> For the complete documentation index, see [llms.txt](https://developer.harness.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://developer.harness.io/database-devops/use-db-devops/create-database-updates/configure-llm-for-database-devops.md). # Configure Author DB Change This guide explains how to enable an LLM-driven change authoring workflow in Harness Database DevOps that generates migration artifacts (e.g., Liquibase changesets or SQL migration scripts), commits them to a temporary branch, and opens a Pull Request (PR) against a target Git branch for review and CI validation. The workflow preserves GitOps principles (auditability, review gates, CI checks) while accelerating authoring through LLM assistance. ### Before you begin Before configuring LLM change authoring, ensure you have the following: * An active Harness account with Database DevOps enabled. * Access to a Git repository where database migration artifacts will be stored. {% tabs %} {% tab title="Interactive Guide" %} {% embed url="" %} Configure Harness AI and DB Schema {% endembed %} {% endtab %} {% tab title="Step-by-Step Guide" %} #### Step 1: Enable Harness AI 1. Log in to your Harness account. 2. Navigate to **Account Settings** > **Default Settings** > **Harness AI**. 3. Toggle the **Harness AI** switch to enable it for your account. 4. Then click on **Database DevOps** and in **LLM Authoring Test and Commit Pipeline** field point to the LLM pipeline identifier, e.g., `dbops_llm_test_preview` which we will create in the next step. #### Step 2: Create a pipeline with Author DB Change steps 1. Go to the **Pipelines** section in Harness. 2. Click on **Create Pipeline**, and click on Add Stage and select the Custom Stage 3. Provide the desired name for your pipeline, and click on **Add**. (Make sure that the pipeline identifier is same as mentioned in LLM pipeline identifier in step 1) 4. Create a New Step Group. 5. Click on **Add Step** and select the **DB Test and Preview** step from the list of available steps. ![Database DevOps Test and Preview](/files/czSf7aHqQVplrYa6xp3S) {% hint style="info" %} **IMPORTANT** Leave schema and instance set as runtime input, i.e., `<+input>`, as these will be provided by the LLM change authoring. {% endhint %} 6. Click on Add Step again and select the Run Step from the list of available steps. 7. Configure the **Run step** to run the following script to create a Pull Request in your Git repository with the generated migration artifacts. * First, add a **pipeline variable** (or step input) named **$REPO\_TYPE** with allowed values: "github", "bitbucket", "harness". (This variable will determine the Git provider and API used for creating the Pull Request). * In the Run step configuration, set the image to `prontotools/alpine-git-curl:latest` which has git and curl installed, and set the shell to `Sh`. * Then use the following unified script, replacing the common placeholders with your actual values: ```bash #!/bin/bash set -e # All configuration is read from step environment variables — no edits needed below. cd dbops/DBTestAndPreview_1 git config --global user.email "$USER_EMAIL" git config --global user.name "$USER_NAME" # ── Set remote URL ──────────────────────────────────────────────────────────── case "$REPO_TYPE" in github) ENCODED_EMAIL="${USER_EMAIL//@/%40}" REMOTE_URL="https://${ENCODED_EMAIL}:<+secrets.getValue("github")>@github.com/${USERNAME}/${REPO}.git" ;; bitbucket) REMOTE_URL="https://${USERNAME}:<+secrets.getValue("bitbucket")>@bitbucket.org/${WORKSPACE}/${REPO}.git" ;; harness) REMOTE_URL="https://${HARNESS_CODE_NAME}:<+secrets.getValue("harness_api_token")>@git.harness.io/${ACCOUNT_ID}/${ORG_ID}/${PROJECT_ID}/${REPO}.git" ;; *) echo "ERROR: Unsupported REPO_TYPE '${REPO_TYPE}'. Must be github, bitbucket, or harness." exit 1 ;; esac # ── Push to a new branch ────────────────────────────────────────────────────── git remote remove origin 2>/dev/null || true git remote add origin "$REMOTE_URL" git fetch origin DEFAULT_BRANCH=$(git remote show origin | grep 'HEAD branch' | awk '{print $NF}') DEFAULT_BRANCH="${DEFAULT_BRANCH:-main}" echo "Default branch: $DEFAULT_BRANCH" git add . git commit -m "Updating changeset in changelog file" NEW_BRANCH="changelog-update-$(date +%Y%m%d%H%M%S)" git checkout -b "$NEW_BRANCH" git push -u origin "$NEW_BRANCH" # ── Open a Pull Request ─────────────────────────────────────────────────────── case "$REPO_TYPE" in github) PR_RESPONSE=$(curl -s -X POST \ -H "Authorization: token '<+secrets.getValue("github")>'" \ -H "Accept: application/vnd.github+json" \ "https://api.github.com/repos/${USERNAME}/${REPO}/pulls" \ -d "{\"title\":\"Automated changelog update\",\"head\":\"${NEW_BRANCH}\",\"base\":\"${DEFAULT_BRANCH}\"}") PR_URL=$(echo "$PR_RESPONSE" | grep '"html_url"' | head -1 | sed -E 's/.*"html_url": *"([^"]+)".*/\1/') ;; bitbucket) PR_RESPONSE=$(curl -s -X POST \ -u "${USERNAME}:<+secrets.getValue("bitbucket")>" \ -H "Content-Type: application/json" \ "https://api.bitbucket.org/2.0/repositories/${WORKSPACE}/${REPO}/pullrequests" \ -d "{\"title\":\"Automated changelog update\",\"source\":{\"branch\":{\"name\":\"${NEW_BRANCH}\"}},\"destination\":{\"branch\":{\"name\":\"${DEFAULT_BRANCH}\"}},\"close_source_branch\":true}") PR_URL=$(echo "$PR_RESPONSE" | grep -o '"html":{"href":"[^"]*"' | sed 's/"html":{"href":"//;s/"$//') ;; harness) PR_RESPONSE=$(curl -s -w "\n__HTTP_STATUS__%{http_code}" -X POST \ "https://app.harness.io/code/api/v1/repos/${REPO}/pullreq?accountIdentifier=${ACCOUNT_ID}&orgIdentifier=${ORG_ID}&projectIdentifier=${PROJECT_ID}" \ -H "Content-Type: application/json" \ -H "x-api-key: <+secrets.getValue("harness_api_token")>" \ -d "{\"title\":\"Automated changelog update\",\"source_branch\":\"${NEW_BRANCH}\",\"target_branch\":\"${DEFAULT_BRANCH}\",\"is_draft\":false}") HTTP_STATUS=$(echo "$PR_RESPONSE" | grep "__HTTP_STATUS__" | sed 's/__HTTP_STATUS__//') BODY=$(echo "$PR_RESPONSE" | grep -v "__HTTP_STATUS__") if [ "$HTTP_STATUS" -lt 200 ] || [ "$HTTP_STATUS" -ge 300 ]; then echo "❌ Failed to create PR (HTTP $HTTP_STATUS)" echo "$BODY" exit 1 fi PR_NUMBER=$(echo "$BODY" | grep -o '"number":[0-9]*' | head -1 | sed 's/"number"://') PR_URL="https://app.harness.io/ng/account/${ACCOUNT_ID}/module/code/orgs/${ORG_ID}/projects/${PROJECT_ID}/repos/${REPO}/pull-requests/${PR_NUMBER}" ;; esac echo "Pull request created: $PR_URL" export CHANGESET_PR_LINK="$PR_URL" ``` Setup the following environment variables in the Run step configuration, ensuring that sensitive values like `github`, `bitbucket`, and `harness_api_token` are stored securely in Harness secrets and referenced appropriately: | **Variable Name** | **Type** | **Description** | **Example** | | ------------------- | -------- | --------------------------------------------- | -------------------------- | | REPO\_TYPE | String | Target Git provider | github, bitbucket, harness | | USER\_EMAIL | String | Git commit author email | | | USER\_NAME | String | Git commit author name | Jane Doe | | USERNAME | String | Git provider username (GitHub/Bitbucket only) | janedoe | | REPO | String | Repository name | my-database-repo | | WORKSPACE | String | Workspace slug (Bitbucket only) | mycompany | | ORG\_ID | String | Organization identifier (Harness Code only) | default | | PROJECT\_ID | String | Project identifier (Harness Code only) | my\_project | | ACCOUNT\_ID | String | Account identifier (Harness Code only) | abc123 | | HARNESS\_CODE\_NAME | String | Harness Code username (Harness Code only) | code\_token\_toewv | ![Database DevOps LLM Authoring Environment Variables](/files/E5mDmanycIom6Fqrt5hh) {% hint style="info" %} **NOTE** If the target Git provider is **Harness Code**, ensure that the `REPO_TYPE` is set to **harness**. {% endhint %} You can also set USER\_NAME and USER\_EMAIL to the user who started the pipeline via expression: `<+pipeline.startingUser.email>` and `<+pipeline.startingUser.name>` to automatically attribute the commit and PR to the user who initiated the change authoring process. ![Database DevOps Pipeline Trigger](/files/5RXxznJj8C6q8yHq3z1v) #### Step 3: Author your first DB change 1. Select your Database DevOps project from the Harness dashboard. 2. Navigate to your desired Database Schema. 3. Click on Edit Schema to open the schema details page and select the **Primary DB Instance (Optional)** from the dropdown. 4. Click on the **Author DB Change** tab. 5. In the **Author DB Change** chat window, mention the changes you want to make to the database schema. For example, you can type: ``` Add a new column named Quantity under the tableName Items which is a non-null, with integer datatype. ``` ![Database DevOps LLM Authoring](/files/f0I5ecojMZi23WMS5olE) 6. Once satisfied with the changes, Click on the **Test and Preview** button. The LLM will process your request and generate the appropriate migration artifacts (e.g., changesets or SQL scripts). 7. You can monitor the pipeline execution status via the Chat Window. Once the pipeline completes successfully, a Pull Request will be created in your Git repository with the generated migration artifacts since the pipeline is configured to create a Pull Request. {% tabs %} {% tab title="Visual Pipeline Example" %} ![Open PR with the changes](/files/I880hChT3bwstXoOFIte) {% endtab %} {% tab title="Pipeline YAML Example" %} ```yaml pipeline: name: LLM Authoring identifier: LLM_Authoring projectIdentifier: DBMarlin orgIdentifier: default tags: {} stages: - stage: name: LLm Authoring identifier: LLm_Authoring description: "This stage is used for testing the LLM generated changeset and creating a PR with the changeset file in the git repository." type: Custom spec: execution: steps: - stepGroup: name: LLM identifier: LLM steps: - step: type: DBTestAndPreview name: DBTestAndPreview_1 identifier: DBTestAndPreview_1 spec: connectorRef: docker dbSchema: <+input> dbInstance: <+input> changeset: <+input> timeout: 10m - step: type: Run name: Git Merge identifier: Git_Merge spec: connectorRef: docker image: prontotools/alpine-git-curl shell: Sh command: |- #!/bin/bash set -e cd dbops/DBTestAndPreview_1 git config --global user.email "$USER_EMAIL" git config --global user.name "$USER_NAME" # ── Set remote URL ──────────────────────────────────────────────────────────── case "$REPO_TYPE" in github) ENCODED_EMAIL="${USER_EMAIL//@/%40}" REMOTE_URL="https://${ENCODED_EMAIL}:<+secrets.getValue("github")>@github.com/${USERNAME}/${REPO}.git" ;; bitbucket) REMOTE_URL="https://${USERNAME}:<+secrets.getValue("account.bitbucket")>@bitbucket.org/${WORKSPACE}/${REPO}.git" ;; harness) # Fix 1: hardcoded username, Fix 2: correct host and path format REMOTE_URL="https://${HARNESS_CODE_NAME}:<+secrets.getValue("harness_api_token")>@git.harness.io/${ACCOUNT_ID}/${ORG_ID}/${PROJECT_ID}/${REPO}.git" ;; *) echo "ERROR: Unsupported REPO_TYPE '${REPO_TYPE}'. Must be github, bitbucket, or harness." exit 1 ;; esac # ── Push to a new branch ────────────────────────────────────────────────────── git remote remove origin 2>/dev/null || true git remote add