MVP: docs on withohm.dev · chat edge = local :8081 (or your deploy) · api.withohm.dev returns edge-pending until AWS cutover. Status

Data Processing Addendum

Version: dpa-2026-07-26
Product: Ohm
Contact: partners@withohm.dev

API field dpa_ack: true and tenant dpa_version at key issue bind this version.

Roles

PartyRole
Customer (tenant)Controller of personal data in prompts, tools, and chosen web targets
Ohm (operator)Processor when processing that data to provide the API
Upstream LLM providersSub-processors (and may be independent controllers under their terms)

Subject matter

Processing of Customer Content submitted to POST /v1/chat/completions and related endpoints, including optional public-web excerpts injected as context.

Nature and purpose

  • Route requests to model providers
  • Optional Redis cache for identical-request replay (TTL-bound)
  • Metering aggregates (not full prompts in ledgers)
  • Public-web fetch when Customer enables fetch_web_context

Not in scope: Ohm does not use Customer Content to train foundation models.

Duration

For the subscription term; cache entries expire per configured TTL unless cache_control: no_store.

Categories of data

Determined by Customer. May include identifiers, message content, and URLs. Ohm applies technical minimisation on web ingest (PII redaction, excerpt caps).

Subprocessors

SubprocessorPurpose
OpenAI / Anthropic (as configured)Model inference
AWS (or operator host)Compute, Redis, networking
StripeCustomer billing
Vercel (marketing/docs host)Public documentation site

International transfers

If Customer Content leaves the UK/EEA, the operator documents the transfer mechanism (e.g. SCCs) before EU-scale go-live.

Security

Hashed API keys at rest; tenant isolation by cache key prefix; compliance gates on ingest. See Security.

Customer instructions

Documented via API: purposes, acks, cache_control, and this DPA version.

Acknowledgement

dpa_ack: true binds dpa-2026-07-26.