close_when_grace_warning_aged_out_and_score_still_low(
+ self, closer_module, _now, monkeypatch
+ ):
+ # Day-1 the closer posted a warning. Day-2 the PR still scores <4
+ # AND the warning is older than `GRACE_PERIOD_SECONDS`, so the
+ # action flips to `close`. This is the "grace expired" path.
+ old_warning = {
+ "user": {"login": "github-actions[bot]"},
+ "body": (
+ "you have 1 day to fix this\n\n" + closer_module.GRACE_COMMENT_MARKER
+ ),
+ "created_at": (
+ _now - dt.timedelta(seconds=closer_module.GRACE_PERIOD_SECONDS + 60)
+ )
+ .isoformat()
+ .replace("+00:00", "Z"),
+ "updated_at": "2026-05-15T00:00:00Z",
+ }
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [
+ _greptile_comment(
+ "Confidence Score: 1/5
",
+ updated_at="2026-05-15T00:00:00Z",
+ ),
+ old_warning,
+ ],
+ )
+ action, score, _ = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=14),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "close"
+ assert score == 1
+
+ def test_should_skip_when_grace_warning_within_window(
+ self, closer_module, _now, monkeypatch
+ ):
+ # Within the 24-hour grace window the closer must NOT close the
+ # PR even if the score is still low. The warning is only an hour
+ # old; give the contributor time to push fixes before destruction.
+ recent_warning = {
+ "user": {"login": "github-actions[bot]"},
+ "body": "warning text\n\n" + closer_module.GRACE_COMMENT_MARKER,
+ "created_at": (_now - dt.timedelta(hours=1))
+ .isoformat()
+ .replace("+00:00", "Z"),
+ }
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [
+ _greptile_comment("Confidence Score: 2/5"),
+ recent_warning,
+ ],
+ )
+ action, score, _ = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=10),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-in-grace-period"
+ assert score == 2
+
+ def test_should_close_immediately_for_swiftwinds_login(
+ self, closer_module, _now, monkeypatch
+ ):
+ # SwiftWinds is in `IMMEDIATE_CLOSE_LOGINS` for dogfooding the bot
+ # from an external account. Skip grace; close on first detection
+ # so the iteration loop is fast.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 1/5")],
+ )
+ action, score, _ = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=0, author_login="SwiftWinds"),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "close"
+ assert score == 1
+
+ def test_should_close_immediately_for_swiftwinds_login_case_insensitive(
+ self, closer_module, _now, monkeypatch
+ ):
+ # GitHub login matching is case-insensitive on GitHub's side; the
+ # API returns the original casing. Make sure the bypass fires
+ # regardless of how the login was registered.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: [_greptile_comment("Confidence Score: 1/5")],
+ )
+ for login in ("SwiftWinds", "swiftwinds", "SWIFTWINDS"):
+ action, _, _ = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=0, author_login=login),
+ now=_now,
+ min_age_days=0,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "close", login
+
+ def test_should_skip_internal_authors(self, closer_module, _now, monkeypatch):
+ # Override the fixture for this one test.
+ monkeypatch.setattr(
+ closer_module, "is_external_pr_author", lambda pr, repo: False
+ )
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_pr_comments",
+ lambda *a, **kw: pytest.fail("should not fetch comments for internal"),
+ )
+ action, score, _ = closer_module.evaluate_pr(
+ self._make_pr(created_days_ago=14),
+ now=_now,
+ min_age_days=7,
+ min_score=4,
+ repo=None,
+ optout_labels=set(),
+ )
+ assert action == "skip-internal"
+ assert score is None
+
+
+class TestMainOptoutLabelDefault:
+ """`--optout-label` must REPLACE the canonical defaults, not append."""
+
+ def _patch_no_op(self, closer_module, monkeypatch):
+ monkeypatch.setattr(closer_module, "fetch_open_prs", lambda repo: [])
+ # `optout_labels` is captured indirectly via evaluate_pr; sniff the
+ # set passed in by stubbing evaluate_pr.
+ captured: dict = {}
+
+ def fake_evaluate(pr, now, min_age_days, min_score, repo, optout_labels):
+ captured["optout_labels"] = set(optout_labels)
+ return ("skip-internal", None, None)
+
+ monkeypatch.setattr(closer_module, "evaluate_pr", fake_evaluate)
+ return captured
+
+ def test_should_use_canonical_defaults_when_flag_omitted(
+ self, closer_module, monkeypatch
+ ):
+ captured = self._patch_no_op(closer_module, monkeypatch)
+ # No PRs -> capture won't fire; instead inject one synthetic PR via
+ # fetch_open_prs so evaluate_pr is invoked at least once.
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_open_prs",
+ lambda repo: [
+ {
+ "number": 1,
+ "title": "p",
+ "createdAt": "2026-05-10T00:00:00Z",
+ "isDraft": True,
+ "labels": [],
+ "author": {"login": "x"},
+ }
+ ],
+ )
+ monkeypatch.setattr(sys, "argv", ["close_low_quality_prs.py"])
+ rc = closer_module.main()
+ assert rc == 0
+ assert captured["optout_labels"] == set(closer_module.DEFAULT_OPTOUT_LABELS)
+
+ def test_should_replace_defaults_when_flag_provided(
+ self, closer_module, monkeypatch
+ ):
+ captured = self._patch_no_op(closer_module, monkeypatch)
+ monkeypatch.setattr(
+ closer_module,
+ "fetch_open_prs",
+ lambda repo: [
+ {
+ "number": 1,
+ "title": "p",
+ "createdAt": "2026-05-10T00:00:00Z",
+ "isDraft": True,
+ "labels": [],
+ "author": {"login": "x"},
+ }
+ ],
+ )
+ monkeypatch.setattr(
+ sys,
+ "argv",
+ [
+ "close_low_quality_prs.py",
+ "--optout-label",
+ "hold",
+ "--optout-label",
+ "needs-discussion",
+ ],
+ )
+ rc = closer_module.main()
+ assert rc == 0
+ # Crucially, none of the canonical defaults leak in.
+ assert captured["optout_labels"] == {"hold", "needs-discussion"}
+ for default in closer_module.DEFAULT_OPTOUT_LABELS:
+ assert default not in captured["optout_labels"], default
+
+
+class TestSecondsSinceLastGraceWarning:
+ """Grace-period detection: only counts comments by the bot identity
+ that contain the shared `GRACE_COMMENT_MARKER`."""
+
+ def _make_marker_comment(
+ self,
+ closer_module,
+ *,
+ login: str = "github-actions[bot]",
+ created_at: str = "2026-05-16T00:00:00Z",
+ include_marker: bool = True,
+ ) -> dict:
+ body = "warning text"
+ if include_marker:
+ body += "\n\n" + closer_module.GRACE_COMMENT_MARKER
+ return {
+ "user": {"login": login},
+ "body": body,
+ "created_at": created_at,
+ }
+
+ def test_should_return_none_when_no_marker_comment(self, closer_module):
+ comments = [
+ {
+ "user": {"login": "github-actions[bot]"},
+ "body": "Some other bot comment",
+ "created_at": "2026-05-16T00:00:00Z",
+ }
+ ]
+ assert closer_module.seconds_since_last_grace_warning(comments) is None
+
+ def test_should_return_none_for_empty(self, closer_module):
+ assert closer_module.seconds_since_last_grace_warning([]) is None
+
+ def test_should_ignore_non_bot_comments_with_marker(self, closer_module):
+ # If a curious user quotes the marker in a comment, we must NOT
+ # treat it as a bot warning. The grace timer would then never fire.
+ comments = [
+ self._make_marker_comment(closer_module, login="random-user"),
+ ]
+ assert closer_module.seconds_since_last_grace_warning(comments) is None
+
+ def test_should_pick_latest_marker_comment(self, closer_module):
+ # When multiple grace warnings exist (e.g. a re-open cycle), use
+ # the most recent one to compute the age.
+ comments = [
+ self._make_marker_comment(closer_module, created_at="2026-05-15T00:00:00Z"),
+ self._make_marker_comment(closer_module, created_at="2026-05-16T23:00:00Z"),
+ ]
+ now = dt.datetime(2026, 5, 17, 0, 0, 0, tzinfo=dt.timezone.utc)
+ age = closer_module.seconds_since_last_grace_warning(comments, now=now)
+ # 1h = 3600s
+ assert age == 3600.0
+
+
+class TestImmediateCloseLoginsConstant:
+ """SwiftWinds is the dogfood account the user explicitly named — pin
+ its presence so a future cleanup that removes the constant or
+ forgets to keep the entry doesn't silently break the test path."""
