Sub-processors
These are the companies that process data on Docent’s behalf today. The list is checked against the same configuration the servers run: a test fails if the two disagree. Audio is never stored by Docent. Telnyx’s conversation retention and call recording are switched off; OpenAI processes audio under its API data-use terms.
| Vendor | What it handles | Where |
|---|---|---|
| Cloudflare | Runs our servers (Workers), stores knowledge files (R2), caches per-app configuration (KV), pools database connections (Hyperdrive), and computes the embeddings used to search your knowledge (Workers AI, model @cf/baai/bge-base-en-v1.5). | Global edge network; storage jurisdiction is configurable per account. |
| Neon | Hosts the Postgres database that is the system of record (workspaces, apps, keys, goals, knowledge metadata, masked turns, analytics events) and the Console sign-in service. | United States, us-east-1. |
| Telnyx | Runs the language model that composes the guide’s replies (Telnyx Inference), and is the alternate voice vendor. Conversation retention and call recording are switched off. | United States. Processing location is best-effort; storage residency is an account setting. |
| OpenAI | Runs the voice session (speech recognition, turn-taking and speech synthesis) when the guide is used by voice. This is the production default voice vendor. | United States. |