ArchiveLM Subprocessor List
Effective date: 2026-05-12 Last updated: 2026-05-12
To provide the ArchiveLM service, Michael De La Guera ("we") engages the third-party service providers below ("subprocessors") to process personal data on our behalf. We have entered into data-processing agreements with each subprocessor that include the protections required by applicable law.
This page is updated when subprocessors change. We will give existing customers reasonable notice of new or replaced subprocessors where required by contract or law; you may object on documented grounds to any new subprocessor by contacting legal@archivelm.com.
Active subprocessors
| Subprocessor | Role | Data processed | Region | DPA |
|---|---|---|---|---|
| Supabase Inc. | Authentication, primary database (Postgres + pgvector), file storage | Account credentials, profile metadata, uploaded documents, extracted text, vector embeddings, beta application data | United States (AWS US-East / EU regions configurable) | Supabase DPA |
| Vercel Inc. | Application hosting, edge networking, serverless function execution, deployment, analytics | Request data, IP addresses (truncated for analytics), session cookies, page-view events | United States and global edge | Vercel DPA |
| Google LLC (Google AI / Gemini) | AI extraction (OCR), structuring, and historical context generation | Document images and extracted text passages submitted as inputs to the model | United States | Google Cloud DPA — Cloud Data Processing Addendum |
| OpenRouter, Inc. | AI gateway routing chat queries to underlying model providers (Anthropic, OpenAI) for the AI Librarian, semantic search, and Research Lab features | User queries and retrieved excerpts from your collection submitted to the routed model | United States; routes to model-provider regions | OpenRouter Privacy |
| Anthropic, PBC | LLM inference for chat and reasoning tasks (routed via OpenRouter) | Same as OpenRouter | United States | Anthropic Commercial Terms; data is not used for model training when routed via API |
| OpenAI, L.L.C. | LLM inference and embeddings (routed via OpenRouter) | Same as OpenRouter | United States | OpenAI Data Processing Addendum |
| Resend, Inc. | Transactional email delivery (beta-application notifications, approval/denial, system emails) | Recipient email address, message content as composed by the platform | United States | Resend DPA |
| Stripe, Inc. | Payment processing for paid plans | Billing information (payment method, transaction details) — full card numbers are not stored by us | United States | Stripe DPA |
| n8n GmbH (self-hosted) | Workflow orchestration for the document-processing pipeline | Document processing job state, source-file URLs, extraction outputs in transit | Self-hosted on Contabo VPS, Germany (Frankfurt region) | n8n Community Edition self-hosted; data stays under our operational control |
| Contabo GmbH | VPS infrastructure hosting our n8n instance and self-hosted inference services | Whatever data the hosted services process while running | Germany (Frankfurt region) | Contabo customer agreement |
| Anthropic, PBC (via Claude Code) | Internal development tooling — does not process production user data | None of the production data tier is sent here | United States | Not applicable to user-data flow |
Operator-only services (not handling user data)
The following services are used by us in the operation of the platform but do not process personal data of platform users:
| Service | Role |
|---|---|
| GitHub | Source-code hosting |
| Cloudflare | DNS for archivelm.com |
| Domain registrar | Registration of archivelm.com |
How user data flows through the subprocessor chain
- You upload a document. Image is stored in our Supabase storage bucket isolated to your account.
- The document is queued for processing. Job metadata is written to the Postgres database; n8n picks up the job from a webhook trigger.
- OCR. n8n sends the document image to Google AI / Gemini for extraction. Output is returned to n8n.
- Structuring. Extracted raw text is sent back to Google AI / Gemini (Flash model) for structuring into articles. Output is written to our Postgres database.
- Indexing. Structured articles are sent to OpenRouter (which routes to OpenAI for embeddings) to generate vector embeddings. Embeddings are written to our pgvector store.
- You search or chat. Your query is sent to OpenRouter, which routes to the chosen model (Anthropic Claude or OpenAI GPT-4) along with retrieved excerpts from your collection. The model's response is returned to you and stored in your account.
- Notifications. Transactional emails (beta notifications, approval/denial, etc.) are dispatched via Resend.
- Payments (paid tiers). Billing events are processed by Stripe.
Changes
We update this list when subprocessors are added, removed, or replaced. Material changes will be notified per Section 1 of this document.
Contact
legal@archivelm.com Michael De La Guera (business mailing address available on request — write to legal@archivelm.com)
Version: 1.0