+
+ def test_should_include_swiftwinds(self, closer_module):
+ assert "swiftwinds" in closer_module.IMMEDIATE_CLOSE_LOGINS
+
+ def test_should_be_lowercase_for_case_insensitive_match(self, closer_module):
+ for login in closer_module.IMMEDIATE_CLOSE_LOGINS:
+ assert login == login.lower(), login
+
+
+class TestGraceWarningCommentText:
+ """Pin the user-facing language in the grace warning comment so the
+ `1 day grace` and `@greptileai still works after close` promises
+ don't get accidentally dropped in a future refactor.
+ """
+
+ def test_should_state_one_day_grace_period(self, closer_module):
+ body = closer_module.format_grace_warning_comment(score=2, threshold=4)
+ # The user's PR explicitly said "specify in the comment" — pin
+ # that the literal "1 day" appears in the comment.
+ assert "1 day" in body
+
+ def test_should_mention_agent_shin_reconsider(self, closer_module):
+ body = closer_module.format_grace_warning_comment(score=2, threshold=4)
+ assert "@agent-shin reconsider" in body
+
+ def test_should_promise_greptileai_works_after_close(self, closer_module):
+ body = closer_module.format_grace_warning_comment(score=2, threshold=4)
+ assert "@greptileai" in body
+ assert "even after the PR is closed" in body
+
+ def test_should_carry_grace_marker(self, closer_module):
+ # The marker is what `seconds_since_last_grace_warning` greps for
+ # to detect a prior warning — dropping it would silently break
+ # the cooldown.
+ body = closer_module.format_grace_warning_comment(score=2, threshold=4)
+ assert closer_module.GRACE_COMMENT_MARKER in body
+
+ def test_close_comment_should_mention_greptileai_post_close(self, closer_module):
+ # The actual close comment should ALSO point at the @greptileai
+ # post-close re-review path so contributors see the same options
+ # whether they read the warning or only catch the close comment.
+ # `close_pr` writes the close comment via `gh pr comment` — we
+ # don't easily call it directly here, but the comment body is
+ # constructed inline. Re-creating it via a no-op `gh` stub is
+ # awkward, so we assert against the same string template by
+ # asserting that the close path's text constant is updated.
+ # `close_pr` source must contain the marker text — guarded by
+ # this whole-module read-and-assert.
+ from pathlib import Path
+
+ source = Path(closer_module.__file__).read_text()
+ assert "even after the PR is closed" in source
+ assert "@greptileai" in source
+
+
+class TestHasOptoutLabel:
+ def test_should_match_label_case_insensitively(self, closer_module):
+ pr = {"labels": [{"name": "Do Not Close"}, {"name": "bug"}]}
+ assert closer_module.has_optout_label(pr, {"do not close"}) is True
+
+ def test_should_return_false_when_no_match(self, closer_module):
+ pr = {"labels": [{"name": "bug"}, {"name": "enhancement"}]}
+ assert closer_module.has_optout_label(pr, {"wip", "keep open"}) is False
+
+ def test_should_handle_missing_labels(self, closer_module):
+ assert closer_module.has_optout_label({}, {"wip"}) is False
diff --git a/tests/test_litellm/test_github_triage_with_llm.py b/tests/test_litellm/test_github_triage_with_llm.py
new file mode 100644
index 00000000000..020de81c391
--- /dev/null
+++ b/tests/test_litellm/test_github_triage_with_llm.py
@@ -0,0 +1,1715 @@
+"""Unit tests for `.github/scripts/triage_with_llm.py` (Agent Shin)."""
+
+from __future__ import annotations
+
+import importlib.util
+import json
+import sys
+from pathlib import Path
+
+import pytest
+
+SCRIPT_PATH = (
+ Path(__file__).resolve().parents[2] / ".github" / "scripts" / "triage_with_llm.py"
+)
+
+
+@pytest.fixture(scope="module")
+def triage_module():
+ spec = importlib.util.spec_from_file_location("triage_with_llm", SCRIPT_PATH)
+ assert spec and spec.loader
+ module = importlib.util.module_from_spec(spec)
+ sys.modules["triage_with_llm"] = module
+ spec.loader.exec_module(module)
+ return module
+
+
+class TestIsInternalContributor:
+ @pytest.mark.parametrize("association", ["OWNER", "MEMBER", "COLLABORATOR"])
+ def test_should_mark_org_associations_as_internal(self, triage_module, association):
+ item = {
+ "author_association": association,
+ "user": {"login": "krrishdholakia"},
+ }
+ assert triage_module.is_internal_contributor(item) is True
+
+ @pytest.mark.parametrize(
+ "association",
+ ["CONTRIBUTOR", "FIRST_TIME_CONTRIBUTOR", "FIRST_TIMER", "NONE"],
+ )
+ def test_should_mark_outside_associations_as_external(
+ self, triage_module, association
+ ):
+ item = {
+ "author_association": association,
+ "user": {"login": "random-oss-dev"},
+ }
+ assert triage_module.is_internal_contributor(item) is False
+
+ @pytest.mark.parametrize(
+ "item",
+ [
+ {"author_association": "", "user": {"login": "random-oss-dev"}},
+ {"user": {"login": "random-oss-dev"}}, # association field absent
+ ],
+ )
+ def test_should_fail_safe_when_author_association_is_missing(
+ self, triage_module, item
+ ):
+ # Fail-safe: an empty/missing association must never make a PR
+ # eligible for the destructive close path. Treat as internal (skip).
+ assert triage_module.is_internal_contributor(item) is True
+
+ @pytest.mark.parametrize(
+ "login",
+ ["dependabot[bot]", "greptile-apps[bot]", "dependabot", "github-actions"],
+ )
+ def test_should_skip_bot_accounts_regardless_of_association(
+ self, triage_module, login
+ ):
+ item = {"author_association": "NONE", "user": {"login": login}}
+ assert triage_module.is_internal_contributor(item) is True
+
+
+class TestHasLinkedIssue:
+ @pytest.mark.parametrize(
+ "body",
+ [
+ "Fixes #1234",
+ "closes #1",
+ "Resolves #99",
+ "fix #42 — this addresses the regression",
+ "Closes https://github.com/BerriAI/litellm/issues/27000",
+ "Resolved https://github.com/BerriAI/litellm/issues/27001",
+ ],
+ )
+ def test_should_detect_common_link_phrases(self, triage_module, body):
+ assert triage_module.has_linked_issue(body) is True
+
+ @pytest.mark.parametrize(
+ "body",
+ [
+ "",
+ "Some change",
+ # Casual mentions must NOT auto-pass — they should fall through to
+ # the LLM judge so the stricter "not a passing mention" rule fires.
+ "See #1234",
+ "see #1234 for context",
+ "ref #1234",
+ "Refs https://github.com/BerriAI/litellm/issues/27000",
+ "this addresses #1234",
+ ],
+ )
+ def test_should_not_auto_pass_casual_mentions(self, triage_module, body):
+ assert triage_module.has_linked_issue(body) is False
+
+ def test_should_not_detect_when_only_html_comment_template(self, triage_module):
+ body = ""
+ assert triage_module.has_linked_issue(body) is False
+
+
+class TestStripHtmlComments:
+ def test_should_remove_single_line_comments(self, triage_module):
+ text = "before after"
+ assert "placeholder" not in triage_module.strip_html_comments(text)
+
+ def test_should_remove_multiline_comments(self, triage_module):
+ text = "kept\n\nkept2"
+ cleaned = triage_module.strip_html_comments(text)
+ assert "Fixes #1" not in cleaned
+ assert "kept" in cleaned and "kept2" in cleaned
+
+ def test_should_handle_none(self, triage_module):
+ assert triage_module.strip_html_comments(None) == ""
+
+
+class TestCloseCommentText:
+ """Pin the user-facing language in close comments so changes are intentional."""
+
+ def test_pr_close_comment_should_recommend_new_pr_primarily(self, triage_module):
+ body = triage_module.format_pr_close_comment(
+ {"verdict": "fail", "missing": ["QA proof"], "explanation": "thin"}
+ )
+ # Primary path: open a new PR (because OSS authors can't reopen a
+ # bot-closed PR). Secondary path: `@agent-shin reconsider`.
+ assert "Open a new PR" in body
+ assert "@agent-shin reconsider" in body
+ # Old advice that no longer works for OSS contributors must NOT
+ # appear (they can't reopen a PR closed by a bot/maintainer).
+ assert "Reopen the PR" not in body
+
+ def test_reopen_comment_should_carry_reconsider_marker(self, triage_module):
+ # The marker is what the rate-limit guard greps for to detect a
+ # prior reconsider verdict on the same PR. If the marker ever
+ # gets dropped from this comment, the cooldown silently breaks
+ # and a contributor can spam `@agent-shin reconsider` to burn
+ # LLM budget.
+ body = triage_module.format_reopen_comment("pr")
+ assert triage_module.RECONSIDER_COMMENT_MARKER in body
+
+ def test_still_failing_comment_should_carry_reconsider_marker(self, triage_module):
+ body = triage_module.format_reconsider_still_failing_comment(
+ "pr",
+ {"verdict": "fail", "missing": ["QA proof"], "explanation": "thin"},
+ )
+ assert triage_module.RECONSIDER_COMMENT_MARKER in body
+
+ def test_pr_close_comment_should_not_promise_automatic_reopen_on_open(
+ self, triage_module
+ ):
+ # The previous comment said "I'll re-evaluate automatically" — that
+ # only worked because the author could reopen, which they often
+ # can't. The new wording must point them at the comment trigger or
+ # a new PR instead.
+ body = triage_module.format_pr_close_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ assert "I'll re-evaluate automatically" not in body
+
+ def test_issue_close_comment_should_use_reconsider_trigger(self, triage_module):
+ body = triage_module.format_issue_close_comment(
+ {"verdict": "fail", "missing": ["repro"], "explanation": "thin"}
+ )
+ assert "@agent-shin reconsider" in body
+ assert "Reopen the issue" not in body
+
+
+class TestWasClosedByAgentShin:
+ """Bot-closed guard: only the bot's own closures are reopen candidates."""
+
+ def test_should_return_true_when_last_close_actor_is_bot(
+ self, triage_module, monkeypatch
+ ):
+ monkeypatch.setattr(
+ triage_module,
+ "fetch_last_close_actor",
+ lambda repo, n: "github-actions[bot]",
+ )
+ assert triage_module.was_closed_by_agent_shin("o/r", 1) is True
+
+ def test_should_return_false_when_last_close_actor_is_maintainer(
+ self, triage_module, monkeypatch
+ ):
+ # A maintainer closed it (e.g. duplicate, security, design). The
+ # bot must refuse to reopen on @agent-shin reconsider.
+ monkeypatch.setattr(
+ triage_module,
+ "fetch_last_close_actor",
+ lambda repo, n: "krrishdholakia",
+ )
+ assert triage_module.was_closed_by_agent_shin("o/r", 1) is False
+
+ def test_should_fail_closed_when_no_close_event(self, triage_module, monkeypatch):
+ # If the events API returns nothing (network blip, repo permission
+ # quirk), the guard must fail-closed: refuse to reopen rather than
+ # assume the bot did it.
+ monkeypatch.setattr(
+ triage_module, "fetch_last_close_actor", lambda repo, n: None
+ )
+ assert triage_module.was_closed_by_agent_shin("o/r", 1) is False
+
+ def test_should_respect_bot_login_override_via_env(
+ self, triage_module, monkeypatch
+ ):
+ # Operators wiring Agent Shin to a PAT (instead of GITHUB_TOKEN)
+ # can override the expected bot login via env. The guard must
+ # respect the override so non-default deployments still work.
+ monkeypatch.setenv("AGENT_SHIN_BOT_LOGIN", "my-bot")
+ monkeypatch.setattr(
+ triage_module, "fetch_last_close_actor", lambda repo, n: "my-bot"
+ )
+ assert triage_module.was_closed_by_agent_shin("o/r", 1) is True
+ # Default "github-actions[bot]" should NOT match when env is set.
+ monkeypatch.setattr(
+ triage_module,
+ "fetch_last_close_actor",
+ lambda repo, n: "github-actions[bot]",
+ )
+ assert triage_module.was_closed_by_agent_shin("o/r", 1) is False
+
+
+class TestSecondsSinceLastReconsiderVerdict:
+ """Rate-limit guard: detects the bot's own reconsider verdict marker."""
+
+ def _make_comment(
+ self, *, login: str, body: str, created_at: str | None = "2026-05-18T05:00:00Z"
+ ) -> dict:
+ comment: dict = {"user": {"login": login}, "body": body}
+ if created_at is not None:
+ comment["created_at"] = created_at
+ return comment
+
+ def test_should_return_none_when_no_bot_reconsider_comments(
+ self, triage_module, monkeypatch
+ ):
+ # An issue with chatter from other users but no bot reconsider
+ # verdict must not be rate-limited.
+ comments = [
+ self._make_comment(login="outside-dev", body="ping?"),
+ self._make_comment(
+ login="github-actions[bot]", body="some other bot message"
+ ),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+ assert triage_module.seconds_since_last_reconsider_verdict("o/r", 1) is None
+
+ def test_should_pick_latest_bot_reconsider_marker(self, triage_module, monkeypatch):
+ # When multiple reconsider verdicts exist, return the AGE of the
+ # most recent one. Using a frozen reference helps pin the math.
+ comments = [
+ self._make_comment(
+ login="github-actions[bot]",
+ body="old verdict " + triage_module.RECONSIDER_COMMENT_MARKER,
+ created_at="2026-05-18T04:00:00Z",
+ ),
+ self._make_comment(
+ login="github-actions[bot]",
+ body="newer verdict " + triage_module.RECONSIDER_COMMENT_MARKER,
+ created_at="2026-05-18T04:55:00Z",
+ ),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+
+ # Freeze "now" via a tiny shim on the module's `dt` import.
+ import datetime as real_dt
+
+ class FrozenDateTime(real_dt.datetime):
+ @classmethod
+ def now(cls, tz=None):
+ return real_dt.datetime(2026, 5, 18, 5, 0, 0, tzinfo=tz)
+
+ frozen_module = type(triage_module.dt)("datetime")
+ frozen_module.datetime = FrozenDateTime
+ frozen_module.timezone = real_dt.timezone
+ monkeypatch.setattr(triage_module, "dt", frozen_module)
+
+ age = triage_module.seconds_since_last_reconsider_verdict("o/r", 1)
+ # newer verdict is 5 minutes (300 seconds) before "now"
+ assert age == 300.0
+
+ def test_should_ignore_non_bot_comments_with_marker(
+ self, triage_module, monkeypatch
+ ):
+ # A user comment that happens to quote the marker (e.g. in
+ # a "what does this hidden marker do?" question) must NOT count.
+ # The rate-limit guard only trusts comments authored by the bot.
+ comments = [
+ self._make_comment(
+ login="curious-user",
+ body=f"Saw this marker: {triage_module.RECONSIDER_COMMENT_MARKER}",
+ ),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+ assert triage_module.seconds_since_last_reconsider_verdict("o/r", 1) is None
+
+ def test_should_ignore_bot_comments_without_marker(
+ self, triage_module, monkeypatch
+ ):
+ # The bot posts other things too (Agent Shin close comments,
+ # CI status, etc.) — only the reconsider-verdict marker should
+ # arm the cooldown.
+ comments = [
+ self._make_comment(
+ login="github-actions[bot]",
+ body="Agent Shin closed this PR (no marker)",
+ ),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+ assert triage_module.seconds_since_last_reconsider_verdict("o/r", 1) is None
+
+
+class TestParseVerdict:
+ def test_should_parse_plain_json(self, triage_module):
+ raw = '{"verdict": "pass", "missing": []}'
+ assert triage_module.parse_verdict(raw)["verdict"] == "pass"
+
+ def test_should_strip_markdown_fence(self, triage_module):
+ raw = '```json\n{"verdict": "fail", "missing": ["foo"]}\n```'
+ result = triage_module.parse_verdict(raw)
+ assert result["verdict"] == "fail"
+ assert result["missing"] == ["foo"]
+
+ def test_should_extract_embedded_json_from_prose(self, triage_module):
+ raw = 'Here you go: {"verdict": "pass", "missing": []}\nThanks.'
