Stripe
Hosted subscription checkout, payment-method management, invoices, renewal, cancellation, and tax calculation when configured.
Data categories
billing references, subscription/payment status, and billing details supplied directly to Stripe
Noezra Provider Register
Subprocessors And AI Processing
Noezra uses the providers below to operate core study, AI, speech, storage, analytics, and research discovery features. Private Workspace Content is not used for shared model training at launch.
AWS infrastructure is primarily configured in Sydney (ap-southeast-2). Bedrock Data Automation is only an emergency fallback and uses the approved APAC cross-region profile, so eligible emergency extraction may be processed in another AWS APAC region. Provider locations and retention depend on the active service configuration and contract; this register does not promise exclusive residency or zero provider retention.
Provider-operated regions under the active Stripe account configuration and contract
Amazon Web Services
Application hosting, PostgreSQL database, object storage, queues, authentication, account email, Titan embeddings, local document processing with PDF Inspector, self-hosted MinerU 3.4 OCR on AWS Batch, Bedrock Data Automation only as an emergency APAC fallback, isolated Python execution for requested research visuals, monitoring, and security.
Data categories
account data, uploaded files, layouts, chunks, embedding input, notes, highlights, citations, logs, and selected normalized datasets, generated calculation recipes and layout dimensions for research visuals
Microsoft Foundry
Managed AI deployment and processing for chat, grounded answers, summaries, generated chat titles, research visual planning and requested illustrative image generation.
Data categories
prompts, relevant document passages, chat history, selected model, citation metadata, bounded selected research data and source passages, layout dimensions, illustration prompts and returned images
OpenAI
Chat and research visual planning, and illustrative image generation through Microsoft Foundry for the requested AI feature.
Data categories
prompts, relevant document passages, chat history, citation metadata, bounded selected research data, layout dimensions, illustration prompts and returned images
Cloudflare
Public DNS management and, where proxying is enabled, edge delivery, traffic protection, and security.
Data categories
hostname and DNS records; when proxied: IP address, request and device metadata, security events, and content transiting the edge
Google Sign-In through Amazon Cognito when selected by the user.
Data categories
Google account identifier, name, email address, and profile information supplied for sign-in
ElevenLabs
Text-to-speech audio generation and browser delivery of ElevenLabs-hosted public-site assets.
Data categories
selected reading text and voice settings; for public browser asset delivery: IP address and request/device metadata
FreeDictionaryAPI
User-requested dictionary definitions and pronunciation lookup.
Data categories
lookup terms, IP address, and browser request metadata
PostHog
Optional product analytics.
Data categories
pseudonymous feature events and device/session metadata only; no private content
Adobe Fonts and rsms.me
Web-font delivery for the Noezra interface.
Data categories
IP address, request headers, and device or browser metadata
Microsoft Bing through Foundry web search
Public web discovery leads for requested paper searches, verified against scholarly indexes before display.
Data categories
public topics and verified public paper metadata; bounded, minimized concepts from available authorised chat sources when you request a paper search; uploaded passages and private notes are not sent to search
Semantic Scholar and public research indexes
Research discovery and paper metadata lookup.
Data categories
search terms, public source metadata, DOI/arXiv identifiers, and saved public references
AI Processing Routes
These are separate processing routes. Provider retention and training practices remain governed by verified Microsoft, OpenAI, and AWS terms and the active service configurations. Anthropic is no longer used for new AI processing. Earlier processing remains subject to the applicable provider terms and retention requirements; model retirement does not itself delete those records.
Local PDF Inspector and self-hosted MinerU 3.4
Local processing first; self-hosted OCR in Sydney
PDF Inspector processes PDFs locally. Only pages needing OCR use self-hosted MinerU 3.4 when safe. If selected-page OCR or composition cannot safely complete, whole-document MinerU is used before any eligible emergency Bedrock Data Automation decision. BDA is never the primary, default, or normal extraction route.
Microsoft Foundry (OpenAI)
United States data zone or global, depending on the deployment
Processes chat prompts, conversation context, and relevant document passages for requested AI features. Research visual planning can include bounded selected data, source passages and layout dimensions. Enabled illustration generation receives the requested illustration prompt and returns an image.
Amazon Bedrock Titan Embeddings V2
Sydney (ap-southeast-2)
Processes document chunks and retrieval queries in the existing AWS production region.
Retention Summary
Account data
Retained while the account is active; deleted or de-identified after account deletion subject to legal requirements.
Private workspace content
Retained while the account is active or until the user deletes it. Derived layouts, chunks, embeddings, and audio cache should be removed with the source.
Analytics
Optional, pseudonymous, and controlled by user settings and analytics workspace retention settings.
Backups and logs
Retained only as reasonably necessary for security, reliability, legal, and operational needs.