Merge pull request #3 from AbeOmor/main

Update docs and files based on comments
Azure · Sep 12, 2023 · 1284d58 · 1284d58
2 parents 10cea03 + 35cf454
commit 1284d58
Show file tree

Hide file tree

Showing 5 changed files with 72 additions and 52 deletions.
diff --git a/.github/workflows/deploy-pf-online-endpoint-pipeline.yml b/.github/workflows/deploy-pf-online-endpoint-pipeline.yml
@@ -11,42 +11,47 @@ on:
 env: 
   GROUP: ${{secrets.GROUP}}
   WORKSPACE: ${{secrets.WORKSPACE}}
-  SUBSCRIPTION: ${{secrets.SUBSCRIPTION}}
+  SUBSCRIPTION_ID: ${{secrets.SUBSCRIPTION_ID}}
   RUN_NAME: web_classification_variant_1_20230816_215600_605116
   EVAL_RUN_NAME: classification_accuracy_eval_default_20230821_111809_077086
   ENDPOINT_NAME: web-classification
+  DEPLOYMENT_NAME: blue
 
 jobs:
   create-endpoint-and-deploy-pf:
     runs-on: ubuntu-latest
     if: ${{ github.event_name == 'workflow_dispatch' || github.event.workflow_run.conclusion == 'success' }} 
     steps:
-    - name: check out repo
+    - name: Check out repo
       uses: actions/checkout@v2
-    - name: install az ml extension
+    - name: Install az ml extension
       run: az extension add -n ml -y
-    - name: azure login
+    - name: Azure login
       uses: azure/login@v1
       with:
         creds: ${{secrets.AZURE_CREDENTIALS}}
-    - name: list current directory
+    - name: List current directory
       run: ls
-    - name: set default subscription
+    - name: Set default subscription
       run: |
          az account set -s ${{env.SUBSCRIPTION}}
-    - name: create Hash
+    - name: Create Hash
       run: echo "HASH=$(echo -n $RANDOM | sha1sum | cut -c 1-6)" >> "$GITHUB_ENV"
-    - name: create unique endpoint name   
+    - name: Create unique endpoint name   
       run: echo "ENDPOINT_NAME=$(echo 'web-classification-'$HASH)" >> "$GITHUB_ENV"
-    - name: display endpoint name
+    - name: Display endpoint name
       run: echo "Endpoint name is:" ${{env.ENDPOINT_NAME}}
-    - name: setup endpoint
+    - name: Setup endpoint
       run: az ml online-endpoint create --file promptflow/deployment/endpoint.yaml  --name ${{env.ENDPOINT_NAME}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: setup deployment
+    - name: Update deployment PRT_CONFIG variable
+      run: |
+        PRT_CONFIG_OVERRIDE=deployment.subscription_id=${{ env.SUBSCRIPTION }},deployment.resource_group=${{ env.GROUP }},deployment.workspace_name=${{ env.WORKSPACE }},deployment.endpoint_name=${{ env.ENDPOINT_NAME }},deployment.deployment_name=${{ env.DEPLOYMENT_NAME }}
+        sed -i "s/PRT_CONFIG_OVERRIDE:.*/PRT_CONFIG_OVERRIDE: $PRT_CONFIG_OVERRIDE/g" promptflow/deployment/deployment.yaml
+    - name: Setup deployment
       run: az ml online-deployment create --file promptflow/deployment/deployment.yaml --endpoint-name ${{env.ENDPOINT_NAME}} --all-traffic -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: check the status of the endpoint
+    - name: Check the status of the endpoint
       run: az ml online-endpoint show -n ${{env.ENDPOINT_NAME}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: check the status of the deployment
+    - name: Check the status of the deployment
       run: az ml online-deployment get-logs --name blue --endpoint-name ${{env.ENDPOINT_NAME}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: invoke model
+    - name: Invoke model
       run: az ml online-endpoint invoke --name ${{env.ENDPOINT_NAME}} --request-file promptflow/deployment/sample-request.json  -g ${{env.GROUP}} -w ${{env.WORKSPACE}} 
diff --git a/.github/workflows/run-eval-pf-pipeline.yml b/.github/workflows/run-eval-pf-pipeline.yml
@@ -8,74 +8,74 @@ on:
 env: 
   GROUP: ${{secrets.GROUP}}
   WORKSPACE: ${{secrets.WORKSPACE}}
-  SUBSCRIPTION: ${{secrets.SUBSCRIPTION}}
+  SUBSCRIPTION_ID: ${{secrets.SUBSCRIPTION_ID}}
   RUN_NAME: web_classification_variant_1_20230816_215600_605116
   EVAL_RUN_NAME: classification_accuracy_eval_default_20230821_111809_077086
 
 jobs:
   login-and-run-and-evalpf:
     runs-on: ubuntu-latest 
     steps:
-    - name: check out repo
+    - name: Check out repo
       uses: actions/checkout@v2
-    - name: install az ml extension
+    - name: Install az ml extension
       run: az extension add -n ml -y
-    - name: azure login
+    - name: Azure login
       uses: azure/login@v1
       with:
         creds: ${{secrets.AZURE_CREDENTIALS}}
     - name: Set up Python
       uses: actions/setup-python@v2
       with:
         python-version: '3.11.4'
-    - name: list current directory
+    - name: List current directory
       run: ls
-    - name: install promptflow
+    - name: Install promptflow
       run: pip install -r promptflow/web-classification/requirements.txt
-    - name: run promptflow
+    - name: Run promptflow
       run: |
         pfazure run create -f promptflow/web-classification/run.yml --subscription ${{env.SUBSCRIPTION}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}} --stream > promptflow/llmops-helper/run_info.txt
         cat promptflow/llmops-helper/run_info.txt
-    - name: set run name
+    - name: Set run name
       run: |
         echo "RUN_NAME=$(python promptflow/llmops-helper/parse_run_output.py run_info.txt)" >> "$GITHUB_ENV"
-    - name: show the current run name
+    - name: Show the current run name
       run: echo "Run name is:" ${{env.RUN_NAME}}
-    - name: show promptflow results
+    - name: Show promptflow results
       run: pfazure run show-details --name ${{env.RUN_NAME}} --subscription ${{env.SUBSCRIPTION}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: run promptflow evaluations
+    - name: Run promptflow evaluations
       run: pfazure run create -f promptflow/web-classification/run_evaluation.yml --run ${{env.RUN_NAME}} --subscription ${{env.SUBSCRIPTION}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}} --stream > promptflow/llmops-helper/eval_info.txt 
-    - name: get eval run name
-      run: export EVAL_RUN_NAME=$(python promptflow/llmops-helper/parse_run_output.py eval_info.txt)
-    - name: show promptflow details
+    - name: Get eval run name
+      run: echo "EVAL_RUN_NAME=$(python promptflow/llmops-helper/parse_run_output.py eval_info.txt)" >> "$GITHUB_ENV"
+    - name: Show promptflow details
       run: pfazure run show-details --name ${{env.EVAL_RUN_NAME}} --subscription ${{env.SUBSCRIPTION}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
-    - name: show promptflow metrics
+    - name: Show promptflow metrics
       run: pfazure run show-metrics --name ${{env.EVAL_RUN_NAME}} --subscription ${{env.SUBSCRIPTION}} -g ${{env.GROUP}} -w ${{env.WORKSPACE}} > promptflow/llmops-helper/eval_result.json 
-  
+
   assert-and-register-model:
     needs: login-and-run-and-evalpf
     runs-on: ubuntu-latest 
     steps:
-    - name: check out repo
+    - name: Check out repo
       uses: actions/checkout@v2
-    - name: install az ml extension
+    - name: Install az ml extension
       run: az extension add -n ml -y
-    - name: azure login
+    - name: Azure login
       uses: azure/login@v1
       with:
         creds: ${{secrets.AZURE_CREDENTIALS}}
     - name: Set up Python
       uses: actions/setup-python@v2
       with:
         python-version: '3.11.4'
-    - name: set default subscription
+    - name: Set default subscription
       run: |
             az account set -s ${{env.SUBSCRIPTION}}
-    - name: list current directory
+    - name: List current directory
       run: ls
-    - name: install promptflow
+    - name: Install promptflow
       run: pip install -r promptflow/web-classification/requirements.txt
-    - name: get assert eval results
+    - name: Get assert eval results
       id: jobMetricAssert
       run: |
             export ASSERT=$(python promptflow/llmops-helper/assert.py result.json 0.6) # NOTE <file>.json is the file name and decimal is the threshold for the assertion
@@ -88,6 +88,6 @@ jobs:
               echo "::warning::Prompt flow run didn't meet quality bar."
               echo "::set-output name=result::false"
             fi
-    - name: register promptflow model
+    - name: Register promptflow model
       if: ${{ steps.jobMetricAssert.outputs.result == 'true' }}
       run: az ml model create --file promptflow/deployment/model.yaml  -g ${{env.GROUP}} -w ${{env.WORKSPACE}}
diff --git a/docs/e2e_llmops_with_promptflow.md b/docs/e2e_llmops_with_promptflow.md
@@ -84,9 +84,7 @@ Before you can set up an Prompt flow project with Machine Learning, you need to
 1. Close the Cloud Shell once the service principals are created. 
 
 ## Set up GitHub repo
-1. Fork example repo. [LLMOps Demo Template Repo](https://github.com/Azure/llmops-gha-demo/) in your GitHub organization
-
-1. Go to https://github.com/Azure/llmops-gha-demo/fork to fork the LLMOps demo repo into your GitHub org. This repo has reusable LLMOps code that can be used across multiple projects. 
+1. Fork example repo. [LLMOps Demo Template Repo](https://github.com/Azure/llmops-gha-demo/fork) in your GitHub organization. This repo has reusable LLMOps code that can be used across multiple projects. 
 
 ## Add secret to GitHub Repo
 
@@ -103,10 +101,16 @@ Before you can set up an Prompt flow project with Machine Learning, you need to
     ![Screenshot of GitHub Secrets String 1.](media/e2e-llmops/github-secrets-string.png)
 
 1. Add each of the following additional GitHub secrets using the corresponding values from the service principal output as the content of the secret:  
-    - **GROUP**  
-    - **WORKSPACE**  
-    - **SUBSCRIPTION**  
+    - **GROUP**: \<Resource Group Name\>
+    - **WORKSPACE**: \<Azure ML Workspace Name\>
+    - **SUBSCRIPTION_ID**: \<Subscription ID\>
 
+    |Variable  | Description  |
+    |---------|---------|
+    |GROUP     |      Name of resource group    |
+    |SUBSCRIPTION_ID     |    Subscription ID of your workspace    |
+    |WORKSPACE     |     Name of Azure Machine Learning workspace     |  
+
 > [!NOTE]
 > This finishes the prerequisite section and the deployment of the solution accelerator can happen accordingly.
 
@@ -120,6 +124,13 @@ Please go to workspace portal, click `Prompt flow` -> `Connections` -> `Create`,
 
 ![Screenshot of Prompt flow Connections](media/e2e-llmops/promptflow_connections.png)
 
+## Setup runtime for Prompt flow 
+Prompt flow's runtime provides the computing resources required for the application to run, including a Docker image that contains all necessary dependency packages.
+
+In this guide, we will use a runtime to run your prompt flow. You need to create your own [Prompt flow runtime](https://learn.microsoft.com/azure/machine-learning/prompt-flow/how-to-create-manage-runtime)
+
+Please go to workspace portal, click `Prompt flow` -> `Runtime` -> `Add`, then follow the instruction to create your own connections
+
 ## Setup variables with for Prompt flow and GitHub Actions 
 
 1. Clone repo to your local machine
@@ -131,13 +142,7 @@ Please go to workspace portal, click `Prompt flow` -> `Connections` -> `Create`,
 ### Update workflow to connect to your Azure Machine Learning workspace
 1. You need to update `run-eval-pf-pipeline.yml` and `deploy-pf-online-endpoint-pipeline.yml` to connect to your Azure Machine Learning workspace. You'll need to update the CLI setup file variables to match your workspace. 
 1. In your cloned repository, go to `.github/workflow/`. 
-1. Edit `env` section in the `run-eval-pf-pipeline.yml` and `deploy-pf-online-endpoint-pipeline.yml` and update these variables in the file. 
-
-    |Variable  | Description  |
-    |---------|---------|
-    |GROUP     |      Name of resource group    |
-    |SUBSCRIPTION     |    Subscription of your workspace    |
-    |WORKSPACE     |     Name of Azure Machine Learning workspace     | 
+1. Verify `env` section in the `run-eval-pf-pipeline.yml` and `deploy-pf-online-endpoint-pipeline.yml` refers to the workspace secrets you added in the previous step. 
 
 ### Update run.yml with your connections and runtime
 

diff --git a/promptflow/deployment/deployment.yaml b/promptflow/deployment/deployment.yaml
@@ -25,7 +25,7 @@ environment_variables:
   PROMPTFLOW_RUN_MODE: serving
 
   # for pulling connections from workspace
-  PRT_CONFIG_OVERRIDE: deployment.subscription_id=<sub-id>,deployment.resource_group=o<resource-group>,deployment.workspace_name=<workspace-name>,deployment.endpoint_name=<endpoint-name>,deployment.deployment_name=<deployment-name>
+  PRT_CONFIG_OVERRIDE: deployment.subscription_id=<sub-id>,deployment.resource_group=<resource-group>,deployment.workspace_name=<workspace-name>,deployment.endpoint_name=<endpoint-name>,deployment.deployment_name=<deployment-name>
 
   # (Optional) When there are multiple fields in the response, using this env variable will filter the fields to expose in the response.
   # For example, if there are 2 flow outputs: "answer", "context", and I only want to have "answer" in the endpoint response, I can set this env variable to '["answer"]'.

diff --git a/promptflow/llmops-helper/service_principal_script.sh b/promptflow/llmops-helper/service_principal_script.sh
@@ -0,0 +1,10 @@
+projectName="<your project name>"
+roleName="Contributor"
+subscriptionId="<subscription Id>"
+environment="<Prod>" #First letter should be capitalized
+servicePrincipalName="Azure-ARM-${environment}-${projectName}"
+# Verify the ID of the active subscription
+echo "Using subscription ID $subscriptionID"
+echo "Creating SP for RBAC with name $servicePrincipalName, with role $roleName and in scopes     /subscriptions/$subscriptionId"
+az ad sp create-for-rbac --name $servicePrincipalName --role $roleName --scopes /subscriptions/$subscriptionId --sdk-auth 
+echo "Please ensure that the information created here is properly save for future use."