+ assert triage_module.parse_verdict(raw)["verdict"] == "pass"
+
+ def test_should_raise_for_unparseable_text(self, triage_module):
+ with pytest.raises(ValueError):
+ triage_module.parse_verdict("not even close to json")
+
+ def test_should_raise_for_empty(self, triage_module):
+ with pytest.raises(ValueError):
+ triage_module.parse_verdict("")
+
+
+class TestBuildPrompts:
+ def test_should_include_pr_title_and_body(self, triage_module):
+ prompt = triage_module.build_pr_prompt(
+ title="Add foo", body=" Real body"
+ )
+ assert "Add foo" in prompt
+ assert "Real body" in prompt
+ assert "comment" not in prompt # HTML comments are stripped
+
+ def test_should_show_empty_marker_for_empty_pr_body(self, triage_module):
+ prompt = triage_module.build_pr_prompt(title="t", body="")
+ assert "(empty)" in prompt
+
+ def test_should_include_issue_title_and_body(self, triage_module):
+ prompt = triage_module.build_issue_prompt(title="Bug", body="repro here")
+ assert "Bug" in prompt
+ assert "repro here" in prompt
+
+ def test_should_not_crash_when_pr_body_contains_curly_braces(self, triage_module):
+ """User-supplied content with `{` / `}` must NOT be re-parsed by
+ `str.format()`. `format` only scans the template literal for
+ replacement fields; values being substituted in are inserted as
+ plain strings, so a body like `{"foo": "bar"}` or `{unmatched`
+ cannot blow up the script. Pinning this here so a future
+ "improvement" to the templating doesn't reintroduce a crash on
+ every PR that quotes JSON.
+ """
+ for body in (
+ 'Here is some JSON: {"foo": "bar", "n": 1}',
+ "Half a brace { left dangling, and a stray }",
+ "Format-spec-looking thing: {0}, {name:>10}, {!r}",
+ "Nested {a: {b: c}} braces",
+ ):
+ pr_prompt = triage_module.build_pr_prompt(title="t", body=body)
+ issue_prompt = triage_module.build_issue_prompt(title="t", body=body)
+ assert body in pr_prompt
+ assert body in issue_prompt
+
+ def test_should_not_crash_when_pr_title_contains_curly_braces(self, triage_module):
+ title = "Fix bug in {0:>10} format-spec handling"
+ pr_prompt = triage_module.build_pr_prompt(title=title, body="x")
+ issue_prompt = triage_module.build_issue_prompt(title=title, body="x")
+ assert title in pr_prompt
+ assert title in issue_prompt
+
+ def test_should_preserve_template_indentation_with_multiline_body(
+ self, triage_module
+ ):
+ """`textwrap.dedent` runs on the static template *before* user
+ content is interpolated, so a multi-line body (whose 2nd+ lines
+ start at column 0) cannot defeat the common-indent computation
+ and leave 8-space indentation on every template line. Pin the
+ dedented shape so the rendered prompt stays consistent for the
+ LLM judge.
+ """
+ body = "first line\nsecond line at column 0\nthird line at column 0"
+ for builder in (
+ triage_module.build_pr_prompt,
+ triage_module.build_issue_prompt,
+ ):
+ prompt = builder(title="t", body=body)
+ # Template lines should NOT carry the 8 leading spaces from
+ # the source-file indentation of the triple-quoted string.
+ assert " You are " not in prompt
+ assert 'You are "Agent Shin"' in prompt
+ assert body in prompt
+
+
+class TestMainModelDefault:
+ """`--model` falls back to DEFAULT_MODEL even when TRIAGE_MODEL is empty."""
+
+ def _stub_triage(self, triage_module, monkeypatch):
+ captured: dict = {}
+
+ def fake_triage(**kwargs):
+ captured.update(kwargs)
+ return {
+ "kind": kwargs["kind"],
+ "number": kwargs["number"],
+ "title": "",
+ "author": "x",
+ "author_association": "NONE",
+ "state": "open",
+ "action": "skip-no-llm-key",
+ }
+
+ monkeypatch.setattr(triage_module, "triage", fake_triage)
+ return captured
+
+ def test_should_fall_back_to_default_when_triage_model_env_empty(
+ self, triage_module, monkeypatch
+ ):
+ captured = self._stub_triage(triage_module, monkeypatch)
+ monkeypatch.setenv("TRIAGE_MODEL", "")
+ monkeypatch.setattr(
+ sys,
+ "argv",
+ ["triage_with_llm.py", "--repo", "o/r", "--pr", "1"],
+ )
+ rc = triage_module.main()
+ assert rc == 0
+ assert captured["model"] == triage_module.DEFAULT_MODEL
+
+ def test_should_respect_explicit_triage_model_env(self, triage_module, monkeypatch):
+ captured = self._stub_triage(triage_module, monkeypatch)
+ monkeypatch.setenv("TRIAGE_MODEL", "gpt-4o-mini")
+ monkeypatch.setattr(
+ sys,
+ "argv",
+ ["triage_with_llm.py", "--repo", "o/r", "--pr", "1"],
+ )
+ rc = triage_module.main()
+ assert rc == 0
+ assert captured["model"] == "gpt-4o-mini"
+
+
+class TestCallLlmJudge:
+ """call_llm_judge sets gpt-5 specific kwargs correctly."""
+
+ def _stub_openai(self, monkeypatch, captured: dict):
+ """Install a fake `openai.OpenAI` client into sys.modules.
+
+ The fake client records the kwargs passed to chat.completions.create
+ and returns a minimal response object whose .choices[0].message.content
+ is "ok".
+ """
+ import types
+
+ class FakeMessage:
+ content = '{"verdict": "pass"}'
+
+ class FakeChoice:
+ message = FakeMessage()
+
+ class FakeResponse:
+ choices = [FakeChoice()]
+
+ class FakeCompletions:
+ def create(self, **kwargs):
+ captured.update(kwargs)
+ return FakeResponse()
+
+ class FakeChat:
+ completions = FakeCompletions()
+
+ class FakeClient:
+ def __init__(self, api_key, base_url=None):
+ captured["__client_kwargs__"] = {
+ "api_key": api_key,
+ "base_url": base_url,
+ }
+ self.chat = FakeChat()
+
+ fake_module = types.ModuleType("openai")
+ fake_module.OpenAI = FakeClient
+ monkeypatch.setitem(sys.modules, "openai", fake_module)
+
+ def test_should_set_reasoning_effort_none_for_gpt5_family(
+ self, triage_module, monkeypatch
+ ):
+ captured: dict = {}
+ self._stub_openai(monkeypatch, captured)
+ triage_module.call_llm_judge(
+ "prompt", model="gpt-5.4-mini", api_key="sk-test", base_url=None
+ )
+ assert captured["model"] == "gpt-5.4-mini"
+ assert captured["temperature"] == 0
+ assert captured["extra_body"] == {"reasoning_effort": "none"}
+
+ def test_should_set_reasoning_effort_for_capitalized_or_dated_gpt5(
+ self, triage_module, monkeypatch
+ ):
+ for model in ("GPT-5.4-mini", "gpt-5.4-mini-2026-03-17", "gpt-5"):
+ captured: dict = {}
+ self._stub_openai(monkeypatch, captured)
+ triage_module.call_llm_judge(
+ "prompt", model=model, api_key="sk-test", base_url=None
+ )
+ assert captured["extra_body"] == {"reasoning_effort": "none"}, model
+
+ def test_should_omit_reasoning_effort_for_non_gpt5(
+ self, triage_module, monkeypatch
+ ):
+ captured: dict = {}
+ self._stub_openai(monkeypatch, captured)
+ triage_module.call_llm_judge(
+ "prompt", model="gpt-4o-mini", api_key="sk-test", base_url=None
+ )
+ assert "extra_body" not in captured
+
+ def test_should_pass_base_url_when_provided(self, triage_module, monkeypatch):
+ captured: dict = {}
+ self._stub_openai(monkeypatch, captured)
+ triage_module.call_llm_judge(
+ "p",
+ model="gpt-5.4-mini",
+ api_key="sk-test",
+ base_url="https://proxy.example.com/v1",
+ )
+ assert (
+ captured["__client_kwargs__"]["base_url"] == "https://proxy.example.com/v1"
+ )
+
+
+class TestTriageOrchestration:
+ """End-to-end-ish tests that mock both gh fetchers and the LLM."""
