MRE - WizeLine instruction

Prerequisites

• python == 3.11
• aws-cli
• aws-cdk >= 2.24.1
• docker
• node = 18.20+
• npm >= 10.2.3
• git

How to install MRE

Step 1: Install MRE Core

export REGION=[specify the AWS region. For example, us-east-1 or us-west-2]
export VERSION=2.9.0
git clone https://git-codecommit.us-east-1.amazonaws.com/v1/repos/aws-media-replay-engine-gen-ai
cd aws-media-replay-engine-gen-ai
git checkout wl-main
cd deployment
./build-and-deploy.sh --enable-ssm-high-throughput --enable-generative-ai --version $VERSION --region $REGION --verbose [--profile <aws-profile>]

to re-deploy skipping the layer generation

./build-and-deploy.sh --enable-ssm-high-throughput --enable-generative-ai --no-layer --version $VERSION --region $REGION --verbose [--profile <aws-profile>]

Step 2: Install MRE's plugins

1. Install lambda layers:

cd lambda-layers
./deploy.sh [aws-profile] [aws-region]

2. Deploy the sample plugins:

cd aws-media-replay-engine-gen-ai/samples/deployment
./build-and-deploy.sh --app plugin-samples --region $REGION [--profile <aws-profile>]

3. Request model access in Bedrock:

• Titan Embeddings G1 - Text
• Titan Text G1 - Express
• Titan Text Embeddings V2
• Claude 3 Sonnet
• Claude 3 Haiku
• Claude
• Claude Instant
• Embed English

4. In the AWS console open OpenSearch Service an open the aws-mre-collection collection and setup the right principals:

• Go to Data access and open de associated policy aws-mre-collection-access-policy
  • In Rule 1 add these principals
    • aws-mre-dataplane-ChaliceRole -> Sample of principal role: arn:aws:iam::851725259499:role/aws-mre-dataplane-ChaliceRoleF025B2F9-f0yHL9FPdvKy
    • aws-mre-plugin-samples-SegmentNewsRole -> Sample of principal role: arn:aws:iam::851725259499:role/aws-mre-plugin-samples-SegmentNewsRoleC24731FD-louGX5HaYDDn
  • In Rule 2 add this principal
    • aws-mre-search-streaming--streamSummaryJsServiceRol -> Sample of principal role: arn:aws:iam::851725259499:role/aws-mre-search-streaming--streamSummaryJsServiceRol-9A1QHkp2k0GY

5. In the AWS console open the mre-vectorsearch-collection collection and setup the right principals:

• Go to Data access and open de associated policy mre-aoss-data-access-policy
  • In Rule 1 add this principal:
    • aws-mre-search-streaming--streamSummaryJsServiceRol -> Sample of principal role: arn:aws:iam::851725259499:role/aws-mre-search-streaming--streamSummaryJsServiceRol-9A1QHkp2k0GY

Step 3: Install live-news-segmenter app API

1. Front End GitHub Setup

   • Repository URL: https://github.com/wizeline/nab-ui-2024
   • Branch: The branch for the IBC demo is already created as demo-ibc.

   Note: If the current UI repository cannot be used, you should create a new one using the latest version of the code provided in ibc-ui-2024.zip.

2. Create the front end Amplify app

• Create a new Amplify app based on the branch created above
  • Git provider: GitHub (After this AWS console will request your GitHub login)
  • Repository
  • Branch: the on created in the previous step (Next)
  • App Name: live-news-segmenter-frontend-{env} i.e live-news-segmenter-frontend-dev
  • Create a New Service Role: (New Browser tab)
    • AWS Service option
    • Use Case: Amplify (Defaul) (Next)
    • Add permissions (Next)
    • Role Name: amplifyconsole-backend-role-{env} i.e amplifyconsole-backend-role-dev (Create Role)
  • Go back to the Amplify app and select the created role (next)
  • Save and Deploy

Wait until the app is fully created to continue with the next steps

2. Run installation

cd aws-media-replay-engine-gen-ai/samples/deployment
./build-and-deploy.sh --app live-news-segmenter  --region $REGION [--profile <aws-profile>]
./build-and-deploy.sh --app wl-function-samples  --region $REGION [--profile <aws-profile>]

3. Amplify frontend app final setup (Env vars and custom headers)

• Copy the file aws-media-replay-engine-gen-ai/samples/source/live-news-segmenter/ui/cdk/template.frontend-setup.json and rename it to frontend-setup.json
• Update the values of the new created file
  • "webAppURL": You can find the app domain, excluding the protocol, in the app overview. For example: demo-nfl.d9kh5a0nalcse.amplifyapp.com.
  • "webAppId": You can find the app ID in the app overview
  • "region": current deployment region
  • "identityPoolId": You can find this in Cognito / Identity Pools under the created Identity pool which name will resemble nabuie8b66a0e_identitypool_e8b66a0e__staging. Use the Identity Pool ID, which will look similar to us-west-2:03effd58-6db4-4ad5-a155-ccfec63b5311.
  • "userPoolId": You can find this under the User pool overview which will look similar to us-west-2_2xuwGhEq9,
  • "appClientId": You can find this under the User pool / App Integration/ App clients and App clients and analytics, get the Client ID the clientWeb one, it will look similar to 17kvktebes0dp4dbhb76qo1dm2
• Run the Amplify update script

  cd aws-media-replay-engine-gen-ai/samples/source/live-news-segmenter/ui/cdk
  python3 -m venv venv
  source venv/bin/activate
  pip install -r requirements.txt
  python3 init-amplify.py $region [ $profile ] 2>&1
  deactivate
  rm -rf venv

• Check the env vars in Amplify app
• Make sure the build image settings are the correct ones:
  • Build image: Amazon Linux 2023 (default)
  • Build timeout: 30
  • Live package updates: Package: Amplify CLI, Version: Latest
• Additional Configuration
  • Ensure that the Amplify authRole (e.g., amplify-livenewssegmenterfro-prod-b61d7-authRole) has the necessary permissions to invoke Lambda URLs. The policy should resemble the following example:

    {
      "Version": "2012-10-17",
      "Statement": [
          {
              "Action": [
                  "execute-api:Invoke",
                  "execute-api:ManageConnections"
              ],
              "Resource": "arn:aws:execute-api:us-east-1:905418424997:*",
              "Effect": "Allow"
          },
          {
              "Action": "lambda:InvokeFunctionUrl",
              "Resource": "arn:aws:lambda:us-east-1:905418424997:function:*",
              "Effect": "Allow",
              "Sid": "VisualEditor0"
          }
      ]
    }

  • Attach the AmazonSSMFullAccess policy to aws-mre-search-streaming--streamSummary role

Step 4: Plugins configuration:

1. Add a new version of the DetectSpeech plugin:

• Log in to the Media Replay Engine Admin tool.

• Navigate to the "Plugins" option in the left-hand menu.

• Search for the desired plugin.

• Click on "Actions" and select "Add New Version" to update the plugin to the latest version.

• Login into the Media Replay Engine Admin tool and go to plugins option in the left rail, search for the plugin and add the new verion clicking in the Actions

• Update the values of the following configuration parameters: training_bucket_name, output_bucket_name, and input_bucket_name.

  • Search for the actual bucket names in your S3 buckets.
  • Replace the current values with the correct bucket names.
  • The bucket names should resemble these formats: mre-transcribe-files-wizeline and mre-voice-samples-training-wizeline.

2. Update lambda functions below for extra configurations:

• SegmentNews: Also change Dynamo table names

  • Locate the DynamoDB table with the prefix wl-mre-custom-api-GenAiTemplates, create an environment variable GEN_AI_TEMPLATES_TABLE_NAME and assign it the name of the found table. The value should resemble wl-mre-custom-api-GenAiTemplate....

  • Locate the DynamoDB table with the prefix aws-mre-dataplane-PluginResult, create an environment variable MRE_PLUGIN_RESULTS_TABLE_NAME and assign it the name of the found table. The value should resemble aws-mre-dataplane-PluginResult....

  • More env vars:

    OPEN_SEARCH_SERVERLESS_CLUSTER_EP: {{Open search endpoint of aws-mre-collection}} (e.g., el38g1x1i9agzdwkws10.us-east-1.aoss.amazonaws.com)
    OPEN_SEARCH_SERVERLESS_CLUSTER_REGION: us-east-1 or us-west-2 depending on the case
    

• DetectCelebrities:

  • Find dynamoDB table with the prefix wl-mre-custom-api-GenAiTemplates and update lambda parameter with corespoding value genai_templates_table = dynamodb.Table('wl-mre-custom-api-GenAiTemplate...')

