Read secrets directly into GitHub secrets from ENV_VARS

.github/actions/aml_real_deployment/action.yml

@@ -24,24 +24,10 @@ inputs:
   KEY_VAULT_NAME:
     description: "Key vault name"
     required: true
-  ENV_VARS:
-    description: 'The API key for authentication'
-    required: false
 
 runs:
   using: composite
-  steps:
-    - name: Create .env file
-      shell: bash
-      run: |
-        echo "${{ inputs.ENV_VARS }}" >> .env
-        for var in $(cat .env); do
-          echo "$var" >> $GITHUB_ENV
-        done
-    - name: load .env file
-      shell: bash
-      run: python -c "from dotenv import load_dotenv; load_dotenv()"
-
+  steps:      
     - name: Provision Managed Endpoint
       uses: ./.github/actions/execute_script
       with:

.github/actions/kubernetes_deployment/action.yml

@@ -29,16 +29,6 @@ inputs:
 runs:
   using: composite
   steps:
-    - name: Create .env file
-      shell: bash
-      run: |
-        echo "${{ secrets.env_vars }}" >> .env
-        for var in $(cat .env); do
-          echo "$var" >> $GITHUB_ENV
-        done
-    - name: load .env file
-      shell: bash
-      run: python -c "from dotenv import load_dotenv; load_dotenv()"
     - name: Provision Kubernetes Online Endpoint
       uses: ./.github/actions/execute_script
       with:

.github/actions/webapp_deployment/action.yml

@@ -9,20 +9,10 @@ inputs:
   DEPLOY_ENVIRONMENT:
     description: "env stage e.g. dev, test, prod"
     required: true
-  ENV_VARS:
-    description: "env stage e.g. dev, test, prod"
-    required: true
 
 runs:
   using: composite
   steps:
-    - name: Convert .env to environment variables
-      shell: bash
-      run: |
-        while IFS='=' read -r key value; do
-          echo "$key=$value" >> $GITHUB_ENV
-        done <<< "${{ inputs.env_vars }}"
-
     - name: create docker image
       shell: bash
       run: use_case_base_path=${{ inputs.USE_CASE_BASE_PATH }} deploy_environment=${{ inputs.DEPLOY_ENVIRONMENT }} build_id=${{ github.run_id }} ./llmops/common/scripts/gen_docker_image.sh 

.github/workflows/platform_cd_dev_workflow.yml

@@ -58,20 +58,25 @@ jobs:
         uses: azure/login@v1
         with:
           creds: ${{ secrets.azure_credentials }}
+
       - name: Configure Azure ML Agent
         uses: ./.github/actions/configure_azureml_agent
         with:
           versionSpec: "3.9"
           base_path: ${{ inputs.use_case_base_path }}
 
-      - name: Create .env file
+      - name: Load secrets from ENV_VARS
         run: |
-          echo "${{ secrets.ENV_VARS }}" >> .env
-            for var in $(cat .env); do
-              echo "$var" >> $GITHUB_ENV
-            done
-      - name: load .env file
-        run: python -c "from dotenv import load_dotenv; load_dotenv()"
+          set -e # fail on error
+          IFS=$'\n' # Set internal field separator to new line
+          for var in $(echo "${{ secrets.ENV_VARS }}"); do
+            KEY=$(echo "$var" | cut -d '=' -f 1)
+            VALUE=$(echo "$var" | cut -d '=' -f 2-)
+            echo "::add-mask::$VALUE"
+            echo "$KEY=$VALUE" >> $GITHUB_ENV
+          done
+        shell: bash
+
       - name: load the current Azure subscription details
         id: subscription_details
         shell: bash
@@ -123,7 +128,6 @@ jobs:
           WORKSPACE_NAME: ${{ vars.WORKSPACE_NAME }}
           KEY_VAULT_NAME: ${{ vars.KEY_VAULT_NAME }}
           MODEL_VERSION: $MODEL_VERSION
-          ENV_VARS: "${{ secrets.ENV_VARS }}"
 
       #=====================================
       # Executes Managed kubernetes deployment when parameter deployment_type == 'aks'
@@ -155,12 +159,6 @@ jobs:
               --env_name ${{ inputs.env_name }} \
               --base_path ${{ inputs.use_case_base_path }} \
 
-      - name: Read .env file
-        id: env_vars
-        run: |
-          LOCAL_VARS=$(cat .env)
-          echo "local_vars=$LOCAL_VARS" >> $GITHUB_OUTPUT
-
         #=====================================
         # Executes Azure Webapp deployment when parameter 
         # deployment_type == 'webapp' using docker image
@@ -172,7 +170,6 @@ jobs:
         with:
           USE_CASE_BASE_PATH: ${{ inputs.use_case_base_path }}
           DEPLOY_ENVIRONMENT: ${{ inputs.env_name }}
-          ENV_VARS: ${{ steps.env_vars.outputs.local_vars }}
         env:
           REGISTRY_DETAILS: ${{ secrets.registry_details }}
 

.github/workflows/platform_ci_dev_workflow.yml

@@ -52,14 +52,18 @@ jobs:
           versionSpec: "3.9"
           base_path: ${{ inputs.use_case_base_path }}
 
-      - name: Create .env file
+      - name: Load secrets from ENV_VARS
         run: |
-          echo "${{ secrets.ENV_VARS }}" >> .env
-          for var in $(cat .env); do
-            echo "$var" >> $GITHUB_ENV
+          set -e # fail on error
+          IFS=$'\n' # Set internal field separator to new line
+          for var in $(echo "${{ secrets.ENV_VARS }}"); do
+            KEY=$(echo "$var" | cut -d '=' -f 1)
+            VALUE=$(echo "$var" | cut -d '=' -f 2-)
+            echo "::add-mask::$VALUE"
+            echo "$KEY=$VALUE" >> $GITHUB_ENV
           done
-      - name: load .env file
-        run: python -c "from dotenv import load_dotenv; load_dotenv()"
+        shell: bash
+
       - name: load the current Azure subscription details
         id: subscription_details
         shell: bash

.github/workflows/platform_pr_dev_workflow.yml

@@ -15,7 +15,7 @@ on:
       azure_credentials:
         description: "service principal auth to Azure"
         required: true
-      env_vars:
+      ENV_VARS:
         description: "env vars for the flow"
         required: false
 
@@ -43,19 +43,24 @@ jobs:
         uses: azure/login@v1
         with:
           creds: ${{ secrets.azure_credentials }}
+
       - name: Configure Azure ML Agent
         uses: ./.github/actions/configure_azureml_agent
         with:
           versionSpec: "3.9"
           base_path: ${{ inputs.use_case_base_path }}
-      - name: Create .env file
+
+      - name: Load secrets from ENV_VARS
         run: |
-          echo "${{ secrets.env_vars }}" >> .env
-          for var in $(cat .env); do
-            echo "$var" >> $GITHUB_ENV
+          set -e # fail on error
+          IFS=$'\n' # Set internal field separator to new line
+          for var in $(echo "${{ secrets.ENV_VARS }}"); do
+            KEY=$(echo "$var" | cut -d '=' -f 1)
+            VALUE=$(echo "$var" | cut -d '=' -f 2-)
+            echo "::add-mask::$VALUE"
+            echo "$KEY=$VALUE" >> $GITHUB_ENV
           done
-      - name: load .env file
-        run: python -c "from dotenv import load_dotenv; load_dotenv()"
+        shell: bash
 
       - name: load the current Azure subscription details
         id: subscription_details

chat_with_pdf/data/bert-paper-qna-1-line.jsonl

@@ -1 +1 @@
-{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf","chat_history":[],"question":"What is the name of the new language representation model introduced in the document?","answer":"BERT","context":"We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers.","config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"aoai","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
+{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf","chat_history":[],"question":"What is the name of the new language representation model introduced in the document?","answer":"BERT","context":"We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers.","config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"gpt-35-turbo","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}

chat_with_pdf/data/bert-paper-qna-3-line.jsonl

@@ -1,3 +1,3 @@
-{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf", "chat_history":[], "question": "What is the main difference between BERT and previous language representation models?", "groundtruth": "BERT is designed to pretrain deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers.", "context": "Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pretrain deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers.", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"aoai","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
-{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf", "chat_history":[], "question": "What is the size of the vocabulary used by BERT?", "groundtruth": "30,000", "context": "We use WordPiece embeddings (Wu et al., 2016) with a 30,000 token vocabulary.", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"aoai","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
-{"pdf_url":"https://grs.pku.edu.cn/docs/2018-03/20180301083100898652.pdf", "chat_history":[], "question": "论文写作中论文引言有什么注意事项？", "groundtruth":"", "context":"", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"aoai","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
+{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf", "chat_history":[], "question": "What is the main difference between BERT and previous language representation models?", "groundtruth": "BERT is designed to pretrain deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers.", "context": "Unlike recent language representation models (Peters et al., 2018a; Radford et al., 2018), BERT is designed to pretrain deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers.", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"gpt-35-turbo","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
+{"pdf_url":"https://arxiv.org/pdf/1810.04805.pdf", "chat_history":[], "question": "What is the size of the vocabulary used by BERT?", "groundtruth": "30,000", "context": "We use WordPiece embeddings (Wu et al., 2016) with a 30,000 token vocabulary.", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"gpt-35-turbo","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}
+{"pdf_url":"https://grs.pku.edu.cn/docs/2018-03/20180301083100898652.pdf", "chat_history":[], "question": "论文写作中论文引言有什么注意事项？", "groundtruth":"", "context":"", "config":{"EMBEDDING_MODEL_DEPLOYMENT_NAME":"text-embedding-ada-002","CHAT_MODEL_DEPLOYMENT_NAME":"gpt-35-turbo","PROMPT_TOKEN_LIMIT":2000,"MAX_COMPLETION_TOKENS":256,"CHUNK_SIZE":1024,"CHUNK_OVERLAP":64}}

chat_with_pdf/flows/evaluation/flow.dag.yaml

@@ -422,7 +422,7 @@ nodes:
     type: code
     path: rag_groundedness_prompt.jinja2
   inputs:
-    deployment_name: aoai
+    deployment_name: gpt-35-turbo
     temperature: 0
     top_p: 1
     max_tokens: 1000
@@ -443,7 +443,7 @@ nodes:
     type: code
     path: rag_retrieval_prompt.jinja2
   inputs:
-    deployment_name: aoai
+    deployment_name: gpt-35-turbo
     temperature: 0
     top_p: 1
     max_tokens: 1000
@@ -463,7 +463,7 @@ nodes:
     type: code
     path: rag_generation_prompt.jinja2
   inputs:
-    deployment_name: aoai
+    deployment_name: gpt-35-turbo
     temperature: 0
     top_p: 1
     max_tokens: 1000

chat_with_pdf/flows/standard/flow.dag.yaml

@@ -16,7 +16,7 @@ inputs:
     type: object
     default:
       EMBEDDING_MODEL_DEPLOYMENT_NAME: text-embedding-ada-002
-      CHAT_MODEL_DEPLOYMENT_NAME: aoai
+      CHAT_MODEL_DEPLOYMENT_NAME: gpt-35-turbo
       PROMPT_TOKEN_LIMIT: 3000
       MAX_COMPLETION_TOKENS: 1024
       VERBOSE: true