+
+ def _make_pr(self, **overrides):
+ base = {
+ "number": 1,
+ "title": "PR title",
+ "body": "PR body",
+ "state": "open",
+ "author_association": "NONE",
+ "user": {"login": "outside-dev"},
+ }
+ base.update(overrides)
+ return base
+
+ def test_should_skip_internal_author(self, triage_module, monkeypatch):
+ pr = self._make_pr(
+ author_association="MEMBER", user={"login": "krrishdholakia"}
+ )
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+
+ def boom(*a, **kw):
+ pytest.fail("LLM should not be called for internal authors")
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=boom,
+ )
+ assert result["action"] == "skip-internal-author"
+
+ def test_should_skip_closed_pr(self, triage_module, monkeypatch):
+ pr = self._make_pr(state="closed")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: pytest.fail("should not run on closed PRs"),
+ )
+ assert result["action"] == "skip-not-open"
+
+ def test_should_short_circuit_on_linked_issue(self, triage_module, monkeypatch):
+ pr = self._make_pr(body="Fixes #1234\n\nFoo bar")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: pytest.fail("LLM should not be called"),
+ )
+ assert result["action"] == "pass-linked-issue"
+ assert result["verdict"]["verdict"] == "pass"
+
+ def test_should_not_short_circuit_on_casual_mention(
+ self, triage_module, monkeypatch
+ ):
+ # "See #1234" is a passing mention, not a closing keyword. The LLM
+ # must get a chance to apply the stricter rubric. With no prior
+ # grace warning, the first failing verdict triggers the warning
+ # path (`would-warn-grace` in dry-run).
+ pr = self._make_pr(body="See #1234 for context. No QA proof here.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_no_warning(triage_module, monkeypatch)
+ called = {"judge": False}
+
+ def judge(prompt):
+ called["judge"] = True
+ return json.dumps(
+ {"verdict": "fail", "missing": ["QA proof"], "explanation": "thin."}
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=judge,
+ )
+ assert called["judge"] is True
+ assert result["action"] == "would-warn-grace"
+
+ def test_should_return_pass_llm_when_judge_passes(self, triage_module, monkeypatch):
+ pr = self._make_pr(body="Long body, no linked issue.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ captured = {}
+
+ def judge(prompt):
+ captured["prompt"] = prompt
+ return json.dumps({"verdict": "pass", "missing": [], "explanation": "ok"})
+
+ result = triage_module.triage(
+ repo="o/r", kind="pr", number=1, close=True, model="m", judge=judge
+ )
+ assert result["action"] == "pass-llm"
+ assert "Long body" in captured["prompt"]
+
+ def test_should_return_would_close_in_dry_run_after_grace_aged_out(
+ self, triage_module, monkeypatch
+ ):
+ # When the grace warning has already aged out (>= GRACE_PERIOD_SECONDS)
+ # AND the rubric still fails, the dry-run preview returns
+ # `would-close` so a step-summary writer can render the close
+ # comment without touching GitHub state.
+ pr = self._make_pr(body="just a sentence.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_aged_out(triage_module, monkeypatch)
+
+ def fake_post(*a, **kw):
+ pytest.fail("should not post comments in dry-run")
+
+ def fake_close(*a, **kw):
+ pytest.fail("should not close in dry-run")
+
+ monkeypatch.setattr(triage_module, "post_comment", fake_post)
+ monkeypatch.setattr(triage_module, "close_pr", fake_close)
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["problem description", "QA proof"],
+ "explanation": "Body is one sentence.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "would-close"
+ assert result["verdict"]["missing"] == ["problem description", "QA proof"]
+
+ def test_should_post_comment_and_close_after_grace_window(
+ self, triage_module, monkeypatch
+ ):
+ # The "real close" path: --close passed AND the grace warning has
+ # aged out AND the rubric still fails. The bot posts the close
+ # comment and closes the PR.
+ pr = self._make_pr(body="just a sentence.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_aged_out(triage_module, monkeypatch)
+ posted = {}
+ closed = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"repo": repo, "n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda repo, n: closed.update({"repo": repo, "n": n}),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Body too thin.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "closed"
+ assert posted["n"] == 42 and closed["n"] == 42
+ assert "Agent Shin" in posted["body"]
+ assert "QA proof" in posted["body"]
+
+ def test_should_skip_on_llm_error_in_close_mode(self, triage_module, monkeypatch):
+ pr = self._make_pr(body="something.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not comment on LLM error"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close on LLM error"),
+ )
+
+ def broken_judge(prompt):
+ raise RuntimeError("upstream 500")
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=broken_judge,
+ )
+ assert result["action"] == "skip-llm-error"
+ assert "upstream 500" in result["error"]
+
+ def test_should_skip_open_pr_in_reconsider_mode(self, triage_module, monkeypatch):
+ # Reconsider only makes sense on a CLOSED PR — running it on an open
+ # one is a no-op (the regular triage flow already evaluated it).
+ pr = self._make_pr(state="open")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=lambda p: pytest.fail("should not run on open PR in reconsider"),
+ reconsider=True,
+ )
+ assert result["action"] == "skip-not-closed"
+
+ @staticmethod
+ def _stub_reconsider_guards(triage_module, monkeypatch):
+ """Default reconsider-guard stubs: pretend bot closed + no cooldown.
+
+ The new safety guards (`was_closed_by_agent_shin`,
+ `seconds_since_last_reconsider_verdict`) hit the GitHub API in
+ production. Tests that exercise the reconsider happy path stub
+ them to "yes the bot closed it, no recent reconsider comment"
+ so the test stays focused on its actual assertion.
+ """
+ monkeypatch.setattr(
+ triage_module, "was_closed_by_agent_shin", lambda *a, **kw: True
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_reconsider_verdict",
+ lambda *a, **kw: None,
+ )
+
+ @staticmethod
+ def _stub_grace_aged_out(triage_module, monkeypatch):
+ """Pretend the grace warning has aged out.
+
+ For tests that exercise the post-grace close path. Set the age
+ to twice the grace window so a future tweak to
+ `GRACE_PERIOD_SECONDS` doesn't accidentally make the stub fall
+ back inside the window.
+ """
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_grace_warning",
+ lambda *a, **kw: triage_module.GRACE_PERIOD_SECONDS * 2,
+ )
+
+ @staticmethod
+ def _stub_grace_no_warning(triage_module, monkeypatch):
+ """Pretend no grace warning has been posted yet (first detection)."""
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_grace_warning",
+ lambda *a, **kw: None,
+ )
+
+ def test_should_reopen_on_reconsider_pass(self, triage_module, monkeypatch):
+ # Reconsider on a closed PR with a passing verdict -> reopen + post a
+ # friendly "re-evaluated" comment. close=True is the production path
+ # (the workflow only adds --close when AGENT_SHIN_ENABLED=true).
+ pr = self._make_pr(
+ state="closed", body="Updated body with QA proof + screenshots."
+ )
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ posted = {}
+ reopened = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda repo, n: reopened.update({"n": n}),
+ )
+ # close_pr / close_issue MUST NOT fire in reconsider mode.
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close on reconsider pass"),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(
+ {"verdict": "pass", "missing": [], "explanation": "ok now"}
+ ),
+ reconsider=True,
+ )
+ assert result["action"] == "reopened"
+ assert reopened["n"] == 42
+ assert posted["n"] == 42
+ assert "reopened" in posted["body"].lower()
+
+ def test_should_dry_run_reconsider_pass_when_close_false(
+ self, triage_module, monkeypatch
+ ):
+ # Reconsider must honor `close=False` (dry-run) just like the
+ # regular triage flow. A local invocation of
+ # `python triage_with_llm.py --reconsider --pr N` (no --close)
+ # must NOT post a comment or reopen the PR — it should return
+ # `would-reopen` so the operator can preview the outcome.
+ pr = self._make_pr(
+ state="closed", body="Updated body with QA proof + screenshots."
+ )
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not post comment in dry-run reconsider"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen PR in dry-run reconsider"),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(
+ {"verdict": "pass", "missing": [], "explanation": "ok now"}
+ ),
+ reconsider=True,
+ )
+ assert result["action"] == "would-reopen"
+ # The previewed comment body is still returned so a step-summary
+ # writer can render exactly what would have been posted.
+ assert "reopened" in result["comment"].lower()
+
+ def test_should_post_still_failing_on_reconsider_fail(
+ self, triage_module, monkeypatch
+ ):
+ pr = self._make_pr(state="closed", body="still empty")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ posted = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ # Neither reopen nor close should fire when reconsider verdict is fail.
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen on fail"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close again on reconsider fail"),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Still no QA proof.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ reconsider=True,
+ )
+ assert result["action"] == "reconsider-still-failing"
+ assert posted["n"] == 42
+ assert "QA proof" in posted["body"]
+
+ def test_should_dry_run_reconsider_fail_when_close_false(
+ self, triage_module, monkeypatch
+ ):
+ # Mirror dry-run behavior for the FAIL branch — `close=False`
+ # must NOT post the "still failing" comment.
+ pr = self._make_pr(state="closed", body="still empty")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail(
+ "must not post still-failing comment in dry-run"
+ ),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Still no QA proof.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ reconsider=True,
+ )
+ assert result["action"] == "would-reconsider-still-failing"
+ assert "QA proof" in result["comment"]
+
+ def test_should_reopen_on_reconsider_with_linked_issue_short_circuit(
+ self, triage_module, monkeypatch
+ ):
+ # The linked-issue short-circuit also has to honor reconsider mode:
+ # if the contributor edited the body to add `Fixes #1234`, the regex
+ # path should reopen the PR without calling the LLM.
+ pr = self._make_pr(state="closed", body="Fixes #1234\n\nAddresses the bug.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ posted = {}
+ reopened = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda repo, n: reopened.update({"n": n}),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=55,
+ close=True,
+ model="m",
+ judge=lambda p: pytest.fail("LLM must not run when linked-issue matches"),
+ reconsider=True,
+ )
+ assert result["action"] == "reopened"
+ assert reopened["n"] == 55
+ assert "reopened" in posted["body"].lower()
+
+ def test_should_dry_run_reconsider_with_linked_issue_when_close_false(
+ self, triage_module, monkeypatch
+ ):
+ # Linked-issue short-circuit must ALSO honor dry-run.
+ pr = self._make_pr(state="closed", body="Fixes #1234\n\nAddresses the bug.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not post in dry-run"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen in dry-run"),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=55,
+ close=False,
+ model="m",
+ judge=lambda p: pytest.fail("LLM must not run when linked-issue matches"),
+ reconsider=True,
+ )
+ assert result["action"] == "would-reopen"
+
+ def test_should_skip_internal_in_reconsider_mode(self, triage_module, monkeypatch):
+ # Internal authors are exempt from triage in both regular and
+ # reconsider mode — Agent Shin should never reopen one of their PRs
+ # automatically, in case a maintainer closed it intentionally.
+ pr = self._make_pr(
+ state="closed",
+ author_association="MEMBER",
+ user={"login": "krrishdholakia"},
+ )
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen for internal author"),
+ )
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=lambda p: pytest.fail("LLM must not run for internal author"),
+ reconsider=True,
+ )
+ assert result["action"] == "skip-internal-author"
+
+ def test_should_skip_reconsider_when_not_bot_closed(
+ self, triage_module, monkeypatch
+ ):
+ # SECURITY: `@agent-shin reconsider` must NOT reopen a PR/issue
+ # that a MAINTAINER closed for non-rubric reasons (e.g. duplicate,
+ # design rejection, security report). Only PRs closed by the bot
+ # itself should ever be candidates for the reconsider reopen path.
+ pr = self._make_pr(state="closed", body="something.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module, "was_closed_by_agent_shin", lambda *a, **kw: False
+ )
+ # Even though there's no rate-limit conflict, the bot-closed guard
+ # alone is sufficient to block. The LLM judge must never run on a
+ # maintainer-closed PR.
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_reconsider_verdict",
+ lambda *a, **kw: None,
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not comment on maintainer-closed PR"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen maintainer-closed PR"),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: pytest.fail("LLM must not run before bot-closed guard"),
+ reconsider=True,
+ )
+ assert result["action"] == "skip-not-bot-closed"
+
+ def test_should_rate_limit_repeated_reconsider_triggers(
+ self, triage_module, monkeypatch
+ ):
+ # COST CONTROL: each `@agent-shin reconsider` event burns CI
+ # minutes + an OpenAI API call. If the bot already posted a
+ # reconsider verdict within the cooldown window
+ # (RECONSIDER_RATE_LIMIT_SECONDS), refuse to run again. This
+ # bounds the damage from a contributor spamming the trigger.
+ pr = self._make_pr(state="closed", body="something with new edits.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module, "was_closed_by_agent_shin", lambda *a, **kw: True
+ )
+ # Pretend the bot posted a reconsider verdict 1 second ago.
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_reconsider_verdict",
+ lambda *a, **kw: 1.0,
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not comment during cooldown"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda *a, **kw: pytest.fail("must not reopen during cooldown"),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: pytest.fail("LLM must not run during cooldown"),
+ reconsider=True,
+ )
+ assert result["action"] == "skip-rate-limited"
+ assert result["rate_limit_age_seconds"] == 1.0
+ assert (
+ result["rate_limit_window_seconds"]
+ == triage_module.RECONSIDER_RATE_LIMIT_SECONDS
+ )
+
+ def test_should_allow_reconsider_after_cooldown_window(
+ self, triage_module, monkeypatch
+ ):
+ # The cooldown is a window, not a one-shot lock — once
+ # RECONSIDER_RATE_LIMIT_SECONDS has elapsed since the last bot
+ # verdict, a fresh `@agent-shin reconsider` is allowed through.
+ pr = self._make_pr(state="closed", body="updated with screenshots now.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module, "was_closed_by_agent_shin", lambda *a, **kw: True
+ )
+ # Last reconsider was 1 hour ago — well outside the 10-min window.
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_reconsider_verdict",
+ lambda *a, **kw: 3600.0,
+ )
+ posted = {}
+ reopened = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_pr",
+ lambda repo, n: reopened.update({"n": n}),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(
+ {"verdict": "pass", "missing": [], "explanation": "ok"}
+ ),
+ reconsider=True,
+ )
+ assert result["action"] == "reopened"
+ assert reopened["n"] == 1
+
+ def test_should_reopen_issue_on_reconsider_pass(self, triage_module, monkeypatch):
+ issue = {
+ "number": 7,
+ "title": "Bug: now with repro",
+ "body": "## Repro\n```bash\ncurl ...\n```\n\nExpected X, got Y.",
+ "state": "closed",
+ "author_association": "NONE",
+ "user": {"login": "outside"},
+ }
+ monkeypatch.setattr(triage_module, "fetch_issue", lambda repo, n: issue)
+ self._stub_reconsider_guards(triage_module, monkeypatch)
+ posted = {}
+ reopened = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "reopen_issue",
+ lambda repo, n: reopened.update({"n": n}),
+ )
+
+ result = triage_module.triage(
+ repo="o/r",
+ kind="issue",
+ number=7,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(
+ {"verdict": "pass", "missing": [], "explanation": "now reproducible"}
+ ),
+ reconsider=True,
+ )
+ assert result["action"] == "reopened"
+ assert reopened["n"] == 7
+ assert "reopened" in posted["body"].lower()
+
+ def test_should_triage_issues_kind(self, triage_module, monkeypatch):
+ issue = {
+ "number": 7,
+ "title": "Bug: X is broken",
+ "body": "no detail",
+ "state": "open",
+ "author_association": "NONE",
+ "user": {"login": "outside"},
+ }
+ monkeypatch.setattr(triage_module, "fetch_issue", lambda repo, n: issue)
+ # Grace already aged out -> close path. (Issues use the same
+ # GRACE_COMMENT_MARKER detection as PRs.)
+ self._stub_grace_aged_out(triage_module, monkeypatch)
+ closed = {}
+ posted = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update(body=body),
+ )
+ monkeypatch.setattr(
+ triage_module, "close_issue", lambda repo, n: closed.update(n=n)
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "kind": "bug",
+ "has_repro": False,
+ "missing": ["reproduction", "expected vs. actual"],
+ "explanation": "No repro provided.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="issue",
+ number=7,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "closed"
+ assert closed["n"] == 7
+ assert "reproduction" in posted["body"]
+
+ # ---- Grace-period flow ------------------------------------------------
+
+ def test_should_post_grace_warning_on_first_failing_run_in_close_mode(
+ self, triage_module, monkeypatch
+ ):
+ # First low-quality detection -> bot posts a warning comment with
+ # the GRACE_COMMENT_MARKER. The PR must NOT be closed yet.
+ pr = self._make_pr(body="just a sentence.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_no_warning(triage_module, monkeypatch)
+ posted = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close on first detection"),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Body too thin.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "warned-grace"
+ assert posted["n"] == 42
+ # Pin the user-facing language pieces the user explicitly asked for.
+ assert "1 day" in posted["body"]
+ assert "@agent-shin reconsider" in posted["body"]
+ assert "@greptileai" in posted["body"]
+ assert "even after the PR is closed" in posted["body"]
+ assert triage_module.GRACE_COMMENT_MARKER in posted["body"]
+
+ def test_should_skip_close_inside_grace_window(self, triage_module, monkeypatch):
+ # A warning was posted recently; do nothing on this run regardless
+ # of close=True. The next run after `GRACE_PERIOD_SECONDS` elapses
+ # is the one that flips to actual close.
