greploop

Greploop

Iteratively fix a PR/MR/CL until Greptile gives a perfect review: 5/5 confidence, zero unresolved comments.

Inputs

• PR/MR/CL number (optional): If not provided, detect the PR/MR for the current branch, or the default pending changelist for p4.

Instructions

0. Detect platform

First check for Perforce, then fall back to git remote detection:

# Check for Perforce environment
if p4 info >/dev/null 2>&1; then
  VCS="perforce"
else
  REMOTE_URL

For self-hosted GitLab instances whose hostname doesn't contain "gitlab", the user can override by passing --vcs gitlab as an input. For Perforce, pass --vcs perforce.

1. Identify the PR/MR/CL

GitHub:

GitLab:

Switch to the PR/MR branch if not already on it.

Perforce:

Ensure the correct workspace (p4 client) is set before proceeding.

Key field differences:

• GitHub: number, headRefName, headRefOid
• GitLab: iid, source_branch, sha
• Perforce: changelist number, P4CLIENT, shelved files

2. Loop

Repeat the following cycle. Max 5 iterations to avoid runaway loops.

A. Trigger Greptile review

Push/shelve the latest changes (if any):

GitHub/GitLab:

Perforce:

Wait for checks to start after push/shelve:

GitHub — check if Greptile is already running before posting a new trigger comment:

If Greptile is not already running (PENDING or IN_PROGRESS), request a fresh review:

Then poll for the Greptile check run to complete:

If polling times out, stop the greploop workflow and report the timeout. Do not continue with stale or missing review results.

GitLab — check if Greptile is already running before posting a trigger comment:

If no pipeline is running, post a trigger comment:

Perforce — Perforce does not have native check runs. If Greptile is integrated via a webhook triggered on p4 shelve, wait for it to process. Check your Greptile installation's webhook endpoint or dashboard for the review status. Poll by re-fetching the Greptile review comment on the CL until a score appears.

Then poll for the Greptile pipeline job to complete (see GitLab API reference):

If polling times out, stop the greploop workflow and report the timeout. Do not continue with stale or missing review results.

B. Fetch Greptile review results

Greptile may surface its score in several places — check all of the relevant sources:

GitHub:

1. PR description (body):

2. General PR comments (issue comments):

Filter for Greptile-authored comments and use the body from the most recently updated comment (updated_at), not the most recently created comment. Greptile may edit the same general PR comment on each review cycle; parse the current body, including the "Prompt to fix all with AI" section, before deciding there are no remaining issues.

3. PR reviews:

Look for the most recent entry from greptile-apps[bot] or greptile-apps-staging[bot].

GitLab:

1. MR description (body):

2. MR notes (comments):

Filter for notes from the Greptile bot user (check the author.username field — the exact username may vary per installation; verify on first run).

Perforce:

1. CL description:

Check the description field for a Greptile-appended score block.

2. CL comments / review notes: If your installation uses a review tool such as Helix Swarm, fetch comments via its API.

Example (Swarm API): GET /api/v11/comments?topic=reviews/<REVIEW_ID>

Response fields of interest typically include:

• user (author username)
• body (comment text)
• flags/state indicating whether the comment is resolved

Filter to comments authored by the Greptile bot:

• Prefer exact username match if known
• Otherwise, use a heuristic where the author name contains "greptile" (case-insensitive)

For all platforms, parse the text for:

• Confidence score: a pattern like 3/5 or 5/5 (or Confidence: 3/5).
• Comment count: Number of inline review comments noted in the summary.

Use whichever source has the most recently updated score. For GitHub, prefer updated_at from issue comments when comparing an edited Greptile summary against older review entries.

Also fetch all unresolved inline comments:

GitHub:

Also carry forward actionable items from the latest Greptile general PR comment, especially the "Prompt to fix all with AI" section, even if the inline comment endpoint returns zero unresolved comments.

GitLab:

Filter to DiffNote type discussions (notes[0].type == "DiffNote") from Greptile that are on the latest commit and not yet resolved ("resolved": false).

Perforce: If using Swarm:

Fetch inline diff comments for the review associated with the CL

GET /api/v11/comments?topic=reviews/<REVIEW_ID>

Filter to comments from the Greptile bot user that have not been marked as resolved/addressed.

C. Check exit conditions

Stop the loop if any of these are true:

• Confidence score is 5/5 AND there are zero unresolved comments
• Max iterations reached (report current state)

D. Fix actionable comments

For each unresolved Greptile comment:

1. Read the file and understand the comment in context.
2. Determine if it's actionable (code change needed) or informational.
3. If actionable, make the fix.
4. If informational or a false positive, note it but still resolve the thread.

E. Resolve threads

GitHub — fetch unresolved review threads and resolve all that have been addressed (see GraphQL reference):

Resolve addressed threads:

GitLab — fetch unresolved discussions and resolve each one (see GitLab API reference):

Filter for "resolved": false discussions. Then resolve each by its id:

Repeat for each unresolved discussion ID. (GitLab has no batch resolution — loop through each one.)

F. Commit and push / re-shelve

GitHub/GitLab:

Perforce:

Wait for checks to start after push/shelve:

Then go back to step A.

3. Report

After exiting the loop, summarize:

If the loop exited due to max iterations, list any remaining unresolved comments and suggest next steps.

Output format

If not fully resolved:

Perforce example: