Skip to content

CONFIGURING_LIMITS.md

Latest commit

 

History

History
305 lines (225 loc) · 10.8 KB

CONFIGURING_LIMITS.md

File metadata and controls

305 lines (225 loc) · 10.8 KB

Configuring Document Limits and Defaults

This guide explains the configurable limits across the PDF Accessibility solution and how to modify them. Limits exist at two levels:

  1. User-Facing Limits (UI) — Per-user quotas for file uploads, page counts, and file size, managed through Cognito custom attributes in the PDF_accessability_UI repository.
  2. Infrastructure Limits (Backend) — Resource-level settings such as Lambda timeouts, memory, chunk sizes, and concurrency, managed in this repository.

Table of Contents

Section Description
User-Facing Limits (UI) File upload quotas, page limits, and size limits per user
Modifying User Limits via Cognito Console How to change limits for an individual user
Modifying Default Group Limits in Code How to change the defaults assigned to new users
Infrastructure Limits (Backend) Lambda, ECS, Step Functions, and processing settings
PDF-to-HTML Processing Defaults Configuration defaults for the PDF-to-HTML pipeline

User-Facing Limits (UI)

When the PDF Accessibility UI is deployed, each user is assigned limits via Cognito custom attributes. These limits control what users can upload through the web interface.

Custom Cognito Attributes

Attribute Description Default (DefaultUsers)
custom:max_files_allowed Maximum number of files a user can upload 8
custom:max_pages_allowed Maximum number of pages per PDF 10
custom:max_size_allowed_MB Maximum file size in MB 25
custom:total_files_uploaded Current upload count (tracked automatically) 0

Default Limits by User Group

The UI creates three Cognito user groups, each with different default limits:

Attribute DefaultUsers AmazonUsers AdminUsers
max_files_allowed 8 15 100
max_pages_allowed 10 10 2500
max_size_allowed_MB 25 25 1000

These defaults are set when a user first signs up and is automatically assigned to a group. Users with an @amazon.com email are assigned to AmazonUsers; all others go to DefaultUsers. Administrators can move users to AdminUsers manually through the Cognito console.

Modifying User Limits via the Cognito Console

To change limits for a specific user without redeploying:

  1. Open the Amazon Cognito Console.
  2. Select the user pool named PDF-Accessability-User-Pool.
  3. Navigate to Users and search for the user by email or username.
  4. Select the user and scroll to User attributes.
  5. Click Edit and modify any of the following attributes:
    • custom:max_files_allowed — Set the new file upload limit
    • custom:max_pages_allowed — Set the new page limit per PDF
    • custom:max_size_allowed_MB — Set the new file size limit in MB
    • custom:total_files_uploaded — Reset to 0 to restore a user's quota
  6. Click Save changes.

The updated limits take effect immediately on the user's next upload attempt.

Note: Changing a user's group membership (e.g., moving them from DefaultUsers to AdminUsers) will automatically apply that group's default limits via an EventBridge-triggered Lambda function.

Modifying Default Group Limits in Code

To change the default limits that are assigned to all new users, you need to update two Lambda functions in the PDF_accessability_UI repository and redeploy.

File 1: Post-Confirmation Lambda

Path: cdk_backend/lambda/postConfirmation/index.py

This Lambda runs when a new user signs up and sets their initial attributes. Edit the group_attributes dictionary:

group_attributes = {
    DEFAULT_GROUP: {
        'custom:first_sign_in': 'true',
        'custom:total_files_uploaded': '0',
        'custom:max_files_allowed': '8',       # Change this value
        'custom:max_pages_allowed': '10',       # Change this value
        'custom:max_size_allowed_MB': '25'      # Change this value
    },
    AMAZON_GROUP: {
        'custom:first_sign_in': 'true',
        'custom:total_files_uploaded': '0',
        'custom:max_files_allowed': '15',       # Change this value
        'custom:max_pages_allowed': '10',       # Change this value
        'custom:max_size_allowed_MB': '25'      # Change this value
    },
    ADMIN_GROUP: {
        'custom:first_sign_in': 'true',
        'custom:total_files_uploaded': '0',
        'custom:max_files_allowed': '100',      # Change this value
        'custom:max_pages_allowed': '2500',     # Change this value
        'custom:max_size_allowed_MB': '1000'    # Change this value
    }
}

File 2: Update Attributes Groups Lambda

Path: cdk_backend/lambda/UpdateAttributesGroups/index.py

This Lambda runs when a user is moved between groups (via EventBridge) and applies the new group's limits. Edit the GROUP_LIMITS dictionary:

GROUP_LIMITS = {
    'DefaultUsers': {
        'custom:max_files_allowed': '3',        # Change this value
        'custom:max_pages_allowed': '10',       # Change this value
        'custom:max_size_allowed_MB': '25'      # Change this value
    },
    'AmazonUsers': {
        'custom:max_files_allowed': '5',        # Change this value
        'custom:max_pages_allowed': '10',       # Change this value
        'custom:max_size_allowed_MB': '25'      # Change this value
    },
    'AdminUsers': {
        'custom:max_files_allowed': '500',      # Change this value
        'custom:max_pages_allowed': '1500',     # Change this value
        'custom:max_size_allowed_MB': '1000'    # Change this value
    }
}

Important: Make sure the values in both files are consistent for each group. After editing, redeploy the UI stack for changes to take effect. Only newly registered users or users whose group changes will receive the updated defaults. Existing users retain their current attribute values unless manually updated via the Cognito console.

