Automate code reviews for GitLab MRs with Gemini AI and JIRA context

What it does

Automates code review by listening for a comment trigger on GitLab merge requests, summarising the diff, and using an LLM to post constructive, line‑specific feedback. If a JIRA ticket ID is found in the MR description, the ticket’s summary is used to inform the AI review.

Use cases

• Quickly obtain high‑quality feedback on MRs without waiting for peers.
• Highlight logic, security or performance issues that might slip through cursory reviews.
• Incorporate project context by pulling in related JIRA ticket summaries.

Good to know

• Triggered by commenting ai-review on a merge request.
• The LLM returns only high‑value findings; if nothing critical is detected, the workflow posts an “all clear” message.
• You can swap out the LLM (Gemini, OpenAI, etc.) or adjust the prompt to fit your team’s guidelines.
• AI usage may incur costs or be geo‑restricted depending on your provider n8n.io.

How it works

• Webhook listener: A Webhook node captures GitLab note events and filters for the trigger phrase.

• Fetch & parse: The workflow retrieves MR details and diffs, splitting each change into “original” and “new” code blocks.

• Optional JIRA context: If your MR description includes a JIRA key (e.g., PROJ-123), the workflow fetches the ticket (and parent ticket for subtasks) and composes a brief context summary.

• LLM review: The parsed diff and optional context are sent to an LLM with instructions to identify logic, security or performance issues and suggest improvements.

• Post results: Inline comments are posted back to the MR at the appropriate file/line positions; if no issues are found, a single “all clear” note is posted.

How to use

• Import the template JSON and open the Webhook node. Replace the REPLACE_WITH_UNIQUE_PATH placeholder with your desired path and configure a GitLab project webhook to send MR comments to that URL.
• Select your LLM credentials in the Gemini (or other LLM) node, and optionally add JIRA credentials in the JIRA nodes.
• Activate the workflow and comment ai-review on any merge request to test it.
• For each review, the workflow posts status updates (“AI review initiated…”) and final comments.

Requirements

• A GitLab project with a generate Personal Access Token (PAT) stored as an environment variable (GITLAB_TOKEN).
• LLM credentials (e.g., Google Gemini) and optional JIRA credentials.

Customising this workflow

• Change the trigger phrase in the Trigger Phrase Filter node.
• Modify the LLM prompt to focus on different aspects (e.g., style, documentation).
• Filter out certain file types or directories before sending diffs to the LLM.
• Integrate other services (Slack, email) to notify teams when reviews are complete.