• DetectSceneLabels:

  • Find dynamoDB table with the prefix wl-mre-custom-api-GenAiTemplates and update lambda parameter with corespoding value genai_templates_table = dynamodb.Table('wl-mre-custom-api-GenAiTemplates...')

Step 5: Re-deploy Gateway API stack

Due to a cyclic dependency with the live-news-segmenter API stack, it is necessary to re-deploy this stack to import the newly created mre-custom-api-url.

1. Open the file aws-media-replay-engine-gen-ai/source/gateway/infrastructure/stacks/chaliceapp.py and search for the CUSTOM_API_URL key. Comment out the line with the value TBD and uncomment the line below it with the final value.

From this:

self.chalice = Chalice(
            self,
            "MreApiChaliceApp",
            source_dir=RUNTIME_SOURCE_DIR,
            stage_config={
                "environment_variables": {
                    "PLUGIN_URL": Fn.import_value("mre-plugin-api-url"),
                    "MODEL_URL": Fn.import_value("mre-model-api-url"),
                    "PROMPT_CATALOG_URL": Fn.import_value("mre-prompt-catalog-api-url"),
                    "CONTENT_GROUP_URL": Fn.import_value("mre-contentgroup-api-url"),
                    "EVENT_URL": Fn.import_value("mre-event-api-url"),
                    "PROFILE_URL" : Fn.import_value("mre-profile-api-url"),
                    "PROGRAM_URL" : Fn.import_value("mre-program-api-url"),
                    "REPLAY_URL" : Fn.import_value("mre-replay-api-url"),
                    "SYSTEM_URL" : Fn.import_value("mre-system-api-url"),
                    "WORKFLOW_URL" : Fn.import_value("mre-workflow-api-url"),
                    "CUSTOM_PRIORITIES_URL": Fn.import_value("mre-custompriorities-api-url"),
                    "CUSTOM_API_URL": "TBD",
                    # "CUSTOM_API_URL": Fn.import_value("mre-custom-api-url"),
                    "API_AUTH_SECRET_KEY_NAME": "mre_hsa_api_auth_secret",
                },
                "tags": {
                    "Project": "MRE"
                },
                "tags": {"Project": "MRE"},
                "manage_iam_role": False,
                "iam_role_arn": self.chalice_role.role_arn,
            },
        )

To this:

self.chalice = Chalice(
            self,
            "MreApiChaliceApp",
            source_dir=RUNTIME_SOURCE_DIR,
            stage_config={
                "environment_variables": {
                    "PLUGIN_URL": Fn.import_value("mre-plugin-api-url"),
                    "MODEL_URL": Fn.import_value("mre-model-api-url"),
                    "PROMPT_CATALOG_URL": Fn.import_value("mre-prompt-catalog-api-url"),
                    "CONTENT_GROUP_URL": Fn.import_value("mre-contentgroup-api-url"),
                    "EVENT_URL": Fn.import_value("mre-event-api-url"),
                    "PROFILE_URL" : Fn.import_value("mre-profile-api-url"),
                    "PROGRAM_URL" : Fn.import_value("mre-program-api-url"),
                    "REPLAY_URL" : Fn.import_value("mre-replay-api-url"),
                    "SYSTEM_URL" : Fn.import_value("mre-system-api-url"),
                    "WORKFLOW_URL" : Fn.import_value("mre-workflow-api-url"),
                    "CUSTOM_PRIORITIES_URL": Fn.import_value("mre-custompriorities-api-url"),
                    # "CUSTOM_API_URL": "TBD",
                    "CUSTOM_API_URL": Fn.import_value("mre-custom-api-url"),
                    "API_AUTH_SECRET_KEY_NAME": "mre_hsa_api_auth_secret",
                },
                "tags": {
                    "Project": "MRE"
                },
                "tags": {"Project": "MRE"},
                "manage_iam_role": False,
                "iam_role_arn": self.chalice_role.role_arn,
            },
        )

2. Re-deploy the Stack, make sure you have activated the right AWS profile and you're pointing to the right region

cd aws-media-replay-engine-gen-ai/source/gateway
python3 -m venv venv
source venv/bin/activate
pip3 install -U -q -r requirements.txt
cd infrastructure
cdk bootstrap
cdk deploy

3. Revert the code to its original status