How to dev/test in local env

Install pacakges

# local dev
pip install python-dotenv

# pdf parsing
pip install unstructured
pip install "unstructured[pdf]"

# Install poppler in MacOS
conda install poppler

# Langchain updated needed
pip install langchain==0.1.20
pip install langchain-community==0.0.38
pip install langsmith==0.1.57

If local tested python package is successful, need to update package version too. For example, lib/shared/file-import-dockerfile

# FROM quay.io/unstructured-io/unstructured:0.11.2
FROM quay.io/unstructured-io/unstructured:0.13.7  # Use the successful version which tested

Also need to update requirements.txt package version which is fit with tested python package version.

• lib/shared/layers/common/requirements.txt
• lib/shared/file-import-batch-job/requirements.txt

Set environment variables

• Need to deploy cdk first
• Create .env file to use opensearch, kendra, dynamodb, ...

AWS_REGION=us-west-2

LOG_LEVEL=INFO
OPEN_SEARCH_COLLECTION_ENDPOINT=https://mrddjnm6dhio5okny662.us-west-2.aoss.amazonaws.com
OPEN_SEARCH_COLLECTION_ENDPOINT_PORT=443
OPEN_SEARCH_COLLECTION_NAME=llm-chat-dev-genaichatbot-wo
POWERTOOLS_DEV=true
POWERTOOLS_LOGGER_LOG_EVENT=true
POWERTOOLS_SERVICE_NAME=chatbot
WORKSPACES_BY_OBJECT_TYPE_INDEX_NAME=by_object_type_idx
WORKSPACES_TABLE_NAME=llm-chat-devGenAIChatBotStack-RagEnginesRagDynamoDBTablesWorkspacesD2D3C0C4-LQQQMVIYR81J

DOCUMENTS_BY_COMPOUND_KEY_INDEX_NAME=by_compound_key_idx
DOCUMENTS_BY_STATUS_INDEX=by_status_idx
DOCUMENTS_TABLE_NAME=llm-chat-devGenAIChatBotStack-RagEnginesRagDynamoDBTablesDocumentsF6F2B272-4ML769BWD33Y
PROCESSING_BUCKET_NAME=llm-chat-devgenaichatbots-ragenginesdataimportproc-bx13ddia9ztc

API_KEYS_SECRETS_ARN=arn:aws:secretsmanager:us-west-2:723597067299:secret:SharedApiKeysSecret9EA666ED-tuhRyQdJdRfK-2D6gfi
CONFIG_PARAMETER_NAME=CFN-SharedConfig358B4A20-e6zFunNqRa2K
SNS_TOPIC_ARN=arn:aws:sns:us-west-2:723597067299:llm-chat-devGenAIChatBotStack-ChatBotApiRealtimeMessagesTopicCC5C5EA4-7DYgkB2f850D
MESSAGES_TOPIC_ARN=arn:aws:sns:us-west-2:723597067299:llm-chat-devGenAIChatBotStack-ChatBotApiRealtimeMessagesTopicCC5C5EA4-7DYgkB2f850D

AWS_NODEJS_CONNECTION_REUSE_ENABLED=1
GRAPHQL_ENDPOINT=https://4dyslp5onzeotprl6bye7dov6m.appsync-api.us-west-2.amazonaws.com/graphql

SAGEMAKER_RAG_MODELS_ENDPOINT=RagEnginesSageMakerModelMultiAB24AEndpoint6DA7D681-AYVexqh887yR
SESSIONS_BY_USER_ID_INDEX_NAME=byUserId
SESSIONS_TABLE_NAME=llm-chat-devGenAIChatBotStack-ChatBotApiChatDynamoDBTablesSessionsTable92B891E3-1MVBWG1LXY8GY

DEFAULT_KENDRA_INDEX_ID=56afa140-59b3-4402-88b5-8a5ada5aa1be
DEFAULT_KENDRA_INDEX_NAME=llm-chat-dev-genaichatbot-wo
DEFAULT_KENDRA_S3_DATA_SOURCE_BUCKET_NAME=llm-chat-devgenaichatbots-ragengineskendraretrieva-remrfwmaip3o
DEFAULT_KENDRA_S3_DATA_SOURCE_ID=ac6c10fa-24f6-4e83-a276-4acdd9371f0f
DELETE_WORKSPACE_WORKFLOW_ARN=arn:aws:states:us-west-2:723597067299:stateMachine:RagEnginesWorkspacesDeleteWorkspace6908C6DA-pgl1BBgcMJ9u

UPLOAD_BUCKET_NAME=llm-chat-devgenaichatbots-ragenginesdataimportuplo-imywistfc3rb
USER_FEEDBACK_BUCKET_NAME=llm-chat-devgenaichatbots-chatbotapichatbucketsuse-zqpwj1ubar98
WEBSITE_CRAWLING_WORKFLOW_ARN=arn:aws:states:us-west-2:723597067299:stateMachine:RagEnginesDataImportWebsiteCrawlingWorkflowWebsiteCrawling9B1CEC96-5HOPtHm2ydoc
X_ORIGIN_VERIFY_SECRET_ARN=arn:aws:secretsmanager:us-west-2:723597067299:secret:SharedXOriginVerifySecret25-7TEnSxWZdkhh-AcP2Q2
CREATE_OPEN_SEARCH_WORKSPACE_WORKFLOW_ARN=arn:aws:states:us-west-2:723597067299:stateMachine:RagEnginesOpenSearchVectorCreateOpenSearchWorkspace2B2FCA5B-n8E0Ga8tlQZD

How to test with AOSS in local

• Default CDK deployment uses private VPC access for AOSS. Also only some lambda can access to data (data access policy)
• Need to change AOSS config
  • Access available from public : private to public
  • Data access policy (Add IAM user which used in local profile)
• Also need to set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN correctly.

Local test with mock data

There are sample mock data in tests/mock-data. The variable is depends on each AWS account, you must change the value for local testing.

Some hard-coded params

• lib/shared/file-import-batch-job/main.py

# If true, complex workspace with PDF will add document summarization to the DB (DynamoDB)
DOC_TYPE_COMPLEX_PDF_SUMMARIZATION = True

• lib/shared/layers/python-sdk/python/genai_core/langchain/workspace_retriever.py

# If true, add document title and summary which summarized using LLM for search result.
ADD_DOC_SUMMARY = True
DOC_TOP_K = 3

• lib/shared/layers/python-sdk/python/genai_core/opensearch/query.py

# Hybrid search score weight
VECTOR_WEIGHT = 0.55
KEYWORD_WEIGHT = 0.45

Frontend app local testing

Check the guide for locally run frontend application (React App)

• Need to modify the distribution url in getDistributionUrl() function, which located in lib/user-interface/react-app/src/common/constants.ts file because it basically uses same CF distribution url with website.

How to modify API for frontend

To modify GraphQL API, need to modify schema.graphql first. And use command below to generate API.ts file.

# From the project root
$ npm run gen   # Check the package.json

# Or you can use codegen directly
$ amplify codegen

If you use MacOS

If you use finch, need to check storage.

• Need to increase storage size when there's no space left on device error to build container image for file import job.
  • Check the document: https://runfinch.com/docs/managing-finch/macos/disk-management/#disk-size
• Need to use CDK_DOCKER env variables when running deployment command.
  • CDK_DOCKER=finch npx cdk deploy
  • Check the scripts/cdk_deploy_finch.sh

If you use docker, make sure there are enough space.

• If not, use docker system prune -a to clear all docker env.

x86/arm64 architecture

• It uses cross platform build like dockerx to build docker image. But sometimes, it won't work.
• You can specify lambda architecture for docker image build in lib/shared/index.ts
  • If you use apple silicon : const lambdaArchitecture = lambda.Architecture.ARM_64;
  • Normal X86 server : const lambdaArchitecture = lambda.Architecture.X86_64;