How does Reply extraction work?
Mail clients quote differently: "On … wrote:" lines, > prefixes, Outlook header blocks, forwarded-message separators. Signatures vary as much. Parsing a reply reliably means recognising these markers and stopping at the first one.
For a language model, quoted history is also a security problem: anyone can type a fake "previous message" from your CFO into a reply. Dropping the quote removes that vector.
How does nofeb use Reply extraction?
- Every message carries
reply_text, its new content with quoted history and signature removed, beside the full text and HTML. safe_bodybuilds on it, removing invisible characters and hidden HTML too, andget_thread_contextreturns a whole thread oldest first with safe_body for each message, so the model sees each message's new words once.
Frequently asked questions
How do I get only the new reply text from an inbound email?
Read reply_text on the message, or safe_body if the text is going to a model. Both are computed for every inbound path, including Postmark, Mailgun and SendGrid.
Status of everything on this page
Every capability named above, with its status as reconciled against the engine.
| Capability | What it does | Status |
|---|---|---|
| Reply extraction | Each message carries its new text with quoted history and signature removed. | Shipped |
| safe_body for untrusted mail | The message's new text with zero-width and direction-altering characters, hidden HTML and forgeable quoted history removed, wrapped in a provenance header that tells the model whose words these are. | Shipped |
| LLM-ready thread context | A whole thread, oldest first, with safe_body text and the thread's metadata, in one call. | Shipped |