Redeploying After Changes

From the PDF_accessability_UI repository root:

cd cdk_backend
npx cdk deploy

Or re-run the deployment script:

chmod +x deploy.sh
./deploy.sh

Infrastructure Limits (Backend)

The following limits are configured in this repository's infrastructure code (app.py) and affect processing capacity. Modifying these requires a redeployment of the backend stack.

Lambda Function Limits

Lambda Function Timeout Memory File
PDF Splitter 900s (15 min) 1024 MB app.py
PDF Merger 900s (15 min) 1024 MB app.py
Title Generator 900s (15 min) 1024 MB app.py
Pre-Remediation Checker 900s (15 min) 512 MB app.py
Post-Remediation Checker 900s (15 min) 512 MB app.py

To modify, edit the timeout and memory_size parameters in app.py. For example:

pdf_splitter_lambda = lambda_.Function(
    self, 'PdfChunkSplitterLambda',
    runtime=lambda_.Runtime.PYTHON_3_12,
    handler='main.lambda_handler',
    code=lambda_.Code.from_docker_build("lambda/pdf-splitter-lambda"),
    timeout=Duration.seconds(900),    # Maximum is 900 seconds (15 min)
    memory_size=1024                  # In MB, range: 128–10240
)

ECS Task Limits

Task Memory CPU File
Adobe AutoTag 1024 MiB 256 (0.25 vCPU) app.py
Alt Text Generator 1024 MiB 256 (0.25 vCPU) app.py

To modify, edit the memory_limit_mib and cpu parameters in app.py:

adobe_autotag_task_def = ecs.FargateTaskDefinition(
    self, "AdobeAutotagTaskDefinition",
    memory_limit_mib=1024,   # Supported values: 512, 1024, 2048, 4096, ...
    cpu=256,                 # Supported values: 256, 512, 1024, 2048, 4096
    ...
)

Step Functions Limits

Setting Value File
State Machine Timeout 150 minutes app.py
Map State Max Concurrency 100 app.py

To modify:

# State Machine overall timeout
pdf_remediation_state_machine = sfn.StateMachine(
    self, "PdfAccessibilityRemediationWorkflow",
    definition=parallel_accessibility_workflow,
    timeout=Duration.minutes(150),    # Change this value
    ...
)

# Maximum parallel chunk processing
pdf_chunks_map_state = sfn.Map(
    self, "ProcessPdfChunksInParallel",
    max_concurrency=100,              # Change this value
    ...
)

PDF Chunk Size (Pages Per Chunk)

The PDF splitter Lambda splits uploaded PDFs into chunks for parallel processing. The number of pages per chunk is set in lambda/pdf-splitter-lambda/main.py:

# Line 146 in lambda/pdf-splitter-lambda/main.py
chunks = split_pdf_into_pages(pdf_file_content, pdf_file_key, s3, bucket_name, 200)

The last argument (200) is the number of pages per chunk. To process in smaller or larger batches, change this value.

Image Size Limit

The maximum image size for Bedrock model invocation is set in pdf2html/content_accessibility_utility_on_aws/remediate/services/bedrock_client.py:

MAX_IMAGE_SIZE = 4_000_000  # 4 MB — maximum allowed image size in bytes

Images exceeding this limit are automatically resized before being sent to Bedrock.

Redeploying After Infrastructure Changes

After modifying any values in app.py or Lambda source code, redeploy the backend:

cdk deploy

Or re-run the deployment script:

chmod +x deploy.sh
./deploy.sh

PDF-to-HTML Processing Defaults

The PDF-to-HTML pipeline has its own set of configurable defaults defined in pdf2html/content_accessibility_utility_on_aws/utils/config_defaults.yaml:

pdf:
  extract_images: true
  image_format: "png"
  embed_fonts: false
  single_file: false
  continuous: true
  embed_images: false
  exclude_images: false
  cleanup_bda_output: false

audit:
  severity_threshold: "minor"
  detailed_context: true
  skip_automated_checks: false

remediate:
  severity_threshold: "minor"
  model_id: "us.amazon.nova-lite-v1:0"

aws:
  region: null
  create_bda_project: false

These defaults can be overridden in three ways (in order of precedence, highest first):

  1. Command-line arguments — When using the CLI directly
  2. Configuration file — Pass a custom YAML file with --config my-config.yaml
  3. Environment variables — Prefix with DOC_ACCESS_ (e.g., DOC_ACCESS_PDF_IMAGE_FORMAT=jpg)

For details on CLI options and configuration file format, see the pdf2html README.

Support

For questions or assistance with configuration: