creation

2026-04-28 00:29:22 +00:00 · 2025-11-12 16:05:46 +01:00
parent 29e3d1da8c
commit 8d2570ace5
1 changed files with 589 additions and 0 deletions
@@ -0,0 +1,589 @@
+Automate Code Reviews for GitLab MRs with Gemini AI and JIRA Context
+
+https://n8nworkflows.xyz/workflows/automate-code-reviews-for-gitlab-mrs-with-gemini-ai-and-jira-context-7924
+
+
+# Automate Code Reviews for GitLab MRs with Gemini AI and JIRA Context
+
+### 1. Workflow Overview
+
+This workflow automates AI-powered code reviews on GitLab Merge Requests (MRs) by integrating Gemini AI and JIRA context. The main use case is to enable teams to obtain automated, high-value feedback on code changes focusing on business logic, correctness, security, and performance. It listens for a specific GitLab MR comment trigger, fetches the MR diff and relevant JIRA ticket context, invokes an LLM for review, and posts inline comments or summary notes back to the MR.
+
+The workflow is logically divided into the following blocks:
+
+- **1.1 Input Reception & Trigger Handling:** Listens for GitLab MR comment events and filters those requesting a review.
+- **1.2 Static Context Initialization:** Stores persistent data such as project and MR IDs for use across nodes and error handling.
+- **1.3 JIRA Context Extraction:** Extracts JIRA issue keys from MR descriptions and fetches issue and parent issue details.
+- **1.4 Merge Request Data Preparation:** Retrieves MR changes, parses diffs, and prepares original vs new code for review.
+- **1.5 AI Review Processing:** Sends code change segments with JIRA context to Gemini AI for automated review.
+- **1.6 Review Results Processing:** Aggregates AI comments, filters relevant issues, and determines if issues were found.
+- **1.7 Posting Comments to GitLab:** Posts inline comments or summary notes to the MR, with fallback for comments without position.
+- **1.8 Error Handling & Cleanup:** Handles errors with contextual posting and clears static context after execution.
+
+---
+
+### 2. Block-by-Block Analysis
+
+#### 2.1 Input Reception & Trigger Handling
+
+**Overview:**  
+This block listens for GitLab MR comment webhook events and triggers the workflow only if the comment matches the configured trigger phrase.
+
+**Nodes Involved:**  
+- Listen For Gitlab Comments (Webhook)  
+- Need Review (If)  
+
+**Node Details:**
+
+- **Listen For Gitlab Comments**  
+  - Type: Webhook  
+  - Role: Receives HTTP POST from GitLab when a note is added to an MR.  
+  - Key Config: webhook path and HTTP method POST (customized by user).  
+  - Output: Emits event data containing note details.  
+  - Edge Cases: Webhook misconfiguration, unauthorized calls, missing note data.  
+
+- **Need Review**  
+  - Type: If  
+  - Role: Checks if the comment text equals the trigger phrase (default: `coro-bot-review`).  
+  - Key Expression: `{{$json.body.object_attributes.note}} === 'coro-bot-review'`  
+  - Output: Passes only triggering comments for further processing.  
+  - Edge Cases: Case sensitivity, empty or malformed comments.
+
+---
+
+#### 2.2 Static Context Initialization
+
+**Overview:**  
+Stores MR identifiers in workflow static data for persistent reference across nodes and in error handling.
+
+**Nodes Involved:**  
+- Set workflow execution information  
+- Post Review Started  
+
+**Node Details:**
+
+- **Set workflow execution information**  
+  - Type: Code  
+  - Role: Extracts project ID and MR IID from webhook payload and stores them keyed by execution ID in static data.  
+  - Key Expression: Uses `$input.first().json.body.project_id` and MR IID path.  
+  - Output: Passes input unchanged.  
+  - Edge Cases: Missing IDs in webhook payload, static data access issues.
+
+- **Post Review Started**  
+  - Type: HTTP Request  
+  - Role: Posts a comment to the MR notifying that AI review has started.  
+  - Config: POST to GitLab MR notes endpoint with private token authentication.  
+  - Body Content: Fixed message about AI review initiation duration.  
+  - Edge Cases: API rate limits, invalid token, network failures.
+
+---
+
+#### 2.3 JIRA Context Extraction
+
+**Overview:**  
+Extracts JIRA issue keys from the MR description and fetches corresponding JIRA issue and parent issue to enrich context for AI review.
+
+**Nodes Involved:**  
+- Extract MR Details  
+- Extract the JIRA Issue ID  
+- Get JIRA issue  
+- If JIRA Subtask  
+- Get JIRA Parent Issue  
+- Format JIRA Context  
+- Format JIRA Parent Context  
+- Merge  
+
+**Node Details:**
+
+- **Extract MR Details**  
+  - Type: Set  
+  - Role: Extracts project ID, MR IID, and MR description from webhook payload for downstream use.  
+  - Output: Adds `projectId`, `iid`, and `description` fields.  
+  - Edge Cases: Missing or malformed fields.
+
+- **Extract the JIRA Issue ID**  
+  - Type: Code  
+  - Role: Uses regex to find JIRA issue keys (e.g., `PROJ-123`) from MR description text.  
+  - Output: Emits JSON with `jiraIssueKey` or empty array if none found (halts downstream).  
+  - Edge Cases: No JIRA key found, malformed description.
+
+- **Get JIRA issue**  
+  - Type: JIRA  
+  - Role: Fetches JIRA issue data using extracted key.  
+  - Credentials: Jira Software Cloud OAuth.  
+  - Output: JIRA issue JSON.  
+  - Edge Cases: Invalid key, auth errors, API downtime.
+
+- **If JIRA Subtask**  
+  - Type: If  
+  - Role: Checks if the fetched JIRA issue has a parent (subtask).  
+  - Output: True branch triggers fetching parent issue.  
+  - Edge Cases: Missing parent field, null values.
+
+- **Get JIRA Parent Issue**  
+  - Type: JIRA  
+  - Role: Fetches parent issue details if subtask.  
+  - Output: Parent issue JSON.  
+  - Edge Cases: Same as Get JIRA issue.
+
+- **Format JIRA Context**  
+  - Type: Set  
+  - Role: Formats the JIRA issue summary, type, and description into a string for the prompt.  
+  - Output: Adds `jiraContext` string.  
+  - Edge Cases: Missing fields in JIRA data.
+
+- **Format JIRA Parent Context**  
+  - Type: Set  
+  - Role: Formats parent JIRA issue similarly.  
+  - Output: Adds `jiraParentContext` string.
+
+- **Merge**  
+  - Type: Merge  
+  - Role: Combines JIRA context data streams into a single object for prompt construction.  
+  - Mode: Combine by position, including unpaired inputs.  
+  - Edge Cases: Missing inputs, partial data.
+
+---
+
+#### 2.4 Merge Request Data Preparation
+
+**Overview:**  
+Fetches the MR code changes from GitLab API, splits the diff into file changes, filters out irrelevant changes, parses diff hunks, and prepares original vs new code segments for review.
+
+**Nodes Involved:**  
+- Get MR Changes  
+- Split Out Changes  
+- Skip File Changes  
+- Parse Diff  
+- Prepare Code Changes  
+- Filter Irrelevant Fields  
+
+**Node Details:**
+
+- **Get MR Changes**  
+  - Type: HTTP Request  
+  - Role: Retrieves the MR changes (diffs) from GitLab API using project ID and MR IID.  
+  - Auth: Private token from environment variable.  
+  - Output: JSON with MR changes.  
+  - Edge Cases: API rate limits, invalid tokens, large MR diffs.
+
+- **Split Out Changes**  
+  - Type: Split Out  
+  - Role: Splits the array of file changes into individual items for granular processing.  
+  - Fields Included: JIRA context, IID, project ID, diff refs.  
+  - Edge Cases: Empty changes array.
+
+- **Skip File Changes**  
+  - Type: If  
+  - Role: Filters out renamed or deleted files or diffs starting with `@@` (likely metadata).  
