Documentation
Model endpoints and training policy
Where the language models REAP uses are served, what they see, and the written policy on training.
Endpoint options
| Option | Where the model runs | Typical use |
|---|---|---|
| REAP-managed endpoint | Operated by REAP from a curated, validated model list; the provider and regions are named in the security packet | SaaS and pilots (available today) |
| Customer endpoint | Your enterprise agreement with a model provider or your cloud’s model service supported providers | Customer Cloud deployments with data-flow requirements |
| On-premises models | not offered as standard today | Environments without a path to a hosted model |
Training policy
Written commitment Customer data is not used to train models without the customer’s written consent; data-use terms are stated in the agreement and in the security packet.
What a model sees
- Evidence relevant to the investigation (command output, diffs, ticket text, document passages).
- Prompt logging with the investigation for audit: scope and retention.
- Consequential actions require approval regardless of what the model concludes; retrieved-content handling is described in the security packet.
Last reviewed: October 7, 2026 (draft) · Owner: solutions engineering · Corrections: security@reapnet.ai