+ pr = self._make_pr(body="just a sentence.")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_grace_warning",
+ lambda *a, **kw: 60.0,
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not comment during grace window"),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close during grace window"),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Body too thin.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=42,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "skip-in-grace-period"
+ assert result["grace_age_seconds"] == 60.0
+ assert result["grace_period_seconds"] == triage_module.GRACE_PERIOD_SECONDS
+
+ def test_should_dry_run_grace_warning_when_close_false(
+ self, triage_module, monkeypatch
+ ):
+ # In dry-run mode the FIRST failing detection returns
+ # `would-warn-grace` (with the previewed comment body) and never
+ # touches GitHub state. Lets a local operator preview the
+ # warning before flipping --close on.
+ pr = self._make_pr(body="thin")
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_no_warning(triage_module, monkeypatch)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: pytest.fail("must not post in dry-run grace warn"),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "thin",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "would-warn-grace"
+ assert "1 day" in result["comment"]
+
+ # ---- IMMEDIATE_CLOSE_LOGINS bypass (e.g. SwiftWinds) -----------------
+
+ def test_should_skip_grace_for_swiftwinds_login(self, triage_module, monkeypatch):
+ # SwiftWinds is in `IMMEDIATE_CLOSE_LOGINS` for dogfooding.
+ # Even though no grace warning has been posted, the bypass must
+ # take the close path immediately.
+ pr = self._make_pr(body="just a sentence.", user={"login": "SwiftWinds"})
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ # The grace check must NOT be consulted at all for an immediate
+ # login — pin that here.
+ monkeypatch.setattr(
+ triage_module,
+ "seconds_since_last_grace_warning",
+ lambda *a, **kw: pytest.fail(
+ "grace check must not run for immediate-close login"
+ ),
+ )
+ posted = {}
+ closed = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda repo, n: closed.update({"n": n}),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "Body too thin.",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=99,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "closed"
+ assert result["immediate_close"] is True
+ assert closed["n"] == 99
+ # The close comment is the standard one (no warning was needed).
+ assert "Agent Shin" in posted["body"]
+
+ def test_should_close_swiftwinds_even_when_close_flag_false(
+ self, triage_module, monkeypatch
+ ):
+ # The `IMMEDIATE_CLOSE_LOGINS` bypass intentionally overrides the
+ # workflow-level dry-run gating. The user explicitly asked to
+ # "turn on full (non dry run) mode with SwiftWinds" because their
+ # personal account has no push permissions to litellm and is the
+ # perfect testing target. With `close=False` (workflow stripped
+ # --close on pull_request_target), the script must STILL post +
+ # close for SwiftWinds.
+ pr = self._make_pr(body="just a sentence.", user={"login": "SwiftWinds"})
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ posted = {}
+ closed = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"n": n, "body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda repo, n: closed.update({"n": n}),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "thin",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=99,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "closed"
+ assert closed["n"] == 99
+
+ def test_should_match_immediate_close_login_case_insensitively(
+ self, triage_module, monkeypatch
+ ):
+ # GitHub returns the original casing of a login. The bypass must
+ # work regardless of the exact case the API returns.
+ for login in ("SwiftWinds", "swiftwinds", "SWIFTWINDS"):
+ pr = self._make_pr(body="thin", user={"login": login})
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda *a, **kw: None,
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: None,
+ )
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "thin",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=False,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "closed", login
+ assert result.get("immediate_close") is True, login
+
+ def test_should_not_treat_random_external_login_as_immediate_close(
+ self, triage_module, monkeypatch
+ ):
+ # Sanity check — the bypass must NOT fire for any login other
+ # than the explicitly-listed test accounts.
+ pr = self._make_pr(body="thin", user={"login": "random-oss-dev"})
+ monkeypatch.setattr(triage_module, "fetch_pr", lambda repo, n: pr)
+ self._stub_grace_no_warning(triage_module, monkeypatch)
+ posted = {}
+ monkeypatch.setattr(
+ triage_module,
+ "post_comment",
+ lambda repo, n, body: posted.update({"body": body}),
+ )
+ monkeypatch.setattr(
+ triage_module,
+ "close_pr",
+ lambda *a, **kw: pytest.fail("must not close non-immediate login"),
+ )
+
+ verdict = {
+ "verdict": "fail",
+ "missing": ["QA proof"],
+ "explanation": "thin",
+ }
+ result = triage_module.triage(
+ repo="o/r",
+ kind="pr",
+ number=1,
+ close=True,
+ model="m",
+ judge=lambda p: json.dumps(verdict),
+ )
+ assert result["action"] == "warned-grace"
+
+
+class TestImmediateCloseLoginsConstant:
+ """SwiftWinds is the dogfood account the user explicitly named — pin
+ its presence so a future cleanup that removes the constant or
+ forgets to keep the entry doesn't silently break the test path."""
+
+ def test_should_include_swiftwinds(self, triage_module):
+ assert "swiftwinds" in triage_module.IMMEDIATE_CLOSE_LOGINS
+
+ def test_should_be_lowercase_for_case_insensitive_match(self, triage_module):
+ for login in triage_module.IMMEDIATE_CLOSE_LOGINS:
+ assert login == login.lower(), login
+
+
+class TestGraceWarningCommentText:
+ """Pin the user-facing promises in the grace warning so a future
+ refactor can't silently drop them."""
+
+ def test_pr_grace_warning_should_state_one_day_grace(self, triage_module):
+ body = triage_module.format_grace_warning_pr_comment(
+ {"verdict": "fail", "missing": ["QA proof"], "explanation": "thin"}
+ )
+ # The user explicitly asked: "specify in the comment" that there
+ # is a 1-day grace.
+ assert "1 day" in body
+
+ def test_pr_grace_warning_should_mention_reconsider_during_grace(
+ self, triage_module
+ ):
+ body = triage_module.format_grace_warning_pr_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ assert "@agent-shin reconsider" in body
+
+ def test_pr_grace_warning_should_promise_greptileai_works_post_close(
+ self, triage_module
+ ):
+ body = triage_module.format_grace_warning_pr_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ # Per user: comment should state @greptileai works even after close.
+ assert "@greptileai" in body
+ assert "even after the PR is closed" in body
+
+ def test_pr_grace_warning_should_carry_grace_marker(self, triage_module):
+ # The marker is what `seconds_since_last_grace_warning` greps for
+ # on subsequent runs to detect that a warning has been posted.
+ # Dropping it would silently break the close-after-grace path.
+ body = triage_module.format_grace_warning_pr_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ assert triage_module.GRACE_COMMENT_MARKER in body
+
+ def test_issue_grace_warning_should_carry_grace_marker(self, triage_module):
+ body = triage_module.format_grace_warning_issue_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ assert triage_module.GRACE_COMMENT_MARKER in body
+ assert "1 day" in body
+ assert "@agent-shin reconsider" in body
+
+ def test_pr_close_comment_should_promise_greptileai_works_post_close(
+ self, triage_module
+ ):
+ # The standard close comment must ALSO point at @greptileai so
+ # contributors see the same options whether they read the warning
+ # or only catch the close comment.
+ body = triage_module.format_pr_close_comment(
+ {"verdict": "fail", "missing": [], "explanation": ""}
+ )
+ assert "@greptileai" in body
+ assert "even after the PR is closed" in body
+
+
+class TestSecondsSinceLastGraceWarning:
+ """Mirror of TestSecondsSinceLastReconsiderVerdict for the new helper.
+ Both helpers share `_seconds_since_latest_marker_comment` underneath
+ so the parsing logic is exercised either way; these tests pin the
+ grace-marker-specific behavior."""
+
+ def _make_comment(
+ self,
+ *,
+ login: str,
+ body: str,
+ created_at: str | None = "2026-05-18T05:00:00Z",
+ ) -> dict:
+ comment: dict = {"user": {"login": login}, "body": body}
+ if created_at is not None:
+ comment["created_at"] = created_at
+ return comment
+
+ def test_should_return_none_when_no_grace_marker(self, triage_module, monkeypatch):
+ comments = [
+ self._make_comment(
+ login="github-actions[bot]",
+ body="Some other bot message",
+ ),
+ self._make_comment(login="random-user", body="ping?"),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+ assert triage_module.seconds_since_last_grace_warning("o/r", 1) is None
+
+ def test_should_ignore_non_bot_comments_with_marker(
+ self, triage_module, monkeypatch
+ ):
+ # A user who quotes the marker in a question must NOT be treated
+ # as the bot warning; otherwise the close-after-grace path would
+ # never fire because the timer keeps resetting.
+ comments = [
+ self._make_comment(
+ login="random-user",
+ body=f"What is {triage_module.GRACE_COMMENT_MARKER}?",
+ )
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+ assert triage_module.seconds_since_last_grace_warning("o/r", 1) is None
+
+ def test_should_pick_latest_grace_marker(self, triage_module, monkeypatch):
+ comments = [
+ self._make_comment(
+ login="github-actions[bot]",
+ body="old warning " + triage_module.GRACE_COMMENT_MARKER,
+ created_at="2026-05-18T03:00:00Z",
+ ),
+ self._make_comment(
+ login="github-actions[bot]",
+ body="newer warning " + triage_module.GRACE_COMMENT_MARKER,
+ created_at="2026-05-18T04:55:00Z",
+ ),
+ ]
+ monkeypatch.setattr(
+ triage_module, "_iter_paginated_json", lambda *a, **kw: iter(comments)
+ )
+
+ import datetime as real_dt
+
+ class FrozenDateTime(real_dt.datetime):
+ @classmethod
+ def now(cls, tz=None):
+ return real_dt.datetime(2026, 5, 18, 5, 0, 0, tzinfo=tz)
+
+ frozen_module = type(triage_module.dt)("datetime")
+ frozen_module.datetime = FrozenDateTime
+ frozen_module.timezone = real_dt.timezone
+ monkeypatch.setattr(triage_module, "dt", frozen_module)
+
+ age = triage_module.seconds_since_last_grace_warning("o/r", 1)
+ # Newer warning is 5 minutes (300s) before "now".
+ assert age == 300.0
diff --git a/tests/test_litellm/test_github_triage_workflows.py b/tests/test_litellm/test_github_triage_workflows.py
new file mode 100644
index 00000000000..da2743da585
--- /dev/null
+++ b/tests/test_litellm/test_github_triage_workflows.py
@@ -0,0 +1,139 @@
+"""Static guardrails for the Agent Shin + Greptile workflow YAML files.
+
+These workflows can post comments and close PRs/issues on
+BerriAI/litellm, so the gating logic that decides "is this a real
+close-on-fail run?" must fail-safe on any unexpected input. The risk
+is mostly maintenance: someone edits the bash gate, drops a quote,
+inverts a comparison, or uses `!= "false"` (which treats "True",
+"yes", "1", and typos as enabling closure) and the regression isn't
+caught until a real OSS contributor's PR gets auto-closed.
+
+The tests below pin two invariants across every workflow that gates a
+destructive `--close`:
+
+ 1. The gate uses the fail-safe `= "true"` comparison — not `!= "false"`,
+ not `!= ""`. Only the literal string "true" should ever enable
+ closure.
+ 2. The gate also requires `AGENT_SHIN_ENABLED = "true"` (or the
+ scheduled-job equivalent) — disabling the variable must always
+ force dry-run.
+
+Static parsing of the YAML + bash text is the right level of test here:
+the gating logic lives in a `run:` block, not in a Python module we can
+import, and end-to-end testing a GitHub Actions workflow from CI is
+infeasible. A YAML-level guardrail is exactly what would have caught
+the original `!= "false"` regression at PR time.
+"""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+import yaml
+
+REPO_ROOT = Path(__file__).resolve().parents[2]
+WORKFLOWS_DIR = REPO_ROOT / ".github" / "workflows"
+
+# Map of workflow file -> the env var name that drives the destructive
+# gate inside that workflow's `run:` block. Keeping this table explicit
+# (rather than scraping every workflow file) means a new workflow file
+# that bypasses the dry-run gating doesn't silently slip past this test.
+DESTRUCTIVE_GATE_ENV: dict[str, str] = {
+ "triage_pr_with_llm.yml": "DISPATCH_CLOSE",
+ "triage_issue_with_llm.yml": "DISPATCH_CLOSE",
+ "close_low_quality_prs.yml": "CLOSE_FLAG",
+ # The reconsider workflow has no per-run "really do it?" knob — its
+ # only kill switch is `AGENT_SHIN_ENABLED`, which already serves as
+ # both the destructive gate and the global enablement gate.
+ "triage_reconsider.yml": "AGENT_SHIN_ENABLED",
+}
+
+
+def _load_workflow(name: str) -> dict:
+ return yaml.safe_load((WORKFLOWS_DIR / name).read_text())
+
+
+def _all_run_blocks(workflow: dict) -> list[str]:
+ """Return every `run:` step's command text, joined."""
+ commands: list[str] = []
+ jobs = workflow.get("jobs") or {}
+ for job in jobs.values():
+ for step in job.get("steps", []) or []:
+ if not isinstance(step, dict):
+ continue
+ run = step.get("run")
+ if isinstance(run, str):
+ commands.append(run)
+ return commands
+
+
+@pytest.mark.parametrize("workflow_file,env_var", sorted(DESTRUCTIVE_GATE_ENV.items()))
+def test_should_use_failsafe_equals_true_comparison(
+ workflow_file: str, env_var: str
+) -> None:
+ """The destructive `--close` gate must use `= "true"` (fail-safe), not
+ `!= "false"` (which would treat "True", "yes", "1", or any typo as
+ enabling closure).
+
+ Both bare `${ENV_VAR}` and `${ENV_VAR:-false}` (with a default) are
+ accepted forms — what matters is the comparison operator. The
+ Greptile closer relies on an outer `AGENT_SHIN_ENABLED` gate so it
+ can use the bare form; the Agent Shin workflows include `:-false`
+ for defense in depth. Either is fine.
+ """
+ workflow = _load_workflow(workflow_file)
+ text = "\n".join(_all_run_blocks(workflow))
+ assert env_var in text, (
+ f"{workflow_file} no longer references {env_var}; was the "
+ "gating env var renamed without updating this test?"
+ )
+ accepted_patterns = (
+ f'"${{{env_var}}}" = "true"',
+ f'"${{{env_var}:-false}}" = "true"',
+ )
+ assert any(p in text for p in accepted_patterns), (
+ f"{workflow_file} must gate the destructive --close flag on the "
+ f'EXACT string "true" (one of: {accepted_patterns!r}). Mirror '
+ 'the Greptile closer pattern; do NOT use `!= "false"` which '
+ 'fail-opens on unknown values like "True", "yes", "1", or typos.'
+ )
+ forbidden_patterns = (
+ f'"${{{env_var}}}" != "false"',
+ f'"${{{env_var}:-false}}" != "false"',
+ f'"${{{env_var}:-true}}" != "false"',
+ )
+ for forbidden in forbidden_patterns:
+ assert forbidden not in text, (
+ f"{workflow_file} uses the fail-open pattern {forbidden!r}. "
+ 'Switch to `= "true"` so unknown values stay dry-run.'
+ )
+
+
+@pytest.mark.parametrize("workflow_file", sorted(DESTRUCTIVE_GATE_ENV))
+def test_should_require_agent_shin_enabled_for_close(workflow_file: str) -> None:
+ """Every destructive gate must also gate on the global enablement
+ variable, so flipping `AGENT_SHIN_ENABLED` off is a kill switch
+ regardless of any per-run input.
+
+ Two patterns are equally fine:
+ - Positive: `[ "${AGENT_SHIN_ENABLED:-false}" = "true" ]` to enter
+ the close branch (Agent Shin workflows).
+ - Negative: `[ "${AGENT_SHIN_ENABLED:-false}" != "true" ]` then
+ bail out / force dry-run (Greptile closer).
+
+ What matters is that the comparison value is the literal "true";
+ `!= "false"` or `= "1"` etc. would not be a true kill switch.
+ """
+ workflow = _load_workflow(workflow_file)
+ text = "\n".join(_all_run_blocks(workflow))
+ accepted_patterns = (
+ '"${AGENT_SHIN_ENABLED:-false}" = "true"',
+ '"${AGENT_SHIN_ENABLED:-false}" != "true"',
+ )
+ assert any(p in text for p in accepted_patterns), (
+ f"{workflow_file} must gate destructive actions on "
+ '`AGENT_SHIN_ENABLED = "true"` (or the inverted `!= "true"` '
+ "guard that forces dry-run). Without this, an unset repo "
+ "variable would not be treated as a kill switch."
+ )