+  - Conditions: renamed_file=false, deleted_file=false, diff starts with "@@"  
+  - Edge Cases: Unexpected diff formats.
+
+- **Parse Diff**  
+  - Type: Code  
+  - Role: Parses the last diff hunk to extract line number information for positioning inline comments.  
+  - Output: Adds `lastOldLine`, `lastNewLine`, and cleaned `gitDiff`.  
+  - Edge Cases: Malformed diffs, no newline at end marker.
+
+- **Prepare Code Changes**  
+  - Type: Code  
+  - Role: Splits diff text into original and new code blocks by line prefix (`-` for original, `+` for new).  
+  - Output: Adds `originalCode` and `newCode` fields to each item.  
+  - Edge Cases: Complex diffs with context lines, empty diffs.
+
+- **Filter Irrelevant Fields**  
+  - Type: Set  
+  - Role: Selects and passes only relevant fields for downstream nodes (diff refs, file paths, line numbers).  
+  - Output: Refined JSON for AI prompt.  
+
+---
+
+#### 2.5 AI Review Processing
+
+**Overview:**  
+Constructs a prompt including JIRA context and code changes, then sends it to Gemini AI for review focusing on logic, security, and performance issues.
+
+**Nodes Involved:**  
+- Basic LLM Chain  
+- Google Gemini Chat Model  
+- Merge LLM Output with Input  
+- Aggregate  
+
+**Node Details:**
+
+- **Basic LLM Chain**  
+  - Type: Langchain LLM Chain  
+  - Role: Builds a structured prompt with JIRA context and a code diff snippet for the LLM.  
+  - Prompt:  
+    ```
+    First, consider the context from the associated JIRA ticket:
+    ------------
+    {{jiraParentContext || jiraContext || 'No JIRA context was provided.'}}
+    ------------
+    
+    File path：{{ new_path }}
+    
+    ```Original code
+    {{ originalCode }}
+    ```
+    change to
+    ```New code
+    {{ newCode }}
+    ```
+    Please review the code changes in this section:
+    ```
+  - Output: Text prompt sent to LLM.  
+  - Edge Cases: Missing context, large diffs.
+
+- **Google Gemini Chat Model**  
+  - Type: Langchain LM Chat (Google Gemini)  
+  - Role: Sends prompt to Gemini AI model for code review.  
+  - Config: Temperature 0, topK 1, model version "gemini-2.5-pro".  
+  - Credentials: Google Palm API credentials.  
+  - Output: AI-generated review text.  
+  - Edge Cases: API quota, rate limits, model errors.
+
+- **Merge LLM Output with Input**  
+  - Type: Merge  
+  - Role: Combines AI review output with original input item to keep context.  
+  - Edge Cases: Missing AI response.
+
+- **Aggregate**  
+  - Type: Aggregate  
+  - Role: Collects all AI responses into one array for further filtering.  
+  - Edge Cases: Empty responses.
+
+---
+
+#### 2.6 Review Results Processing
+
+**Overview:**  
+Filters AI comments to exclude empty or non-actionable items and determines whether any issues were found.
+
+**Nodes Involved:**  
+- Process Comments  
+- Any Issues Found?  
+- Get Static Context  
+- Split Out Comments  
+
+**Node Details:**
+
+- **Process Comments**  
+  - Type: Code  
+  - Role: Filters aggregated AI comments removing those with empty text, `ALL_CLEAR`, or variations of no findings.  
+  - Output: `commentsToPost` array and `noIssuesFound` boolean.  
+  - Edge Cases: Incorrect filtering logic.
+
+- **Any Issues Found?**  
+  - Type: If  
+  - Role: Checks if no issues were found by evaluating `noIssuesFound` boolean.  
+  - Output: True branch triggers posting a no-issues note, false branch proceeds to post comments.
+
+- **Get Static Context**  
+  - Type: Code  
+  - Role: Retrieves project and MR IDs from static data for posting comments.  
+  - Edge Cases: Missing static data due to execution context issues.
+
+- **Split Out Comments**  
+  - Type: Split Out  
+  - Role: Splits filtered comments into individual items for posting.  
+  - Edge Cases: Empty comments array.
+
+---
+
+#### 2.7 Posting Comments to GitLab
+
+**Overview:**  
+Posts AI review comments to GitLab MR discussions, either as inline comments with position or as fallback top-level comments if positioning fails.
+
+**Nodes Involved:**  
+- Prepare Request  
+- Post Gitlab MR Comments  
+- Prepare Request Without Position  
+- Post Gitlab MR Comments WIthout Position  
+- Post No Issues Found  
+- Cleanup Static Context  
+
+**Node Details:**
+
+- **Prepare Request**  
+  - Type: Code  
+  - Role: Builds request body including nested GitLab MR diff position info (file paths, commit SHAs, line numbers).  
+  - Output: `requestBody` with `body` and `position` keys for inline comments.  
+  - Edge Cases: Missing line info, invalid data types.
+
+- **Post Gitlab MR Comments**  
+  - Type: HTTP Request  
+  - Role: Posts inline comments to GitLab MR discussions with position info.  
+  - Auth: Private token from environment variable.  
+  - On Error: Continues without stopping workflow to allow fallback.  
+  - Edge Cases: API errors, invalid positioning.
+
+- **Prepare Request Without Position**  
+  - Type: Code  
+  - Role: Builds request body without position info to post as general discussion comment (fallback).  
+  - Output: `requestBody` with only `body` key.
+
+- **Post Gitlab MR Comments WIthout Position**  
+  - Type: HTTP Request  
+  - Role: Posts fallback comments at MR discussion level without positioning.  
+  - Edge Cases: API limits, missing data.
+
+- **Post No Issues Found**  
+  - Type: HTTP Request  
+  - Role: Posts a summary comment indicating the AI found no issues.  
+  - Body: Fixed "LGTM" message.  
+  - Edge Cases: Posting failures.
+
+- **Cleanup Static Context**  
+  - Type: Code  
+  - Role: Deletes stored static context data for the current execution to avoid stale data.  
+  - Output: Empty data array.  
+
+---
+
+#### 2.8 Error Handling & Cleanup
+
+**Overview:**  
+Handles errors during execution by retrieving static context and posting an error message to the MR, followed by cleanup.
+
+**Nodes Involved:**  
+- Error Trigger  
+- Get Static Context Including Error  
+- Post Error Occurred  
+- Cleanup Static Context  
+
+**Node Details:**
+
+- **Error Trigger**  
+  - Type: Error Trigger Node  
+  - Role: Catches any workflow errors to initiate error handling path.  
+  - Edge Cases: Errors in error handling nodes.
+
+- **Get Static Context Including Error**  
+  - Type: Code  
+  - Role: Retrieves static context combined with error message for reporting.  
+  - Output: JSON including `projectId`, `mrId`, and `error` message.
+
+- **Post Error Occurred**  
+  - Type: HTTP Request  
+  - Role: Posts an error comment to GitLab MR notifying users of the failure and suggesting to retry.  
+  - Edge Cases: API errors during error reporting.
+
+- **Cleanup Static Context**  
+  - (Shared node with posting block) Deletes static data after error handling.
+
+---
+
+### 3. Summary Table
+
+| Node Name                      | Node Type                    | Functional Role                          | Input Node(s)                        | Output Node(s)                      | Sticky Note                                                                                   |
+|--------------------------------|------------------------------|----------------------------------------|------------------------------------|-----------------------------------|----------------------------------------------------------------------------------------------|
+| Listen For Gitlab Comments      | Webhook                      | Entry point, listens for MR comments   | -                                  | Need Review                       | "### 🟢 Step 1: Listen & Capture\nA Webhook node listens for merge request note events. ..."  |
+| Need Review                    | If                           | Checks if comment equals trigger phrase| Listen For Gitlab Comments          | Set workflow execution information| "Change the trigger word in the **IF** node to match your own.\n..."                         |
+| Set workflow execution information | Code                       | Stores MR and project IDs in static data | Need Review                      | Post Review Started, Extract MR Details | "### 🔵 Static Context & Error Handling\nWorkflow Static Data is used to store persistent values..." |
+| Post Review Started            | HTTP Request                 | Posts "AI review started" comment      | Set workflow execution information | Extract MR Details                |                                                                                              |
+| Extract MR Details             | Set                         | Extracts MR project ID, IID, description | Set workflow execution information | Get MR Changes, Extract the JIRA Issue ID |                                                                                              |
+| Extract the JIRA Issue ID      | Code                        | Extracts JIRA key from MR description   | Extract MR Details                 | Get JIRA issue                   |                                                                                              |
+| Get JIRA issue                | JIRA                        | Fetches JIRA issue by key               | Extract the JIRA Issue ID          | Format JIRA Context, If JIRA Subtask |                                                                                              |
+| Format JIRA Context           | Set                         | Formats JIRA issue info for prompt      | Get JIRA issue                    | Merge                            |                                                                                              |
+| If JIRA Subtask              | If                           | Checks if JIRA issue is a subtask       | Get JIRA issue                    | Get JIRA Parent Issue            |                                                                                              |
+| Get JIRA Parent Issue        | JIRA                        | Fetches parent issue details            | If JIRA Subtask                   | Format JIRA Parent Context       |                                                                                              |
+| Format JIRA Parent Context   | Set                         | Formats parent JIRA context             | Get JIRA Parent Issue             | Merge                           |                                                                                              |
+| Merge                       | Merge                       | Combines JIRA contexts                   | Format JIRA Context, Format JIRA Parent Context | Get MR Changes                |                                                                                              |
+| Get MR Changes               | HTTP Request                 | Fetches MR file changes                  | Extract MR Details, Merge         | Split Out Changes                |                                                                                              |
+| Split Out Changes            | Split Out                   | Splits file changes array                | Get MR Changes                   | Skip File Changes               |                                                                                              |
+| Skip File Changes            | If                           | Filters out renamed/deleted/irrelevant files | Split Out Changes              | Parse Diff                     |                                                                                              |
+| Parse Diff                  | Code                        | Parses diff hunks for line info          | Skip File Changes                | Prepare Code Changes            |                                                                                              |
+| Prepare Code Changes         | Code                        | Splits diff into original and new code  | Parse Diff                     | Basic LLM Chain, Filter Irrelevant Fields |                                                                                              |
+| Filter Irrelevant Fields     | Set                         | Selects relevant fields for review       | Prepare Code Changes            | Merge LLM Output with Input     |                                                                                              |
+| Basic LLM Chain             | Langchain LLM Chain          | Builds prompt for AI review               | Prepare Code Changes            | Google Gemini Chat Model        |                                                                                              |
+| Google Gemini Chat Model    | Langchain LM Chat            | Sends prompt to Gemini AI model           | Basic LLM Chain                | Basic LLM Chain                |                                                                                              |
+| Merge LLM Output with Input | Merge                        | Combines AI output with original input   | Filter Irrelevant Fields, Basic LLM Chain | Aggregate                    |                                                                                              |