origin "$REMOTE_URL" git fetch origin DEFAULT_BRANCH=$(git remote show origin | grep 'HEAD branch' | awk '{print $NF}') DEFAULT_BRANCH="${DEFAULT_BRANCH:-main}" echo "Default branch: $DEFAULT_BRANCH" echo "── Current directory ──" pwd echo "── Git status ──" git status echo "── All files ──" find . -type f | grep -v ".git" | sort git add . git commit -m "Updating changeset in changelog file" NEW_BRANCH="changelog-update-$(date +%Y%m%d%H%M%S)" git checkout -b "$NEW_BRANCH" git push -u origin "$NEW_BRANCH" # ── Open a Pull Request ─────────────────────────────────────────────────────── case "$REPO_TYPE" in github) PR_RESPONSE=$(curl -s -X POST \ -H "Authorization: token '<+secrets.getValue("github")>'" \ -H "Accept: application/vnd.github+json" \ "https://api.github.com/repos/${USERNAME}/${REPO}/pulls" \ -d "{\"title\":\"Automated changelog update\",\"head\":\"${NEW_BRANCH}\",\"base\":\"${DEFAULT_BRANCH}\"}") PR_URL=$(echo "$PR_RESPONSE" | grep '"html_url"' | head -1 | sed -E 's/.*"html_url": *"([^"]+)".*/\1/') ;; bitbucket) PR_RESPONSE=$(curl -s -X POST \ -u "${USERNAME}:<+secrets.getValue("account.bitbucket")>" \ -H "Content-Type: application/json" \ "https://api.bitbucket.org/2.0/repositories/${WORKSPACE}/${REPO}/pullrequests" \ -d "{\"title\":\"Automated changelog update\",\"source\":{\"branch\":{\"name\":\"${NEW_BRANCH}\"}},\"destination\":{\"branch\":{\"name\":\"${DEFAULT_BRANCH}\"}},\"close_source_branch\":true}") PR_URL=$(echo "$PR_RESPONSE" | grep -o '"html":{"href":"[^"]*"' | sed 's/"html":{"href":"//;s/"$//') ;; harness) PR_RESPONSE=$(curl -s -w "\n__HTTP_STATUS__%{http_code}" -X POST \ "https://app.harness.io/code/api/v1/repos/${REPO}/pullreq?accountIdentifier=${ACCOUNT_ID}&orgIdentifier=${ORG_ID}&projectIdentifier=${PROJECT_ID}" \ -H "Content-Type: application/json" \ -H "x-api-key: <+secrets.getValue("harness_api_token")>" \ -d "{\"title\":\"Automated changelog update\",\"source_branch\":\"${NEW_BRANCH}\",\"target_branch\":\"${DEFAULT_BRANCH}\",\"is_draft\":false}") HTTP_STATUS=$(echo "$PR_RESPONSE" | grep "__HTTP_STATUS__" | sed 's/__HTTP_STATUS__//') BODY=$(echo "$PR_RESPONSE" | grep -v "__HTTP_STATUS__") if [ "$HTTP_STATUS" -lt 200 ] || [ "$HTTP_STATUS" -ge 300 ]; then echo "❌ Failed to create PR (HTTP $HTTP_STATUS)" echo "$BODY" exit 1 fi PR_NUMBER=$(echo "$BODY" | grep -o '"number":[0-9]*' | head -1 | sed 's/"number"://') PR_URL="https://app.harness.io/ng/account/${ACCOUNT_ID}/module/code/orgs/${ORG_ID}/projects/${PROJECT_ID}/repos/${REPO}/pull-requests/${PR_NUMBER}" ;; esac echo "Pull request created: $PR_URL" export CHANGESET_PR_LINK="$PR_URL" envVariables: REPO_TYPE: harness USER_EMAIL: <+pipeline.triggeredBy.email> USER_NAME: <+pipeline.triggeredBy.name> REPO_TYPE: <+input> USER_EMAIL: <+pipeline.triggeredBy.email> USER_NAME: <+pipeline.triggeredBy.name> ACCOUNT_ID: <+input> ORG_ID: <+input> PROJECT_ID: <+input> HARNESS_CODE_NAME: <+input> REPO: <+input> stepGroupInfra: type: KubernetesDirect spec: connectorRef: gke tags: {} ``` {% endtab %} {% endtabs %} {% endtab %} {% endtabs %} ### Next steps * Go to [Get started with changelogs](/database-devops/setup-db-devops/get-started-with-changelogs.md) to understand how Liquibase changesets are structured and executed in Harness Database DevOps. * Go to [Create a pipeline in Database DevOps](/database-devops/use-db-devops/deployment-pipeline-configuration/create-a-pipeline.md) to build the pipeline that runs your LLM-authored migrations. * Go to [Automated rollback for database schemas](/database-devops/use-db-devops/deployment-pipeline-configuration/rollback-and-failure-strategies/rollback-for-database-schemas.md) to configure rollback strategies for failed deployments.