+| Aggregate                  | Aggregate                    | Aggregates all AI review outputs          | Merge LLM Output with Input     | Process Comments               |                                                                                              |
+| Process Comments           | Code                        | Filters out empty or irrelevant AI comments | Aggregate                    | Any Issues Found?              |                                                                                              |
+| Any Issues Found?          | If                           | Checks if any issues found by AI          | Process Comments               | Get Static Context, Split Out Comments |                                                                                              |
+| Get Static Context         | Code                        | Fetches stored project and MR IDs         | Any Issues Found? (true branch)  | Post No Issues Found           |                                                                                              |
+| Post No Issues Found       | HTTP Request                 | Posts 'No issues found' comment           | Get Static Context              | Cleanup Static Context         |                                                                                              |
+| Split Out Comments         | Split Out                   | Splits actionable comments for posting    | Any Issues Found? (false branch) | Prepare Request              |                                                                                              |
+| Prepare Request            | Code                        | Builds comment body with position info    | Split Out Comments             | Post Gitlab MR Comments        |                                                                                              |
+| Post Gitlab MR Comments    | HTTP Request                 | Posts inline comments to MR                | Prepare Request               | Cleanup Static Context, Prepare Request Without Position |                                                                                              |
+| Prepare Request Without Position | Code                     | Builds comment body without position info | Post Gitlab MR Comments       | Post Gitlab MR Comments WIthout Position |                                                                                              |
+| Post Gitlab MR Comments WIthout Position | HTTP Request       | Posts fallback comments without position   | Prepare Request Without Position | Cleanup Static Context         |                                                                                              |
+| Cleanup Static Context     | Code                        | Clears static context at workflow end      | Post No Issues Found, Post Gitlab MR Comments, Post Gitlab MR Comments WIthout Position, Post Error Occurred | - |                                                                                              |
+| Error Trigger             | Error Trigger               | Captures workflow errors                    | -                            | Get Static Context Including Error |                                                                                              |
+| Get Static Context Including Error | Code                  | Retrieves static data and error message     | Error Trigger                | Post Error Occurred            |                                                                                              |
+| Post Error Occurred       | HTTP Request                 | Posts error message comment to MR           | Get Static Context Including Error | Cleanup Static Context         |                                                                                              |
+
+---
+
+### 4. Reproducing the Workflow from Scratch
+
+1. **Create a Webhook node** named "Listen For Gitlab Comments":  
+   - HTTP Method: POST  
+   - Path: Set your unique webhook path  
+   - This node listens for GitLab MR note events.
+
+2. **Add an If node** named "Need Review":  
+   - Condition: `{{$json.body.object_attributes.note}} === 'coro-bot-review'` (or your preferred trigger phrase)  
+   - Connect "Listen For Gitlab Comments" → "Need Review" (main output).
+
+3. **Add a Code node** named "Set workflow execution information":  
+   - Code extracts `projectId` and `mrId` from webhook payload  
+   - Stores them in global static data keyed by execution ID.  
+   - Connect "Need Review" → "Set workflow execution information".
+
+4. **Add an HTTP Request node** named "Post Review Started":  
+   - POST to `https://gitlab.com/api/v4/projects/{{ $json.body.project_id }}/merge_requests/{{ $json.body.merge_request.iid }}/notes`  
+   - Headers: `PRIVATE-TOKEN` from env variable (e.g., `GITLAB_TOKEN`)  
+   - Body: `{"body":"🤖 AI code review initiated. This may take up to 30 minutes for large merge requests. I'll post my findings as comments on the relevant files."}`  
+   - Connect "Set workflow execution information" → "Post Review Started".
+
+5. **Add a Set node** named "Extract MR Details":  
+   - Extract from webhook JSON:  
+     - `projectId` = `{{$json.body.project_id}}`  
+     - `iid` = `{{$json.body.merge_request.iid}}`  
+     - `description` = `{{$json.body.merge_request.description}}`  
+   - Connect "Set workflow execution information" → "Extract MR Details".
+
+6. **Add an HTTP Request node** named "Get MR Changes":  
+   - GET `https://gitlab.com/api/v4/projects/{{ $json.projectId }}/merge_requests/{{ $json.iid }}/changes`  
+   - Headers: `PRIVATE-TOKEN` from env variable  
+   - Connect "Extract MR Details" → "Get MR Changes".
+
+7. **Add a Split Out node** named "Split Out Changes":  
+   - Split field: `changes`  
+   - Include fields: `jiraParentContext`, `jiraContext`, `iid`, `project_id`, `diff_refs`  
+   - Connect "Get MR Changes" (main output index 2) → "Split Out Changes".
+
+8. **Add an If node** named "Skip File Changes":  
+   - Conditions:  
+     - `renamed_file` is false  
+     - `deleted_file` is false  
+     - `diff` starts with "@@"  
+   - Connect "Split Out Changes" → "Skip File Changes".
+
+9. **Add a Code node** named "Parse Diff":  
+   - Parses last diff hunk and extracts `lastOldLine` and `lastNewLine` line numbers.  
+   - Connect "Skip File Changes" (true branch) → "Parse Diff".
+
+10. **Add a Code node** named "Prepare Code Changes":  
+    - Splits diff lines into `originalCode` and `newCode` by prefix '-' and '+'.  
+    - Connect "Parse Diff" → "Prepare Code Changes".
+
+11. **Add a Set node** named "Filter Irrelevant Fields":  
+    - Pass only relevant fields: `iid`, `project_id`, `diff_refs`, `lastNewLine`, `lastOldLine`, `new_path`, `old_path`.  
+    - Connect "Prepare Code Changes" → "Filter Irrelevant Fields".
+
+12. **Add a Langchain LLM Chain node** named "Basic LLM Chain":  
+    - Prompt includes JIRA context, file path, original and new code blocks, with instructions focusing on business logic, security, and performance.  
+    - Connect "Prepare Code Changes" → "Basic LLM Chain".
+
+13. **Add a Google Gemini Chat Model node**:  
+    - Model name: `models/gemini-2.5-pro`  
+    - Temperature: 0  
+    - TopK: 1  
+    - Credentials: Google Palm API (set up in n8n credentials).  
+    - Connect "Basic LLM Chain" → "Google Gemini Chat Model".  
+    - Connect "Google Gemini Chat Model" → "Basic LLM Chain" (ai_languageModel input).
+
+14. **Add a Merge node** named "Merge LLM Output with Input":  
+    - Combine mode: combine by position  
+    - Connect "Filter Irrelevant Fields" → "Merge LLM Output with Input" (input 1)  
+    - Connect "Basic LLM Chain" → "Merge LLM Output with Input" (input 2).
+
+15. **Add an Aggregate node** named "Aggregate":  
+    - Aggregate all item data.  
+    - Connect "Merge LLM Output with Input" → "Aggregate".
+
+16. **Add a Code node** named "Process Comments":  
+    - Filters out comments with empty text, `ALL_CLEAR`, or "no issues" phrases.  
+    - Outputs `commentsToPost` and `noIssuesFound`.  
+    - Connect "Aggregate" → "Process Comments".
+
+17. **Add an If node** named "Any Issues Found?":  
+    - Condition: `noIssuesFound === true`  
+    - Connect "Process Comments" → "Any Issues Found?".
+
+18. **Add a Code node** named "Get Static Context":  
+    - Retrieves `projectId` and `mrId` from static data using execution ID.  
+    - Connect "Any Issues Found?" (true branch) → "Get Static Context".
+
+19. **Add an HTTP Request node** named "Post No Issues Found":  
+    - POST to GitLab MR notes endpoint  
+    - Body: `"🤖 AI review complete. No significant issues were found. LGTM!"`  
+    - Headers: `PRIVATE-TOKEN`  
+    - Connect "Get Static Context" → "Post No Issues Found".
+
+20. **Add a Code node** named "Cleanup Static Context":  
+    - Deletes static data for current execution.  
+    - Connect "Post No Issues Found" → "Cleanup Static Context".
+
+21. **Connect "Any Issues Found?" (false branch) → Split Out node "Split Out Comments":**  
+    - Field to split out: `commentsToPost`  
+    - Include fields: `iid`, `project_id`, `diff_refs`.  
+
+22. **Add a Code node** named "Prepare Request":  
+    - Builds GitLab comment request body including position info (file paths, commit SHAs, line numbers).  
+    - Connect "Split Out Comments" → "Prepare Request".
+
+23. **Add an HTTP Request node** named "Post Gitlab MR Comments":  
+    - POST to GitLab MR discussions endpoint with JSON body from `requestBody`.  
+    - Headers: `PRIVATE-TOKEN`  
+    - Retries disabled, on error continue output.  
+    - Connect "Prepare Request" → "Post Gitlab MR Comments".
+
+24. **Add a Code node** named "Prepare Request Without Position":  
+    - Builds comment body without position for fallback posting.  
+    - Connect "Post Gitlab MR Comments" (error output) → "Prepare Request Without Position".
+
+25. **Add an HTTP Request node** named "Post Gitlab MR Comments WIthout Position":  
+    - Posts fallback comments without position info.  
+    - Headers: `PRIVATE-TOKEN`  
+    - Connect "Prepare Request Without Position" → "Post Gitlab MR Comments WIthout Position".
+
+26. **Connect both "Post Gitlab MR Comments" (success) and "Post Gitlab MR Comments WIthout Position" → "Cleanup Static Context"** to clear data after posting.
+
+27. **Add an Error Trigger node** named "Error Trigger":  
+    - Captures workflow errors.
+
+28. **Add a Code node** named "Get Static Context Including Error":  
+    - Retrieves static context and error message for error reporting.  
+    - Connect "Error Trigger" → "Get Static Context Including Error".
+
+29. **Add an HTTP Request node** named "Post Error Occurred":  
+    - Posts error message comment to GitLab MR including error details.  
+    - Headers: `PRIVATE-TOKEN`.  
+    - Connect "Get Static Context Including Error" → "Post Error Occurred".
+
+30. **Connect "Post Error Occurred" → "Cleanup Static Context"** to clear data after error reporting.
+
+31. **JIRA Context Integration:**  
+    - From "Extract MR Details", connect to "Extract the JIRA Issue ID" (Code node with regex to find JIRA key).  
+    - Connect "Extract the JIRA Issue ID" → "Get JIRA issue" (JIRA node).  
+    - Connect "Get JIRA issue" →  
+      - "Format JIRA Context" (Set node) and  
+      - "If JIRA Subtask" (If node).  
+    - "If JIRA Subtask" (true branch) → "Get JIRA Parent Issue" → "Format JIRA Parent Context".  
+    - "Format JIRA Context" and "Format JIRA Parent Context" → "Merge".  
+    - Connect "Merge" → "Get MR Changes" (to add JIRA context to MR data).
+
+---
+
+### 5. General Notes & Resources
+
+| Note Content                                                                                                                                                         | Context or Link                                                                                     |
+|--------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------|
+| ⭐️ Try It Out! Welcome! This template adds AI-powered code reviews to your GitLab Merge Requests. Simply comment `ai-review` on any MR to trigger the workflow.     | Sticky note "Try It Out!" at the workflow start.                                                  |
+| 🟢 Step 1: Listen & Capture - A Webhook node listens for MR note events; extracts project & MR IDs; posts a review start message.                                  | Sticky note "Step 1: Listen & Capture".                                                           |
+| 🟣 Step 2: Prepare & Review - Fetches MR diffs, parses code changes, obtains JIRA context, and sends to LLM for review focusing on logic, security, and performance. | Sticky note "Step 2: Prepare & Review".                                                           |
+| 🟠 Step 3: Post & Fallback - Posts inline comments with position or fallback thread comments if position missing.                                                  | Sticky note "Step 3: Post & Fallback".                                                            |
+| 🔵 Static Context & Error Handling - Uses workflow static data to persist MR info across nodes and to enable posting error messages on failure.                    | Sticky note "Static Context & Error Handling".                                                    |
+| 🟡 Need Help & Customise - Change trigger word, swap LLM, adjust prompts, or filter files/dirs as needed.                                                          | Sticky note "Need Help & Customise".                                                              |
+| GitLab API documentation: https://docs.gitlab.com/ee/api/merge_requests.html#list-merge-request-changes                                                              | Useful for understanding MR changes API.                                                         |
+| JIRA Cloud API documentation: https://developer.atlassian.com/cloud/jira/software/rest/api-group-issue-fields/                                                      | For JIRA issue data retrieval.                                                                    |
+| Gemini AI (Google PaLM) API docs: https://developers.generativeai.google/api/rest/gemini                                                                                 | For LLM model usage and parameters.                                                              |
+
+---
+
+This completes the detailed reference documentation for the "Gitlab Code Review Template" workflow integrating Gemini AI and JIRA context to automate code reviews on GitLab Merge Requests.